Web Technologies phần 8 pot

Extending merely symbolic SWS descriptions with context information on a conceptual level through MSS enables similarity-based matchmaking between real-world situation characteristics an

tation and rule processing (with XML formats,

OWL ontologies and SWRL rule processors)

A future option for the Semantic web, but one

we pursue now in Case Study 1, may be electronic

dashboards to link scientific publications and

electronic medical records to associate disease,

drug compounds, biology and pathway knowledge

between R&D groups As a final concern for HCL-SIG, there is today no widely recognized machine

accessible semantic differentiation between a

manuscript and publication; illustration and

ex-perimental image data; or between an experiment,

data, interpretations, and the hypothesis that an

experiment was designed to validate Initially, our

first web-based study in Case Study 1 addresses

only parts of these problems with an adaptive

electronic Institutional Review Board (eIRB) for

research protocols rather than medical records;

but associated with the eIRB, we are considering

business intelligence for individual organization

and system-wide performance metrics, and

link-ing scientific publications from multiple military

R&D groups to improve patient care

Brief Literature Review

In addition to the literature reviewed in the

back-ground, an additional but brief review is provided

here to place our work in a historical context On

its face, Durkheim’s (1893) “social facts” stand

against Weber’s (1958) methodological

individu-alism, today ingrained in game theory, where

the choices available to those playing games

are influenced by the social and religious norms

existing within a culture (Körding, 2007) As an

example, the choice to cooperate with a partner in

the Prisoner’s Dilemma Game is configured with

a higher value than the choice to defect from a

partner, even though from an information theory

perspective, society often gains significantly more

social benefits from competition than cooperation

(Lawless & Grayson, 2004) While social norms

should not be disparaged but studied, neither

should scientists favor the norm of cooperation by

configuring it with a higher social welfare value, similar to an industrial policy that chooses the winners for its society But there are limitations to Durkehim’s view, too If reality is bi-stable, social facts are open to multiple interpretations Parsons and Luhman contributed to cybernet-ics and control theory as an information approach

to controlling and modeling society Parsons (1966) developed a systems approach as a grand theory of society He used systems as a tool to analyze the function of society, concluding the sys-tems that adapt to their environment had evolved into more efficient systems; however, in that the environment is ever changing, adaptation is not

an optimal control strategy (Conant & Ashby, 1970) Parsons influenced Luhmann’s (1984) theory of autopoietic, or self-organizing, systems Luhman believed that autopoietic systems filtered information from the environment, independently

of other systems, making them autonomous, but also apart from society Elias (1969) contributed

to cybernetics with his ideas on figurational, networked or interconnected structures as the source of power over other systems Crozier and Friedberg (1981) used game structures to explicitly analyze power and strategy between organizations and their members as interdependent actors But the limitations remain for game theory from the influence of social norms and the lack of a theory

by building a constitutional government based

on checks and balances (Hamilton, Madison, & Jay, 1787-1788), concluding that social structure controls and stabilizes government independently

of the social norms in existence Further, not only

do checks and balances recognize the limits of situational awareness, motivated by the search for meaning at the individual level (Carley, 2002); but also, consensus rules and compromise dilute the

added information provided to society by checks

and balances, their strength However,

compro-mise leads to an “action consensus” based on a

concrete plan of action, compared to the unified

worldview of consensus seeking, which reduces

the likelihood of action (Lawless et al., 2008b)

This is not to conclude that Weber’s ideas missed

the mark Just the opposite Weber understood

that the tradeoffs between the incommensurable

beliefs of Confucianism and Puritanism produced

profound differences in the control of and social

welfare benefits for two social systems, which

agrees with the uncertainty relations presented

below

MAIN FOCUS OF THE CHAPTER

In general, most of social science is predicated

on the assumption that observations of behavior,

especially the self-observations made in response

to questionnaires, provide perfect or near perfect

information about a target behavior, thereby

leav-ing no room for an uncertainty principle However,

striking problems exist with asking agents about

the causes of their behavior (self-reports, surveys,

structured interviews, case studies) Baumeister

et al (2005) found that a 30-year meta-analysis

of survey data on self-esteem correlated poorly

with academic and work performance, casting

doubt on one of the most studied phenomena in

psychology and also on the ability of self-reports

to capture targeted phenomena Similarly, in an

attempt to prove the value of air combat

maneu-vering for Air Force educators, Lawless and his

colleagues (2000) found no association between

air combat outcomes (wins-losses) and

examina-tion scores on air-combat knowledge And at the

end of his distinguished career in testing game

matrices, Kelley (1992) found no association

between the preferences as measured by surveys

before games were played and the choices actually

made during games Along the same line, Axsom

and Lawless (1992) found that scientists easily

misinterpreted the causes of behavior measured

in effort justification experiments designed to reduce public speaking anxiety even when the scientists observed the changes directly

In their review of decision theory, Shafir and LeBoef (2002) concluded that justifications of actions already taken were not consistent with the actions taken, including for expert judges In addition, they found that the widely held belief

by theoreticians that expectations of well-being lead to well-being was systematically violated, even for experts But even though the evidence in support of widespread claims based on self-reports does not exist, many social models continue to endorse the belief that cooperation enhances the value of social welfare more than competition

In agreement with Pfeffer and Fong, the lack of fundamentals has produced a subjective social science In response, we take a more theoretical approach based on the impact that cooperation and dissonance have on the diminution or generation

of information (Lawless & Grayson, 2004)

To summarize, metrics must not interfere with the process of measurement; doing so collapses interdependence and invokes the organizational uncertainty principle (e.g, surveys of self-esteem

at the individual level by Baumeister et al., 2004; decision-making at the organizational level; Law-less & Grayson, 2004) Perceptions are integral to behavior, as the Coca-Cola Company discovered when it decided to close out its traditional Coca-Cola brand due to its inability to best Pepsi-Cola

in internal taste tests (en.wikipedia.org/wiki/New_Coke) But following considerable public criticism, the firm brought back its traditional cola and re-branded it “Classic Coke” As Baumeister has re-discovered, the measurement of percep-tions in interdependent states with behavior collapses the interdependence, producing static information

We plan to study organizations with tational models However, Bankes (2002) and Conzelmann and his colleagues (2004) have both concluded that current computational models

compu-of organizations are not predictive, principally

with Agent-Based Models (ABMs) We plan

two correctives: first, to test models using social

configurations addressed by our organizational

uncertainty model to reproduce the results of

col-lapsed interdependent states that we have predicted

In contrast to traditional social science, we have

attempted to combine individuals with

organiza-tions and systems, statics with dynamics, and

empirical approaches with theory We incorporate

dynamics in our model with the effects of

feed-back on oscillations within an organization, but

as a metric for its performance We incorporate

organizations in our model by introducing control

as organizations seek to improve in performing or

revising their mission (Mattick & Gagen, 2005;

also, May, 1973) Finally, in our approach, an

empirical approach alone precludes formal

ap-proaches and optimal solutions; our immediate

goal, then, is to build and be guided by theory

and empirical findings

To implement control theory (Csete & Doyle,

2002), we need to quantify an organizational or

system's level model In line with earlier

argu-ments, an organization controls at least four

aspects of the decision-making process First, by

helping to set or choose its reference or threshold

set-points (e.g., culture, decision processes,

expec-tations, planning; and in Case Study 1, mission

and vision) Second, by damping unexpected

disturbances Third, by filtering and transform-ing incomdisturbances Third, by filtering and transform-ing information about system internal

states, inputs, and responses to form patterns

and decisions Finally, by taking actions then

collecting feedback to revise decisions However,

Conant and Ashby (1970) concluded that feedback

to minimize errors is not an optimal solution for

control, that the optimum solution avoided errors (e.g, with a plan that produces the most efficient operation possible)

As metrics for our control theory, we have proposed inverting the organizational uncer-tainty principle in Figure 2 to link uncertainty between planning and execution as well as be-tween resource availability and the duration of plan execution

In Figure 2, uncertainty in the social

interac-tion is represented by an interdependence between

business models, strategy, plans, or knowledge

un-certainty (∆BM x , where the knowledge or business

model is a function of the social location where it

was learned; from Latané, 1981 and Sukthankar,

2008) and uncertainty in the rate of change in

knowledge or its execution as ∆v = ∆ (∆BM/∆t)

This relationship agrees with Levine and land (2004) that as consensus for a concrete plan

More-increases (∆BM x reduces), the ability to execute

the plan increases (∆v increases) By extension,

Uncertainty in the execution of a plan

Uncertainty in resources to execute a plan

ΔvΔBM x ≈ c ≈ ΔRΔt

Uncertainty in plans, worldview beliefs or Business Models

Uncertainty in time

to execute a plan

Figure 2 Measurement problem

The measurement problem occurs as the result of the nizational uncertainty principle The measurement problem arises from the interdependence between the two factors

orga-on each side of the equatiorga-on It states that both factors

on either side of the equation cannot simultaneously be known exactly For example, a decrease in the uncertainty

in the strategy for an organization results in an increase in uncertainty for the execution of that strategy In practice, decreasing strategic uncertainty increases action; increas- ing strategic uncertainty slows action (Busemeyer, 2008)

At the same time, the uncertainty principle informs us that only one of the factors on either side of the equation can

be known with certainty (Lawless et al., 2007).

interdependence also exists in the uncertainty in

the resources expended to gain knowledge, ∆R,

and by uncertainty in the time it takes to enact

knowledge, ∆t That these two sets of bistable

fac-tors are interdependent means that a simultaneous

exact knowledge of the two factors in either set is

precluded, due to a collapse of interdependence

However, a partial or proportional collapse is not

ruled out (i.e., tradeoffs)

We have used the model in Figure 2 to study

human organizations making decisions under

uncertainty by addressing complex situations like

the environmental cleanup of its nuclear facilities

undertaken by the Department of Energy, or

merg-ers and acquisitions The primary characteristic

of this interdependence is reflected in tradeoffs

between coordinating social objects

communicat-ing to solve problems while in states of uncertainty

(Lawless & Grayson, 2004) In Case Study 1, we

apply Organizational Uncertainty theory to a

system of seven MDRCs (Medical Department

Research Training Center) Our goal is to help

those MDRCs become more productive in

meet-ing their assigned mission This means that the

MDRC system would shift from a fragmented to

a more ordered group of organizations, thereby

increasing productivity In the future, to exploit

the power of the semantic web, we propose to

use a rate equation to measure in real-time with

machines the system performance, thus offering

management insight as to the factors to change

in a tradeoff that enhances organizational

per-formance

In addition, we have proposed that alignment

of humans and thinking machines (agents) in an

organization ranges from disordered in the lowest

energy or individual state to one focused on the

mission (Lawless et al., 2007) But, by focusing

on the mission exclusively as in the latter case,

organizations become vulnerable to change

Therefore, it is important to use feedback not only

to fine tune an organization's effectiveness over

the short term, but to restructure by revising its

mission over the long term (Smith & Tushman,

2005) We propose that the tension can be best constructed, maintained and controlled over time

by using semantic web-based metrics

uncer-to support DOE’s plans uncer-to speed the shipments

of transuranic wastes to its repository in New Mexico (i.e., the WIPP facility; see www.wipp.energy.gov ) as part of its mission to accelerate the cleanup of DOE facilities across the U.S These nine DOE Boards were geographically separated and located at the DOE sites where the transuranic wastes were being removed and shipped to WIPP DOE’s plans were entailed in 13 concrete recom-mendations and explained to the various Boards

by DOE engineers (e.g., recommendation #8:

“DOE in consultation with stakeholders and lators initiate action to assure that WIPP has the capacity to accommodate all of the above listed TRU waste”) As predicted, four-fifths of DOE’s majority-rule boards endorsed these recommenda-tions, while three-fourths of its consensus-ruled boards rejected them In addition, the time spent

regu-in decidregu-ing for majority-ruled boards was about

one-fourth of the amount of time taken by the

consensus-ruled boards

In a follow-on field study of consensus deci-sions by the Hanford Board in Washington State

and majority rule decisions at the Savannah River

Site Board in South Carolina, Boards located at the

two DOE sites with the largest cleanup budgets,

we found that consensus rule decisions produced

a cognitive congestion that resulted in behavioral

“gridlock” when the single worldview of the Board

conflicted with DOE’s vision, increasing social

volatility (Lawless & Whitton, 2007) We have

found that cognitive congestion is more likely

under cooperative decision making because of the

inability to accept challenges to illusions

(Law-less et al., 2008b) In contrast, we have found that

the cognitive disambiguation from competition

improves social welfare with practical decisions

that feedback amplifies or accelerates

Relative to the SRS-CAB, Bradbury and her

colleagues (2003) analyzed interviews and other

self-reported measures to conclude that Hanford

CAB members felt very positive about their

consensus-seeking process, that they very much

wanted a cleaned-up environment, and they felt

that DOE at its Hanford site was very responsive

to their demands However, the results from field

metrics at DOE Hanford and DOE Savannah River

Site (SRS) across three measures of cleanup

(high-level radioactive wastes, transuranic wastes, and

the environmental remediation of contaminated

sites) indicated the opposite (e.g., Lawless et al.,

2005) Compared to the SRS CAB and the SRS site,

this difference between perceptions at the Hanford

CAB and the results in the field represented an

increase in risk perceptions (i.e, an unchecked

increase in the number of illusions) among the

Hanford CAB members that had kept them from

making concrete recommendations to accelerate

the environmental cleanup at Hanford

Evidence: Laboratory

Preliminary data from a laboratory experiment nearing completion with college students making recommendations to improve their college experi-ences appears to have fully replicated the DOE CAB study In this study, we asked college students in

3-person groups (N = 53 groups) at a Historically

Black College and a nearby University to proposed open-ended recommendations to improve op-erations affecting them at their schools (e.g., with cafeteria food, library, student government, etc.) Students were randomly assigned to three-person groups who made recommendations either under consensus (CR) or majority rules (MR) Time for both types of groups was held constant Tentatively,

we predicted and found that while CR produces significantly more discussions (oscillations or

jω), indicating less time available to craft

recom-mendations, MR produces significantly more total recommendations (our analyses are ongoing)

Evidence: Case Study 1: Military Medical Department Research Training Centers (MDRCs)

plying the organizational uncertainty principle,

Guided by our theoretical and field results in ap-we have been assisting a system of seven military MDRCs (Wood et al., 2008) to become more pro-ductive; e.g., produce more research with greater scientific impact; improve patient care; and reduce the costs of care Specifically, when we began this case study, we found little knowledge existed at the organizational level that directly linked each research product (publications, presentations, workshops) with MDRCs assigned mission In-stead, MDRC collected basic citations for each publication; not all publications were captured in its data-base; nor were all conferences attended captured in their data base

We began with a preliminary set of metrics that

indicated the efficiency in meeting MDRCs mission

per research protocol across the factors of scholarly

activity, personnel availability, space, and funding

But at the same time, these Centers wanted to be

able to transform their mission as necessary These

two goals are contradictory But Smith and

Tush-man (2005) concluded that satisfying contradictory

goals like these could make an organization more

productive now, and more transformative in the

future (see Figure 3)

Based on feedback from metrics of

organiza-tional performance linked to eIRB's, administrators

have the ability to execute their mission effectively

and efficiently; e.g., with Lean Six Sigma

pro-cesses But efficiency alone reduces adaptability

to uncertain future missions (Smith & Tushman,

2005) Thus, concomitantly, a group internal to

each MDRC and a national group of elite

profes-sionals from all MDRC units could gather annually

to transform its mission, goals, and rules guided

by the same feedback and metrics As these two systems compete in a bistable relationship to control the Mission, the two systems operate in tension (interdependence), producing a natural evolution

in the product evaluation selection process in the hope that the benefits of the funded eIRB will

Mission Tradeoffs:

1 Well-crafted mission supported by consensus-seeking versus consensus- action Includes procedures, rules &

metrics; e.g., fragmentation p innovation, impedes consensus.

2 Execution

Vision Tradeoffs:

1 Vision transforms mission New vision & mission are constructed by consensus-seeking versus consensus- action

2 Top professionals at each MDRC propose vision and mission revisions based on mission demands &

outcomes

3 National meetings held to debate proposals HQ adopts and publishes best, integrated proposal(s)

4 Revisions voted on by all MDRC,

HQ & MRG professionals every 3 years versus Command Decision- Makin

4 Timeliness with execu bringing assets to bear versus duration

g (CDM).

Negative Feedback Positive Feedback

Metrics

Feedback =

= Planned Mission - Actual Mission

= Actual Mission - Mission Vision

Figure 3 Future proposal for a semantic-web based system of seven MDRCs

The initial guidance based on theory were: Mission success makes a lean organization more efficient but also more able to change; change in a business model or its execution in reaction to environmental change was not optimum (Conant

vulner-& Ashby, 1970); and a sweet spot exists where mission performance is optimum, errors are at a minimum, and at the same time the mission and the organizations it guides are modernized

includes routing of submissions to IRB members;

receipt of comments from IRB reviewers;

trans-mission of modification requests to investigators;

development of IRB meeting minutes; tracking of

protocol status; automatic notification of investiga-tors of continuing review deadlines; and tracking

metrics The technology provides a platform for

collaboration across the organization between

Principal Investigators and team members when

drafting protocol proposals It provides feedback

among IRB reviewers, the PI and study team, and

Administrators It tracks Adverse Events (medical

and drugs); provides guided electronic input and

assistance and error checking and reporting to

PI’s and Administrators; but more importantly,

it is a platform for integrated management and

reporting

The vision for this eIRB project is to achieve

an end state to:

allow all research proposals, supporting

docu-ments, and scholarly products to be submitted and

managed by a secured web based electronic system

that allows for the real time calculation of research

metrics of workload, productivity and quality

Ad-ditionally, this kind of system will allow for better

management of the necessary documentation for

human research protection and ensure a better

environment of operational security oversight for

potentially actionable medical information This

will be developed with joint execution in mind

and have input from our DoD counterparts A

system that effectively captures all aspects of the

research process, from protocol submission and

processing to publication of scholarly products

or novel therapeutics will generate the highest

quality data for productivity analysis and metric

development We believe this can best be achieved

by development of an electronic protocol

submis-sion and management system with the capacity

to generate real time metrics of productivity and

quality (Wood, 2007, pp 4-5)

In installing the eIRB, MDRC will be better positioned to leverage business intelligence (BI) tools that automatically pull together data for metrics with machines from this new electronic system and from other disparate database systems already in place (e.g., electronic medical records) However, only until MDRC has database systems across all aspects of biomedical research and medical care delivery and the BI tools to link these often incongruent systems together will it be able to generate real time data for semantic-web machines to study, define and improve their pro-cesses Once in place, MDRC can make decisions

in real-time rather than with data many months old thereby closing the gap between the mission and the vision and pushing the organization faster towards innovation The natural tension and gap between the mission and vision, as it closes, will decrease the cycle time between these two per-spectives propelling MDRC along the pathway of necessary transformation We believe the ability

to quickly and effectively manage knowledge is the key to organizational change

Knowledge management is one of the est growing sectors in the business community

fast-In parallel with the rapid growth of knowledge generated by automation systems, organizations having the capability and diversity of BI tools

to analyze their performance against their own chosen metrics should help to accelerate system-wide transformation These tools can afford a seamless reach across different platforms to easily allow for the automatic generation of dashboards that can visually depict metrics of organizational importance in a manner not previously available

As the present web evolves into the Semantic web,

so will the capability of knowledge management with BI tools

Current Status A case in point to demonstrate

the power of web-based technology and edge management has been the virtual collabora-tion systems used by the MDRC working group planning for an eIRB Leaders geographically separated were able to meet approximately thirty

knowl-times over almost two years and work together to

solve common problems in a manner that would

have been cost-prohibitive in the past MDRC

leaders from Hawaii, Washington State, Texas,

Washington DC, and Georgia worked as a

net-worked virtual organization for approximately 60

hours using web-based collaboration technology

with visual and audio communication that lead

ultimately to the successful funding of the eIRB

system (for a review of Networked and Virtual

Or-ganizations, see Lawless et al., 2008a) Members

simply logged onto the web from the convenience

of their own office to participate in problem solv-ing and closof their own office to participate in problem solv-ing the gap of tension between their

mission and vision Using this virtual

collabora-tion in conjunccollabora-tion with a mind-mapping program

(similar to a semantic network) for more effective

brainstorming allowed the saving of thousands of

dollars in travel and personnel time

Assessment of Case Study One We began

Case Study 1 by contrasting the organizational

performance of MDRC against the specifics listed

in its assigned mission: improving patient care in

the field; reducing the costs of care; and increasing

the impact of research products We found no clear

link between research products and the mission; no

measure of publication impacts; and no direct way

to measure organizational productivity against its

peers (reduced or negligible states of

interdepen-dence) In general, the organizations in the MDRC

network appeared to be fragmented, with each

pursuing its own path to mission success No

over-arching measure of system performance existed

for the MDRCs that the separate organizations

could follow to guide their collective behavior

As a consequence, long-term work practices and

cultural differences predominated Subsequently,

the move to adopt a web-based eIRB has set the

stage to turn around the lack of organizational and

system-wide knowledge MDRC is prepared for

real-time organizational and system-wide based

metrics, improvements and future transformations

(based on maintaining interdependent states) We

believe that the semantic web can enhance these

metrics by operating in real-time with data lected by machines to distinguish between classes

col-of data sources (using OWL’s vocabulary to label separately a site’s physician students, physician scientists and medical scientists across the dif-ferent sites, etc.) At the same time, we will be diligent in preventing web machines from either the inadvertent disclosure of patient records or the premature release, identity and location of researcher data

Evidence: Case Study 2: Application

of Theoretical Model to a College

After developing and applying metrics for a government organization whose primary mis-sion is training military physicians in medical research practices, it was helpful to apply similar web-based metrics to an organization with a very different purpose The subject of Case Study 2 is

an organization whose primary function is higher education Although all institutions of higher education are tasked with the production of new knowledge within fields where it offers degrees, this organization’s primary purpose is to train the next generation of citizens through the use of a liberal arts curriculum In its Vision statement, technology is highlighted and indicates that the institution “provides information technologies that link its students’ total academic and social experi-ences to the global world.” (Bradley, 2008)Today’s institutions of higher education are faced with an interesting dilemma with faculty members who have come of age during a period

of tremendous technological upheaval During the last twenty years, institutions of higher educa-tion have started making significant investments

in administrative information systems Higher education institutions are being asked by policy makers, accrediting bodies, the business com-munity and the public for evidence that college graduates have a demonstrated knowledge base predicated on their degree With the mounting cost of higher education, consumers are asking

for accountability from colleges and universities

(Bradley, 2008) Institutions of higher education

as well as most organizations must focus on

systems that must be in place to ensure that the

decisions made in the future take advantage of

the best data possible

Institutions are engaged in a delicate dance

of remaining true to their purpose in society

while responding to calls for accountability for

their actions Laws such as the Family Education

Rights and Privacy Act (FERPA) caused some

campus officials to develop extremely strict poli-cies regarding information about student records

These policies were strictly enforced even when

it was known that the aggregation and analysis

of data from student records would provide the

institution with invaluable information for

mak-ing informed decisions about ways to improve

academic programs, increase retention, and

address issues being raised by outside entities

Institutional research projects were strangled

by the fear of litigation regarding the privacy of

student information (Green, 2007)

According to Green (2007) “institutions of

higher education have seen an emergence of

a wide, rich, and mission-critical array of

stu-dent and institutional services that are directly

linked to core campus information services (or

Enterprise Resource Planning (ERP) functions)

Yet these new functions and servicesalumni

services, course/learning management systems,

digital content, e-portfolios, e-services (online

registration, fee payment), and portalsare all

firmly dependent not only on the Web but also

on real-time interaction with the core elements of

the “old” management information system (MIS),

particularly students records and institutional

finances.” Many of these functions at institutions,

particularly small institutions are informal and

units within the organization form their own

fiefdoms many times as a way of managing the

complexity of a system that is governed by external

policies and procedures as well as the end users

of the services In an earlier age when students

walked from one office to another to engage personnel in the business of enrolling in courses, acquiring financial aid, paying their bills, and ob-taining housing, these systems worked However,

in an age where information drives decisions for the organization as well as the consumer, the earlier model is no longer feasible

The organization employs approximately

200 individuals with the majority of als serving as instructional personnel providing instruction for a student body of less than 1,000 individuals studying at the undergraduate level Besides instructional staff (faculty), there are administrative staff members, staff who provide support services to students, a unit that manages the fiscal enterprise of the organization, and a unit responsible for external partnerships and fund raising All units of the institution rely on the efficient function of the other areas but are limited in the operational knowledge generated

individu-by these other units from the lack of technological (web-based) interconnectivity

Current StatusComputing and technology support in an academic environment provide the technology infrastruc-ture for academic and administrative activities that have become essential for the operational effectiveness of institutions of higher education There is a need to analyze the current technol-ogy infrastructure due to the present isolation between subsystems and organizational opera-tions Multiple systems exist but each organiza-tional unit works with its own “preferred” one, producing fragmentation The different systems are not integrated causing record sharing and management problems Currently, information technology (IT) support is being done by two staff members, one deals with hardware issues and the other with support software plus the network as part of the college’s infrastructure There is no system request form or work-list Priority is given

to network issues and calls from very important

persons (VIPs) within the organization, likely

impeding performance

With a new administration, this organization

has realized the need to evaluate the current

IT infrastructure and the need for changes to

fulfill its vision and mission After the

pre-liminary investigation, the first need identified

was to overhaul and redesign the website The

previous version did not represent the academic

organization due to its commercial feel Then

an IT inventory survey was conducted to find

out what systems are available, which system is

being utilized by which unit (or not at all), the

merit of these choices, and the costs associated

with each system To find an enterprise-wide

solution, the institution is considering having an

IT-consultant company to evaluate the current

infrastructure (conceptual model), and suggest

the best solution The institution also needs a

chief information officer (CIO) (or MIS director)

who is capable of implementing the plan

All institutional areas that are impacted by or

use technology should be evaluated Either after

purchasing an enterprise information system (EIS)

or after choosing from currently available systems

for a single “main” system that supports most unit

functions plus a Transaction Process System (TPS)

for business/financial unit online transactions, per-formance measurement should be enacted Focus,

however, would not be placed on the network per se,

but on the organization’s performance as measured

with its EIS Critical Success Factors for an EIS

in a higher education institute like this one which

should be measured are:

• Instructional support, as measured by the

number of courses offered or supported via

the Internet or other electronic methods,

num-ber of instructional classrooms supported,

number of student computer labs, student

accounts, technology in residence halls and

shared spaces (i.e campus center) or other

means

• Research support, as measured by access to research databases, high speed network con-nectivity, other data collection and analysis mechanisms, and specialized research func-tions

• gate, or on a per-student full-time equivalent (FTE) or per-faculty FTE basis, including comparisons with peer institutions

Cost of services, either measured in the aggre-• sion

Number of graduates compared with admis-• port

Student learning outcomes: assessments sup-Assessment of Case Study TwoWhile it is too early in the process to assess this college, and while a measurable semantic-web based baseline is being built, certain areas to measure performance are already obvious For example, after implementation of the EIS, do faculty publication numbers and the impact of research (quantity and quality) improve? Does the new IT web system improve or assist the College

in its assessment processes? After the EIS system

is operational, we plan to review its performance

as well as the College’s

FUTURE TRENDS

The most important future trend is the use of based models (ABM’s) to model social and orga-nizational effects to measure their effectiveness with the semantic web Agent based systems have been endowed with auction based mechanisms for distributing their resources (Gibson and Troxell, 2007) In this scenario, the various entities would

agent-“bid” for the information they require, ensuring that the entity that valued the information the most would receive it in the timeliest manner for their decision making Double auctions have been used for similar analyses with genetic algorithms (Choi, Ahn and Han, 2008)

Working with mathematics, ABM’s, and

arti-ficial intelligence, the organizational uncertainty

principle can be generalized to interdependent

probability distribution functions with the

stan-dard deviation of Fourier transform pairs (i.e., the

standard deviations from a Gaussian distribution

and its Fourier transform form a “Fourier pair”; in

Cohen, 1995; Rieffel, 2007) Next, we construct

circuits to provide a basic model of social decision

making (Yu & Efstathiou, 2002) Circuits can be

modeled using virtual natural selection processes

(e.g., machine learning, natural computation)

Rate equations would then provide a detailed

prediction of outcomes that we plan to estimate

with Monte Carlo simulations Completing the

process, sensitivity analyses with the rate

equa-tion parameters provides a direct link back to the

organizational uncertainty principle

Circuits

Based on entropy measures, Yu and Efstathiou

(2002) found that series network circuits

under-performed parallel circuits We expect to find

that group decision-making, especially around a

table, is similar to a series circuit, with subgroups

or subcommittees acting like parallel circuits

However, we also expect that consensus rules (CR)

will be serial and sequential, producing the time

lags observed in the field and laboratory, but that

majority rules (MR) with discussion drivers will

act like a parallelization of neutrals, producing

the time speedup also observed

Natural Computation

Natural computation models will permit us to test

field data and model the organizations that produce

this data, especially the MDRC system in Case

Study 1 and later the college in Case Study 2 We

propose to test the data and organizational models

with artificial agents evolved using biologically

inspired natural selection (De Jong, 2008) and

social methods of decision-making (e.g

“vot-ing” mechanisms, ensembles) Based on our field research, we predict longer decision times and more oscillations under consensus rule (CR) than majority rule (MR) That is, we expect CR to model serial sequential individual decision processes Surowiecki (2004) presented evidence and case studies of why agent ensembles often outperform individual experts Earlier, Opitz and Maclin (1999) empirically showed that ensembles often outperform individuals, with theoretical support provided by Brown (2005) and Tang (2006)

Monte Carlo Simulations

Monte Carlo simulation is a technique that lows the simultaneous iteration of many uncer-tain variables to understand the impact of input uncertainties on one or more outcome variables Developed during the 1940s as part of the Manhat-tan Project, and named after the famous casino in Monaco, Monte Carlo techniques are used today

al-in fields ranging from manufacturing to finance, engineering and life sciences

The basic concept in Monte Carlo simulation

is that each uncertain variable, which we call a random variable, is simulated by a probability distribution For each trial of a simulation, each random variable is sampled from its correspond-ing probability distribution and the sampled value

is used to compute the output variable(s) for the model Many such trials are conducted and a value

is collected for each outcome variable for each simulation trial At the conclusion of all trials a distribution of outcomes can be constructed to better understand the distribution of uncertain-ties for an outcome given the uncertainties in the input variables

Rate Equation

Lawless and his colleagues (2007) devised a mathematical model of social interaction rates (this approach will allow future tests of this model constructed with machine learning using

recombination operators; De Jong, 2008) We

propose to adapt this model to guide our future

research on organizations, e.g., training MDRC

physicians with the experimental method or

edu-cating students unprepared for college courses

with enhancement classes In the latter case, the

model becomes,

Γ = N 1 N 2 v 12 σ 12 exp (-∆A/<A>), (1)

where Γ is the college graduation rate; N 1 the

popu-lation in the group of those who have learned; N 2

those in the population who have not yet learned;

v 12 represents the velocity of knowledge passed

between them, with the minimum effect occurring

under censorship; σ12 represents how well the two

groups match their beliefs, with the maximum

ef-fect occurring under social agreement (resonance);

and exp (-∆A/<A>) represents the probability of

graduation or knowledge exchanges, where ∆A

represents the energy or effort required for the

knowledge to be acquired, and <A> represents

the average amount of effort being expended by

the targeted HBCU, its professors and support

staff, and its fellow students Before we address

the implications of equation (1), let’s rearrange

it If χ represents the knowledge required before

a student can be declared to become a graduate,

then Γ =∂χ/∂t ≈ ∆χ/∆t, and

∆χ = ∆t N 1 N 2 v 12 σ 12 exp (-∆A/<A>) (2)

From equation (2), given an average time

to matriculate from the target HBCU, various

opportunities exist as tradeoffs for it as an

orga-nization to improve the probability that one of

its students will graduate (∆χ) from this college

Increasing the numbers of those who actively

support the student increases the occurrence of

teacher-support group (N 1 ) to student (N 2) speech

acts Increasing the velocity (v 12) of knowledge

passed between the two groups improves the

acquisition of knowledge Increasing the match

(σ12) between faculty-support groups and student

groups can dramatically increase the knowledge

gained (e.g., study groups; student focus groups;

faculty-student focus groups; enhancement

groups) But also the probability of graduation can be increased by reducing barriers for students

(-∆A; e.g., either lowering requirements, choosing better qualified entrants, or enhancing the skills

of the weaker entrants) Finally, by increasing the overall average effort or excitement by the HBCU

directed toward learning and graduation (<A>),

a college can strongly improve the odds that its students will graduate Inversely, changing these factors can also decrease or adversely increase the time required for graduation

Using the equations that we have laid out, with machines automatically collecting the data over the semantic web, we believe that real-time metrics will become possible This information will not only be able to inform colleges or MDRCs whether they are on-target to achieve their mission

as they themselves have defined it, but whether they are making progress evolving into the vi-sion that they themselves have also proposed With machine readable data feeding real-time metrics, organizations like MDRC will also be able to tune their performance For the first time,

we will know the actual cost of controlling their organizations to realize their benefits

CONCLUSION

A preliminary web-based metric modeled after the plans for the new semantic web Health Care and Life Sciences Interest Group (HCL SIG) using electronic spreadsheets indicates that researcher protocol effectiveness can be established and measured as part of an organization’s mission

In the metric, for theoretical reasons we have chosen the interdependent factors of planning-execution and resources-timing As a result, the organizational uncertainty principle has proven to

be a fertile source for theory and a tool to guide a system of military units in the field as they move into a new web-based collaboration system, and

for a college as it begins to establish a web-based

EIS system with real-time metrics Future trends

and our next steps along the path forward with

natural computation, Monte Carlo simulation and

Agent-Based Models (ABM’s) were also reviewed

Finally, we will assure that semantic web machines

do not inadvertently disclose patient records nor

prematurely release data from researchers

REFERENCES

Axelrod, R (1984) The evolution of cooperation

New York: Basic

Axsom, D., & Lawless, W F (1992) Subsequent

behavior can erase evidence of

dissonance-induced attitude change Journal of Experimental

Social Psychology, 28, 387-400.

Bandura, A (1989) Human agency in social

cognitive theory American Psychologist, 44(9),

1175-1184

Bandura, A (1977) Social learning theory New

York: General Learning Press

Bankes, S C (2002) Perspective Agent-based

modeling In Proceedings of the National Academy

of Sciences, 99(3), 7199-7200.

Baumeister, R F., Campbell, J D., Krueger, J

I., & Vohs, K D (2005, January) Exploding the

self-esteem myth Scientific American.

Berners-Lee, T (2007, March 27) Video: Tim

Berners-Lee on the Semantic Web Retrieved

February 20, 2008, from

http://www.technolo-gyreview.com/Infotech/18451/

Bradley, G (2008) Confidential source:

Presen-tation to the faculty by the new president of the

subject college while articulating that building

technology infrastructure must be one of the

goals for the college.

Bohr, N (1955) Science and the unity of

knowl-edge In L Leary (Ed.), The unity of knowledge

(pp 44-62) New York: Doubleday

Brown, G., Wyatt, J., Harris, R., & Yao, X (2005) Diversity creation methods: A survey

and categorization Journal of Information sion, 6, 5-20.

Fu-Busemeyer, J (2008) Second quantum interaction symposium (2008) In P Bruza, W F Lawless, K

von Rijsbergen, D Sofge, B Coecke, & S Clark (Eds.), Oxford, UK: Oxford University

Cacioppo, J T., Berntson, G G., & Crites, S L., Jr., (Eds.) (1996) Social neuroscience: Principles,

psychophysiology, arousal and response In Social psychology handbook of basic principles New

Coase, R (1937) The nature of the firm Eco-Cohen, L (1995) Time-frequency analysis: Theory and applications Upper Saddle River,

NJ: Prentice Hall Signal Processing Series.Conant, R C., & Ashby, W R (1970) Every good regulator of a system must be a model of

that system International Journal of Systems

Science, 1(2), 899-97.

Conzelmann, G., Boyd, R., Cirillo, V., Koritarov, C., Macal, M., North, P., Thimmapuram, R & Veselka, T (2004) Analyzing the potential for market power using an agent-based modeling ap-proach: Results of a detailed U.S power market

simulation In Proceedings of the International Conference on Computing, Communication and Control Technologies, Austin, TX.

Crozier, M., & Friedberg, E (1981) Actors and systems (l’acteur et le système) Chicago: Chicago

University Press

Csete, M E., & Doyle, J C (2002) Reverse

engineering of biological complexity Science,

295, 1664-69.

De Jong, K A (2008, February) Evolving

intel-ligent agents: A 50 year quest Computational

Intelligence Magazine, 3(1), 12-17.

Durkheim, E (1893/1997) The division of labor

in society The Free Press

Elias, N (1969/2000) The civilizing process

(Üder den prozess der zivilisation) Oxford, UK:

Blackwell

Green, K C (2007) Prodding the ERP turtle

EDUCAUSE Review, 148-149.

Kelley, H H (1992) Lewin, situations, and

inter-dependence J Social Issues, 47, 211-233.

Körding, K (2007) Decision theory: What

“should” the nervous system do? Science, 318,

606-10

Latane, B (1981) The psychology of social impact

American Psychologist, 36, 343-356.

Lawless, W F., Castelao, T., & Ballas, J A (2000)

Virtual knowledge: Bistable reality and solution

of ill-defined problems IEEE Systems, Man, &

Cybernetics, 30(1), 119-124

Lawless, W F., & Grayson, J M (2004) A

quantum perturbation model (QPM) of

knowl-edge and organizational mergers In L van Elst

& V Dignum (Eds.), Agent mediated knowledge

management (pp 143-161) Berlin, Germany:

Springer

Lawless, W F., Bergman, M., & Feltovich, N

(2005) Consensus-seeking versus truth-seeking

ASCE Practice Periodical of Hazardous, Toxic,

and Radioactive Waste Management, 9(1),

59-70

Lawless, W F., Bergman, M., & Feltovich, N

(2006) The physics of organizational uncertainty:

Perturbations, measurement and computational

agents In S H Chen, L Jain, & C C Tai (Eds.),

Computational economics: A perspective from

computational intelligence (pp 286-298)

Her-shey, PA: Idea Group Publishing

Lawless, W F., Bergman, M., Louçã, J., Kriegel,

N N., & Feltovich, N (2007) A quantum metric

of organizational performance: Terrorism and

counterterrorism Computational & cal Organizational Theory, 13, 241-281.

Mathemati-Lawless, W F., Howard, C R., & Kriegel, N N (2008a) A quantum real-time metric for NVO’s

In G D Putnik & M M Cunha (Eds.), pedia of networked and virtual organizations

Encyclo-Hershey, PA: Information Science Reference.Lawless, W F., Whitton, J., & Poppeliers, C (2008b) Case studies from the UK and US of stakeholder decision-making on radioactive

waste management ASCE Practice Periodical

of Hazardous, Toxic, and Radioactive Waste Management, 12(2), 70-78.

Levine, J M., & Moreland, R L (1998) Small groups In D T Gilbert, S T Fiske, & G Lindzey

(Eds.), Handbook of social psychology (pp

415-469) Boston, MA: McGraw-Hill Luhmank, N

(1984), Soziale systeme: Grundriß einer meinen theorie, Frankfurt: Suhrkamp

allge-Hamilton, A., Madison, J., & Jay, J (1787-1788)

The federalist papers New York newspapers.

Mattick, J S., & Gagen, M J (2005) Accelerating

networks Science, 307, 856-8.

May, R M (1973/2001) Stability and complexity

in model ecosystems Princeton, NJ: Princeton

University Press

Metropolis, N (1987) The beginning of the Monte

Carlo method Los Alamos Science, Special sue, 125-130.

Is-Montesquieu, C.-L., II (1949) The spirit of the laws (T Nugent trans.) New York: MacMillan

Opitz, D., & Maclin, R (1999) Popular ensemble

methods: An empirical study Journal of Artificial Intelligence Research, 11, 169-198.

Parsons, T (1966) Societies: Evolutionary and

comparative perspectives Englewood-Cliffs, NJ:

Prentice Hall

Pfeffer, J., & Fong, C T (2005) Building

organiza-

tion theory from first principles: The self-enhance-

ment motive and understanding power and influ-ence Organizatonal Science, 16(4), 372-388.

Rieffel, E G (2007) Certainty and uncertainty

in quantum information processing In

Proceed-ings of the Quantum Interaction: AAAI Spring

Symposium, Stanford University AAAI Press.

Sanfey, A G (2007) Social decision-making:

Insights from game theory and neuroscience

Science, 318, 598-602.

Smith, W K., & Tushman, M L (2005) Managing

strategic contradictions: A top management model

for managing innovation streams Organizational

Science, 16(5), 522-536.

Shafir, E., & LeBoeuf, R A (2002) Rationality

Annual Review of Psychology, 53, 491-517.

Sukthankar, G (2008, June 10) Robust and eficient

plan recognition for dynamic multi-agent teams

Presentation to Information Technology Division,

Naval Research Laboratory, Washington, DC

Surowiecki, J (2004) The wisdom of crowds:

Why the many are smarter than the few and how

collective wisdom shapes business, economies,

societies and nations Little, Brown, & Co

Tang, E K., Suganthan, P N., & Yao, X (2006)

An analysis of diversity measures Machine

Learning, 65, 247-271.

Troxell, W., & Gerald, G (2007) U.S Patent No

20,070,124,026 Washington, DC: U S Patent and

Trademark Office

Von Neumann, J., & Morgenstern, O (1953)

Theory of games and economic behavior

Princ-eton, NJ: Princeton University Press

W3C-SWHCLS (2008) W3C Semantic Web health care and life sciences interest group

Retrieved February 20, 2008, from http://www.w3.org/2001/sw/hcls/

Weber, M (1904-5/1930) The protestant ethic and the spirit of capitalism (Die protestantische ethik und der ‘geist’ des kapitalismus) (T Parsons

trans.) Scribner

Winston, W L., & Albright, S C (2007) Practical management science (3rd ed.) Thomson Press

Wood, J (2007) PROPOSAL TITLE Southeast

Regional Medical Command (SERMC) tional Review Board (IRB) Automation Solution (PI-LTC Joseph Wood)

Institu-Wood, J., Tung, H.-L., Grayson, J., Poppeliers, C.,

& Lawless, W F (2008) A classical uncertainty principle for organizations In M Khosrow-Pour

(Ed.), Encyclopedia information science & nology reference (2nd ed.) Hershey, PA: Informa-tion Science Reference

tech-Yu, S B., & Efstathiou, J (2002, April 9-10)

An introduction of network complexity In ceedings of Tackling Industrial Complexity: The ideas that make a difference, Downing College,

Pro-Cambridge, UK

KEY TERMS AND DEFINITIONS

Bistability: Bistability occurs when one data

set produces two mutually exclusive interpretations that cannot be held in awareness simultaneously (Cacioppo et al., 1996) Bohr (1955) concluded that multiple interpretations support the existence of different cultures Further, given the importance of feedback to social dynamics (Lawless et al., 2007), rapid shifts between bistable perceptions increase uncertainty in the non-observed perception which not only underwrites social problems between dif-ferent groups, but also supports the existence of an uncertainty principle

Game Theory: Game theory was invented in

the 1940’s by Von Neuman & Morgenstern (1953)

It is a one-shot game, or repeated games, played by

2 or more agents In its most basic form, the game

configuration presents two choices for payoffs

to each player Payoffs are interdependent The

values in the configuration of choices offered to

participants are arbitrary and normative

Health Care and Life Sciences Interest

Group: The Semantic Web includes a Health

Care and Life Sciences Interest Group (HCLSIG,

2008) to establish interoperable data standards for

“connected knowledge” to improve collaboration

across the health care and life sciences The goal

for HCLSIG is to reduce medical errors, increase

physician efficiency and advance patient care and

satisfaction It includes document annotation and

rule processing (with XML formats, OWL

ontolo-gies and SWRL rule processors)

Organizations: Organizations are social

col-lectives performing a function that often cannot be

done by an individual alone Organizations do this

by assigning interdependent roles to individuals,

which requires coordinating the output of

indi-viduals, but also amplifies the capabilities of the

individual working alone (Ambrose, 2001)

Organizational Uncertainty Principle:

The organizational uncertainty principle acts as

a tradeoff in attention directed at reducing the

uncertainty in one factor, such as a worldview,

with the result that the uncertainty in a second

interdependent factor is increased inversely It

is based on Bohr’s (1955) famous notion that the uncertainty principle at the atomic level applied

to social situations is captured by human action and observation That is, the more focused indi-viduals are on acting out a series of steps, the less observant they become of their action Applied to societies, action-observation uncertainty couples that open the path to multiple interpretations of the same social behavior lie at the root of differ-ent cultures

Semantic Web: The Semantic Web is an

on-going project to extend the World Wide Web (WWW) to permit humans and machines to col-laborate efficiently As envisioned by Berners-Lee (2007), inventor of WWW (and web languages URI, HTTP, and HTML), the future Web should evolve into a universal exchange for data, informa-tion and knowledge Without a universal standard for machine access, HTML data is difficult to use

on a large scale The Semantic Web solves this problem with an efficient global mesh for informa-tion access by humans and machines

Social Learning Theory: SLT is a term coined

by Bandura (1977) that includes the three different schools of ideas that accounted for learning by organisms, but with a primary focus on humans These three schools were classical conditioning (Pavlovian associations), operant conditioning (Skinnerian rewards and punishments), and mod-eling, Bandura’s own school of thought

This work was previously published in Handbook of Research on Social Dimensions of Semantic Technologies and Web vices, edited by M M Cruz-Cunha; E F Oliveira; A J Tavares; L G Ferreira, pp 469-488, copyright 2009 by Information Science Reference (an imprint of IGI Global).

Context-awareness is highly desired, particularly

in highly dynamic mobile environments Semantic

Web Services (SWS) address context-adaptation

by enabling the automatic discovery of distributed

Web services based on comprehensive semantic

capability descriptions Even though the

ap-propriateness of resources in mobile settings is

strongly dependent on the current situation, SWS

technology does not explicitly encourage the

representation of situational contexts Therefore,

whereas SWS technology supports the allocation

of resources, it does not entail the discovery of

appropriate SWS representations for a given

situational context Moreover, describing the

complex notion of a specific situation by utilizing

symbolic SWS representation facilities is costly,

prone to ambiguity issues and may never reach

semantic completeness In fact, since not any real-world situation completely equals another,

a potentially infinite set of situation parameters has to be matched to a finite set of semantically defined SWS resource descriptions to enable context-adaptability To overcome these issues, the authors propose Mobile Situation Spaces (MSS) which enable the description of situation characteristics as members in geometrical vector spaces following the idea of Conceptual Spaces (CS) Semantic similarity between situational contexts is calculated in terms of their Euclidean distance within a MSS Extending merely symbolic SWS descriptions with context information on a conceptual level through MSS enables similarity-based matchmaking between real-world situation characteristics and predefined resource represen-tations as part of SWS descriptions To prove the feasibility, the authors provide a proof-of-concept prototype which applies MSS to support context-adaptation across distinct mobile situations

Chapter 7.2

Bridging the Gap between

Mobile Application Contexts

and Web Resources

Current and next generation wireless

communica-tion technologies will encourage a widespread use

of available resources – data and services - via

a broad range of mobile devices resulting in the

demand for a rather context-adaptive resource

retrieval Context-adaptation is a highly important

feature across a wide variety of application

do-mains and subject to intensive research throughout

the last decade (Dietze, Gugliotta & Domingue,

2007; Schmidt & Winterhalter, 2004; Gellersen,

Schmidt & Beigl, 2002) Whereas the context is

defined as the entire set of surrounding situation

characteristics, each individual situation represents

a specific state of the world, and more precisely, a

particular state of the actual context (Weißenberg,

Gartmann & Voisard, 2006) Particularly, a

situa-tion descripsitua-tion defines the context of a specific

situation, and it is described by a combination of

situation parameters, each representing a particular

situation characteristic Following this definition,

context-adaptation can be defined as the ability

of Information Systems (IS) to adapt to distinct

possible situations

To achieve this, we base on a promising

technology for distributed and highly dynamic

service oriented applications: Semantic Web

Services (SWS) SWS technology (Fensel et al.,

2006) addresses context-adaptation by means of

automatic discovery of distributed Web services

as well as underlying data for a given task based

on comprehensive semantic descriptions First

results of SWS research are available in terms of

reference ontologies – e.g OWL-S (Joint US/EU

ad hoc Agent Markup Language Committee, 2004)

and WSMO (WSMO Working Group, 2004) – as

well as comprehensive frameworks (e.g DIP

proj-ect1 results) However, whereas SWS technology

supports the allocation of appropriate services for

a given goal based on semantic representations,

it does not entail the discovery of appropriate

SWS goal representations for a given situation

Particularly in mobile settings, the current

situa-tion of a user heavily determines the intensitua-tional scope behind a user goal and consequently, the appropriateness of particular resources For instance, when attempting to retrieve localized geographical information, the achievement of a respective goal has to consider the location and device of the user

Despite the strong impact of a (mobile) context

on the semantic meaning and intention behind a user goal, current SWS technology does not ex-plicitly encourage the representation of domain situations Furthermore, the symbolic approach

- describing symbols by using other symbols out a grounding in the real world - of established SWS and Semantic Web (SW) representation standards in general, such as RDF (World Wide Web Consortium, W3C, 2004a), OWL (World Wide Web Consortium, W3C, 2004b), OWL-S (Joint US/EU ad hoc Agent Markup Language Committee, 2004), or WSMO (WSMO Working Group, 2004), leads to ambiguity issues and does not entail semantic meaningfulness, since mean-ing requires both the definition of a terminology

with-in terms of a logical structure (uswith-ing symbols) and grounding of symbols to a conceptual level (Cregan, 2007; Nosofsky, 1992).Moreover, while not any situation or situation parameter completely equals another, the description of the complex no-tion of a specific situation in all its facets is a costly task and may never reach semantic completeness Apart from that, to enable context-adaptability,

a potential infinite set of (real-world) situation characteristics has to be matched to a finite set of semantically defined parameter representations Therefore, we claim, that fuzzy classification and matchmaking techniques are required to extend and exploit the current functionalities provided

by SWS and match the specific requirements of context-aware mobile applications

Conceptual Spaces (CS), introduced by Gärdenfors (Gärdenfors, 2000; Gärdenfors, 2004) follow a theory of describing entities at the con-ceptual level in terms of their natural character-istics similar to natural human cognition in order

to avoid the symbol grounding issue CS enable

representation of objects as vector spaces within a

geometrical space which is defined through a set

of quality dimensions For instance, a particular

color may be defined as point described by vectors

measuring the quality dimensions hue, saturation,

and brightness Describing instances as vector

spaces where each vector follows a specific metric

enables the automatic calculation of their semantic

similarity, in terms of their Euclidean distance,

in contrast to the costly representation of such

knowledge through symbolic SW representations

Even though several criticisms have to be taken

into account when utilizing CS (Section 0) they

are considered to be a viable option for knowledge

representation

In this chapter, we propose Mobile

Situa-tion Spaces (MSS) as a specific derivaSitua-tion of

Conceptual Situation Spaces (CSS) MSS utilize

CS to represent situations and are mapped to

standardized SWS representations to enable first,

the situation-aware discovery of appropriate SWS

descriptions and finally, the automatic discovery

and invocation of appropriate Web services to

achieve a given task within a particular situation

Extending merely symbolic SWS descriptions

with context information on a conceptual level

through MSS enables a fuzzy, similarity-based

matchmaking methodology between real-world

situation characteristics and predefined SWS

representations within mobile environments Since

semantic similarity between situation parameters

within a MSS is indicated by the Euclidean

distance between them, real-world situation

pa-rameters are classified in terms of their distance

to predefined prototypical parameters, which are

implicit elements of a SWS description Whereas

current SWS technology addresses the issue of

allocating services for a given task, our approach

supports the discovery of SWS task representations

within a given mobile situation Consequently,

the expressiveness of current SWS standards is

extended and fuzzy matchmaking mechanisms

are supported

To prove the feasibility of our approach, a proof-of-concept prototype is provided which uses MSS to support context-adaptation by taking into account context parameters such as the current location and desired knowledge subject

The paper is organized as follows The ing Section 2 provides background information on SWS, whereas Section 3 introduces our approach

follow-of Conceptual Situation Spaces which are aligned

to current SWS representations Section 4 trates the application of CSS to mobile settings by introducing MSS Utilizing MSS, we introduce a context-adaptive prototype in Section 5 Finally,

illus-we conclude our work in Section 6 and provide

an outlook to future research

SEMANTIC WEB SERVICES AND WSMO

SWS technology aims at the automatic ery, orchestration and invocation of distributed services for a given user goal on the basis of comprehensive semantic descriptions SWS are supported through representation standards such

discov-as WSMO and OWL-S We refer to the Web Service Modelling Ontology (WSMO), a well established

SWS reference ontology and framework The conceptual model of WSMO defines the follow-ing four main entities:

• Domain Ontologies provide the

founda-tion for describing domains semantically They are used by the three other WSMO elements WSMO domain ontologies not only support Web service related knowl-edge representation but semantic knowl-edge representation in general

•

Goals define the tasks that a service request-er expects a Web service to fulfill In this sense they express the requester’s intent

• Web service descriptions represent the

functional behavior of an existing ployed Web service The description also

de-outlines how Web services communicate

(choreography) and how they are

com-posed (orchestration).

• Mediators handle data and process

interop-erability issues that arise when handling

heterogeneous systems

WSMO is currently supported through several

software tools and runtime environments, such as

the Internet Reasoning Service IRS-III (Cabral et

al., 2006) and WSMX (WSMX Working Group,

2007) IRS-III is a Semantic Execution

Environ-ment (SEE) that also provides a developEnviron-ment and

broker environment for SWS following WSMO

IRS-III mediates between a service requester and

one or more service providers Based on a client

request capturing a desired outcome, the goal,

IRS-III proceeds through the following steps utilizing

the set of SWS capability descriptions:

1 Discovery of potentially relevant Web

services

2 Selection of set of Web services which best

fit the incoming request

3 Invocation of selected Web services whilst

adhering to any data, control flow and Web

service invocation constraints defined in the

SWS capabilities

4 Mediation of mismatches at the data or

process level

In particular, IRS-III incorporates and extends

WSMO as core epistemological framework of the

IRS-III service ontology which provides semantic

links between the knowledge level components

describing the capabilities of a service and the

restrictions applied to its use

However, even though SWS technologies

en-able the dynamic allocation of Web services for

a given goal, it does not consider the adaptation

to different user contexts In order to fully enable

context-aware discovery of resources as required

by mobile settings (Section 1), the following

shortcomings have to be considered:

I1 Lack of explicit notion of context: current

SWS technology does not entirely specify how to represent domain contexts For ex-ample, WSMO addresses the idea of con-text: Goal and web service represent the user and provider local views, respective-ly; the domain ontologies define the termi-nologies used in each view; and the media-tors are the semantic bridges among such distinct views However, WSMO does not specify what a context description should define and how the context elements should

sym-of a logical structure (using symbols) and grounding of symbols to a conceptual level (Cregan, 2007; Nosofsky, 1992)

I3 Lack of fuzzy matchmaking gies: Describing the complex notion of a

methodolo-specific situation in all its facets is a costly task and may never reach semantic com-pleteness Whereas not any situation and situation parameter completely equals an-other, the number of (predefined) semantic representations of situations and situation parameters is finite Therefore, a possibly infinite set of given (real-world) situation characteristics has to be matched to a fi-nite set of predefined parameter instance representations which are described within

an IS Consequently, fuzzy classification and matchmaking techniques are required

to classify a real-world situation based

on a limited set of predefined parameter descriptions

CONCEPTUAL SITUATION SPACES

To address the issues I1 - I3 introduced in Section

0, we propose Mobile Situation Spaces (MSS) as

a setting-specific realisation of our metamodel

for Conceptual Situation Spaces (CSS) (Dietze,

Gugliotta & Domingue, 2008)

CSS Formalisation

CSS enable the description of a particular

situa-tion as a member of a dedicated CS As defined

in (Weißenberg et al., 2006) a situation is defined

as:

={ (1, ,2 1, 2, , ) Î }

Where t 1 is the starting time of a situation,

t 2 represents the end time of a situation and cp i

being situation parameters which are invariant

throughout the time interval defined through t 1

and t 2 Referring to (Gärdenfors, 2004; Raubal,

2004), we define a CSS (css:Conceptual Situation

Space in Figure 1) as a vector space:

dimensions c i Please note, that we do not guish between dimensions and domains - beings sets of integral dimensions (Gärdenfors, 2004)

distin but enable dimensions to be detailed further in terms of subspaces Hence, a dimension within one space may be defined through another conceptual space by using further dimensions (Raubal, 2004)

In such a case, the particular quality dimension

c j is described by a set of further quality sions with

sev-prominence value p (css:Prominence) for each

dimension Therefore, a CSS is defined by

where P is the set of real numbers However, the

usage context, purpose and domain of a particular

CSS strongly influence the ranking of its quality

dimensions This clearly supports our position

of describing distinct CSS explicitly for specific

domains only

Particular members (css:Member) in the CSS

are described through a set of valued dimension

vectors (css:Valued Dimension Vectors)

Sym-bolic representations of domain situations and

parameters, such as css:Situation Description

and css:Situation Parameter, refer to particular

CSS (css:Conceptual Situation Space) whereas

parameter instances are represented as members

(css:Member).

Moreover, referring to Gärdenfors (2004) we

consider prototypical members (css:Prototypical

Member) within a particular space Prototypical

members enable the classification of any arbitrary

member m within the a specific CSS, by simply

calculating the Euclidean distances between m

and all prototypical members in the same space to

identify the closest neighbours of m For instance,

given a CS to describe apples based on their shape,

taste and colour, a green apple with a strong and

fruity taste may be close to a prototypical

mem-ber representing the typical characteristics of the

Granny Smith species Figure 1 depicts the CSS

metamodel

The metamodel introduced above has been

formalized into a Conceptual Situation Space

Ontology (CSSO), utilizing OCML (Motta, 1998)

In particular, each of the depicted entities is

repre-sented as a concept within CSSO whereas

associa-tions are reflected as their properties in most cases

The correlation relationship indicates whether two

dimensions are correlated or not For instance,

when describing an apple the quality dimension

describing its sugar content may be correlated with

the taste dimension Information about correlation

is expressed within the CSSO through axioms

related to a specific quality dimension instance

CSSO is aligned to a well-known foundational

ontology: the Descriptive Ontology for Linguistic

and Cognitive Engineering (DOLCE) (Gangemi, Guarino, Masolo, Oltramari, Schneider, 2002) and,

in particular, its module Descriptions and tions (D&S) (Gangemi, Mika, 2003) The aspect

Situa-of gradually refining a CSS through subspaces corresponds to the approach of DOLCE D&S to gradually refine a particular description by using parameters where each parameter can be described

or Manhattan distance (Krause, 1987), could be considered, even though the nature of the space and its possible metrics suggests the Euclidean distance as a useful metric to calculate similarities Applying a formalization of CS proposed in Raubal (2004) to our definition of a CSS, we formalize the Euclidean distance between two members

in a CSS as follows Given a CSS definition C

and two members represented by two vector sets

V and U, defined by vectors v 0 , v 1 , …,v n and u 1 ,

u 2 ,…,u n within C, the distance between V and U

can be calculated as:

where z(u i ) is the so-called Z-transformation or standardization (Devore, Peck, 2001) from u i. Z-transformation facilitates the standardization of distinct measurement scales which are utilized by different quality dimensions in order to enable the calculation of distances in a multi-dimensional and multi-metric space The z-score of a par-

ticular observation u i in a dataset is calculated

as follows:

-where u is the mean of a dataset U and s uis the

standard deviation from U Considering

promi-nence values p i for each quality dimension i, the

Euclidean distance d(u,v) indicating the semantic

similarity between two members described by

vec-tor sets V and U can be calculated as follows:

s

s i

i u

i v i

Utilizing CSS for SWS Selection

Whereas the discovery of distributed Web

services for a given user goal is addressed by

current SWS technology, such as WSMO, and

corresponding reasoners, the context-aware

selection of a specific SWS goal representation

for a given situation is a challenging task to be

tackled when developing SWS-driven

applica-tions By providing an alignment of CSS and

SWS, we address this issue by enabling the

classification of an individual situation along

predefined situation descriptions - used within

SWS descriptions - based on semantic

similar-ity calculation Therefore, CSS are aligned to

WSMO to support the automatic discovery of the

most appropriate goal representation for a cific situation Since both metamodels, WSMO

spe-as well spe-as CSS, are represented bspe-ased on the OCML representation language (Motta, 1998), the alignment was accomplished by defining relations between concepts of both ontologies

gine utilizes capability descriptions to identify

SWS (wsmo:Web Service) which suit a given

Goal In contrast, the preliminary selection

of the most appropriate goal description for a given situation is addressed by classification of situation parameters through CSS For instance, given a set of real-world situation parameters, described as members in a CSS, their semantic similarity with predefined prototypical param-

eters (css:Prototypical Member) is calculated

Given such a classification of a particular world situation, a goal representation which assumes matching prototypical parameter instances is selected and achieved through the reasoning engine

real-Figure 2 Alignment of CSS and WSMO.

Deriving CSS for Certain

Application Contexts

As stated in Gärdenfors (2000), the definition and

prioritization of quality dimensions within a CS

is highly dependent on the purpose and context

of the space For instance, when describing an

apple, dimensions may be differently weighted,

dependent on whether the apple is subject to visual

cognition exclusively or to full sensory perception,

what would be the case if the apple is supposed

to be eaten Whereas in the first case, dimensions

such as color and shape are highly ranked, taste

and texture may additionally be important in the

latter case

Consequently, the derivation of an

appropri-ate space for a certain purpose is considered an

important task which usually should be carried

out by a qualified individual such as an

applica-tion designer We particularly foresee a procedure

consisting of the following steps:

S1

Identification of situation parameters eligi-ble for representation as quality dimension

c i

S2 Assignment of prominence values p i to

each quality dimension c i

S3 Assignment of metrics to each quality

di-mension c i

With respect to S1, one has to take into account

which aspects of a situation are relevant from an

application perspective, i.e which characteristics

have an impact on the applied context adaptation

strategy or rules In the case of our intended

us-age of CSS for SWS selection, only parameters

are important, which are considered within SWS

capability representations (Section 0)

Since several dimensions might have a different

impact factor on the entire space, S2 is aimed at

assigning a prominence value p i to each dimension

c i Prominence values should usually be chosen

from a predefined value range, such as 0 1

However, since the assignment of prominences to

quality dimensions is of major importance for the semantic meaning of calculated distances within

a space, this step is not straightforward and most probably requires ex post re-adjustment

During the final step S3, a quantitative metric

has to be assigned to each previously defined mension Whereas certain dimensions naturally are described using qualitative measurements, such

di-as a size or a weight, other dimensions are ally described using rather qualitative values The latter applies for instance to the notion of a color

usu-In case no quantitative metric can be assigned to

a certain quality dimension c i, a subspace has to

be defined which refines the particular dimension through further dimensions For instance, in the case of the color dimension, a subspace could be defined using the quantitative dimensions hue, saturation and brightness Hence, the proposed procedure has to be repeated iteratively until a sufficient description depth has been achieved

leading to the definition of a CSS C of the form

(Section 0):

A MOBILE SITUATION SPACE

Following the steps introduced in Section 0, we derive a CSS aimed at representing situations in mobile settings A mobile situation is defined by parameters such as the technical environment used by a user, his/her current objectives and par-ticularly the current location Since each of these parameters apparently is a complex theoretical construct, most of the situation parameters cannot

be represented as a single quality dimension within the CSS, but have to be represented as dedicated subspaces which are defined by their very own dimensions (Section 0) Moreover, applying CSS

to represent a particular concept is only reasonable

in cases where similarity calculation is possible and semantically meaningful, i.e a particular measure-

ment can be applied to each quality dimension

For instance, the native language of a user is a

crucial important situation parameter, but in this

case, only a direct match is reasonable in order

to provide appropriate information resources in

the correct language to the user

Therefore, this section focuses exemplarily

on the representation of two parameters through

a CSS subspace, which are of particular interest:

the location and the subject a user is interested

in Due to the complex and diverse nature of a

particular subject or spatial location, traditional

symbolic representation approaches of the

Se-mantic Web are supposed to fail since it is nearly

impossible to define either a subject or a location

in a non-ambiguous and comprehensive way by

just following a symbolic approach

Moreover, a one-to-one matchmaking between

different locations and subjects is hard to achieve,

since fairly not any instance of these parameters

completely equals another one Therefore, fuzzy

similarity detections, as enabled through MSS,

have to be utilized

To represent spatial locations, we define a

CSS subspace L with 2 quality dimensions l i

representing the latitude and longitude of a

par-ticular location

1 2

In order to represent a particular subject,

we currently consider 4 dimensions (history, geography, culture, languages) which are used

to describe the semantic meaning of a particular

subject within subspace S:

Figure 3 depicts the key concepts of the

ontol-ogy describing L and S as subspaces (css:Location Space, css:Subject Space) within the mobile space (css:Mobile Situation Space).

Moreover, Figure 3 depicts the relation of the

subspace L (css:Location Space) and subspace

S (css:Subject Space) with WSMO-based SWS

descriptions, represented via grey-colored cepts (Section 0)

con-Instances of a situation parameter representing

a subject are defined by particular members within

the space S (css:Subject Space), which itself uses

4 quality dimension c i, whereas instances of a rameter representing a spatial location are defined

pa-by members within the space L (css:Location Space), which itself uses 4 quality dimension l i.

The metric scale, datatype and value range for each

dimension s i and l i are presented in Table 1:

As depicted in Table 1, each quality

dimen-sion l i is ranked on an interval scale with value ranges being float numbers between -90 and +90

in case of the latitude and between -180 and +180

Figure 3 Key concepts representing mobile situation subspaces.

in case of the longitude Furthermore, each

qual-ity dimension c i is ranked on a ratio scale with

value ranges being float numbers between 0 and

100 The authors would like to highlight, that no

prominence values have been assigned since each

dimension has an equal impact to define a

particu-lar member It is obvious, that the assignment of

prominence values is a highly subjective process,

strongly dependent on the purpose, context and

individual preferences Therefore, future work is

aimed at enabling users to assign rankings of

qual-ity dimensions themselves in order to represent

their individual priorities regarding the service

retrieval process

To classify an individual mobile situation, we

define prototypical members (css:Prototypical

Member) in the Mobile Situation Space For

in-stance, to describe particular cities as members

within L, we utilized geodata, retrieved from

GoogleMaps2, to describe a prototypical member

for each location which is targeted by a particular

SWS A few examples of prototypical location

members used in the current prototype application

are represented in Table 2:

An example of how such parameters are

repre-sented in a formal knowledge modeling language

is given in Section 0 Moreover, we predefined

several prototypical subjects in S, each

represent-ing the maximum value of a particular quality

dimension s i what resulted in the following 4

prototypical subjects

Apart from the depicted subjects, each subject

which is described as part of a symbolic SWS

capability representation had been referred to an

individual member in S.

SIMILARITY-BASED SWS

SELECTION AND ACHIEVEMENT

IN A MOBILE SETTING

To prove the feasibility of our approach, we

pro-vide a proof-of-concept prototype application,

which utilizes MSS (Section 4) - based on the

CSS metamodel introduced in Sections 0 - and supports context-adaptation in a mobile environ-ment based on SWS and CSS

Runtime Support for CSS and SWS

The following Figure 4 depicts the general chitecture adopted to support reasoning on MSS and SWS in distinct domain settings through a Semantic Execution Environment (SEE), which

ar-in our case is IRS-III (Section 0)

Multiple mobile devices - such as PDAs, mobiles or any other portable device hosting a Web browser - can serve as user interface of the

Table 1 Metric scale, range, and data type of quality dimensions l i and s i

Quality Dimension Metric Scale Data-type Range

Table 2 Prototypical members within L

Prototype l 1 (Latitude) l 2 (Longitude)

SEE, enabling the user (and the device itself) to

provide information about his/her goal and the

current real-world situation

The SEE makes use of semantic representations

of the CSS formalisation (CSS ontology, CSSO),

specifically derived for mobile settings, and of

SWS annotations based on WSMO in order to

discover and allocate the most appropriate resource

for a given user goal within a current situation

Ontologies had been represented using the OCML

knowledge modeling language (Motta, 1998)

WSMO capabilities are represented by

defin-ing the assumptions and effects of available SWS

and goals in terms of certain situation description

or situation parameter instances (Section 0) Such

situation descriptions are refined as particular

prototypical members of an associated CSS,

such as prototypical members of the MSS S and

L introduced in Section 4

As mentioned in Section 3, CSSO allows us

to describe a specific mobile situation

descrip-tion instance in terms of a collecdescrip-tion of situadescrip-tion

parameter instances Mobile situation description

instances are automatically and gradually defined

at runtime by the SEE as the result of the user

interaction with the mobile device On the basis of

the detected context parameters, the SEE performs the following steps:

1 Computation of similarities between the detected real-world context parameters—obtained from the user and its device—and symbolic representation of prototypical situation parameters;

2 Progressive update of the current mobile situation description with the closest pro-totypical situations parameters;

3 Determination of (WSMO) goal matching the refined situation description;

4 Achievement of selected goals by means

of discovery and orchestration of available web services

Consequently, we enable the classification

of real-world context parameters along able predefined parameters in order to enable

avail-a similavail-arity-bavail-ased selection avail-and orchestravail-ation WSMO goals

Figure 4 Architecture to support runtime reasoning on CSS and SWS.

Context Classification

and Adaptation

As outlined in the previous section, the SEE

automatically detects the semantic similarity

of specific situation parameters with a set of

predefined prototypical parameters to enable

the allocation of context-appropriate resources

In this section, we further detail these aspects,

since they are central in the contribution of this

chapter In particular, we specify the concepts of

classification and adaptation

Referring to CSS subspaces L and S described

in Section 0, given a particular member U in L or

S, its semantic similarity with each of the

proto-typical members is indicated by their Euclidean

distance Since we utilize spaces described by

dimensions which each use the same metric scale

and no prominence value, the distance between two

members U and V can be calculated disregarding

a Z-transformation (Section 0) for each vector:

Please note, that it would be possible to

cal-culate distances either between entire situations

(members within css:Mobile Situation Space) or

between particular parameter (members in

sub-spaces such as L and S) Since individual semantic

similarities between instances of parameters such

as the current location or the desired subject are

usually important knowledge when deciding about

the appropriateness of resources for a given

con-text, the current application calculates distances

between each parameter, i.e between members

within each individual subspace

The calculation of Euclidean distances

us-ing the formula shown above is performed by a

standard Web service, which is annotated as SWS

and invoked through IRS-III at runtime Given a

particular CSS description, a member

(represent-ing a specific parameter instance) as well as a set

of prototypical member descriptions (representing

prototypical parameter instances), similarities are calculated by the Web service at runtime in order

to classify a given situation parameter

For instance, a user is currently located in Eastbourne (UK) and is interested in historical information about the surrounding area Con-sequently, the particular situation description

(css:MobileSituation Desccription) includes a location parameter which is defined by a member E

in the specific location space (css:Location Space)

with the following vectors describing latitude and longitude of Eastbourne:

Figure 5 Mobile device showing semi-automatic location detection.

Based on the current situation description, SWS

are selected which are able to address the

situa-tion Whereas parameters which are not defined

by members in a specific CSS require a direct

match with a corresponding SWS description, a

similarity-based match is computed for parameters

which are described in a CSS, e.g the location

or the subject Hence, distance calculation was

utilized to identify similarities between current

context parameters – such as E and S1 – and

prototypical parameters which had been defined

as part of SWS capability descriptions in order

to represent the parameters targeted by available

SWS In order to illustrate the representation of

prototypical CSS members, the following OCML

code defines a location parameter instance

rep-resenting the geospatial location Brighton, as

well as the respective prototypical member (L3)

in the MSS L.

Calculating distances between E and targeted

locations – represented as prototypical MSS

mem-bers - led to the identification of the following

distances to the three closest matches: (Table 4)

Since not any SWS targets historical interests

(S1) exclusively – as desired by the user - no direct

match between the situation and subjects targeted

by available SWS was achieved However, larity calculation identified related subject areas, which partially target historical information Table 5 indicates their vectors and distances to

simi-the required subject S1.

The subjects S5, S6 and S7 as well as the tions L1, L2, and L3 shown in Table 4 and Table

loca-5 had been described as prototypical members in the MSS (Section 0) during the development of SWS representations targeting certain subjects and locations By following our alignment from Section 0, this task could be performed by either the Web service provider or any SWS expert who

is providing and publishing a semantic tion of available Web services

representa-As indicated by the Euclidean distances picted in Tables 4 and 5, the closest matching SWS

de-provides historical and cultural (S7) resources for the Brighton (L3) area, as these show the lowest

distances Provided these similarities, a user is able to select predefined parameters that best suit his/her specific preferences within the cur-rent situation In that, the use of similarity-based classification enables the gradual refinement of

Listing 1 Partial OCML code defining location parameter instance and respective MSS member

(def-instance brighton-location location

(has-valued-dimension (brighton-valued-lat-vector brighton-valued-long-vector))))

(def-instance brighton-valued-lat-vector location-valued-dimension-vector

a situation description and fuzzy matchmaking

between real-world situations, and prototypical

parameters predefined within a SWS description

For example, the following OCML code defines

the partial capability description of a Web service

that provides historic and cultural information for

the area of Brighton:

In fact, the assumption expression presented

above describes that situation description

repre-senting the current situation (has-situation)

con-sider the location Brighton and the subject S7.

As a result, in our approach, the actual mobile

situation description (i.e the actual context) is

the result of an iterative process that involves

several distance calculations to map symbolic

representations and real world characteristics

Notice that this process actively involves the end

users in providing observables and validating the

distance calculations According to the obtained situation parameters and the selected user goal, the SEE discovers and orchestrates annotated Web services, which show the capabilities to suit the given situation representation Whereas discov-ery and orchestration are addressed by existing SWS technology, the context-aware selection of

a specific SWS goal representation is addressed through CSS by enabling similarity-based clas-sifications of individual situations as described

in the previous sections

RELATED WORK

Since our work relates to several different but related research areas, we report here related work on (i) Semantic Web Services, (ii) Context-adaptive systems, and (iii) Context-adaptation

in mobile environments Moreover, by ing our approach with related work in (iii) we describe our contribution to the current state of the art in context-adaptive mobile and ubiquitous computing

compar-SWS: OWL-S (OWL-S Coalition 2004) is a

comparatively narrow framework and ontology for adding semantics to Web service descriptions

In order to identify problematic aspects of OWL-S and suggest possible enhancements, a contextual-ized core ontology of services has been described

in Mika et al (2004) Such an ontology is based

on DOLCE (Gangemi et al., 2002) and its specific module D&S (Gangemi, Mika, 2003) Even though

we followed a similar approach, we adopt WSMO (WSMO Working Group, 2004) instead of OWL-

Table 4 Distances between E and targeted

loca-tions

Prototype Euclidean Distance

L1: Milton Keynes (UK 1.6125014961413195

Listing 2 Partial OCML code representing SWS capability in terms of assumed MSS members

(def-class lpmo-get-brighton-his-and-cult-LOs-ws-capability (capability) ?capability

((used-mediator:value lpmo-get-brighton-his-and-cult-LOs-mediator)

(has-assumption:value

(KAPPA (?web-service)

(and (= (get-location (wsmo-role-value ?web-service ‘has-situation)) “ Brighton”))

(= (get-subject (wsmo-role-value ?web-service ‘has-situation)) “S7”)))))

S as reference ontology for SWS Moreover, the

aim of our resulting ontology is not proposing

changes to WSMO, but creating domain-specific

models which incorporate WSMO-based SWS

representations

Context-adaptive systems: in Bouquet et al

(2003) the authors define contexts as the local

models that encode a party’s view of a domain

They distinguish contexts from ontologies, since

the latter are shared models of some domain that

encode a view which is common to a set of

dif-ferent parties Contexts are best used in those

applications where the core problem is the use

and management of local and autonomous

rep-resentations with a need for a lack of centralized

control For example, the notion of contexts is used

in some applications of distributed knowledge

management Bonifacio et al (2003), pervasive

computing environments (Chen, Finin & Joshi,

2003) and peer-to-peer applications (Serafini et

al., 2003) According to the definition introduced

in Bouquet et al (2003), we propose a novel use

of contexts The local models encode party’s view

of SWS-based process descriptions

Context-adaptation in mobile

environ-ments: Weissenberg et al (2006) adopt an

ap-proach to context-adaptation in mobile settings

which shows some similarities to ours: given a

set of context parameters – based on sensor data

– first a context is identified and then a

match-ing situation However, they rely on manually

predefined axioms which enable such a reasoning

compared to the automatic detection as proposed

in this paper Korpipaa et al (2003) propose a

related framework but firstly, require client-side

applications to be installed and, secondly, relies

on Bayesian reasoning for matching between

measured lower-level contexts and higher-level

context abstractions represented within an

on-tology Hence, as a major lack, it is required to

provide information about contexts and their

relations within a Bayesian Network in order to

perform the proposed reasoning Gu, Wang, Pung

& Zang (2004) propose a context-aware

middle-ware which also distinguishes between lower-level and higher-level contexts However, there is no mechanism to automatically identify relationships between certain contexts or context parameters The same criticism applies to the approaches to a semantic representation of user contexts described

in Toivinen, Kolari & Laako (2003) and Sathish, Pavel & Trossen (2006)

Finally, it can be highlighted, that current proaches to context-adaptation in mobile settings usually rely on the manual representation of map-pings between a given set of real-world context data and predefined context representations Since this approach is costly and time-consuming, our approach could contribute there significantly by providing a similarity-based and rather fuzzy method for automatically identifying appropriate symbolic context representations given a set of detected context parameters

ap-CONCLUSION

In this paper, we proposed an approach to support fuzzy, similarity-based matchmaking between real-world situation parameters in mobile settings and predefined semantic situation descriptions by incorporating semantic context information on a conceptual level into symbolic SWS descriptions based on Conceptual Situation Spaces Given a particular mobile situation, defined by param-eters such as the location and device of the user, the most appropriate resources, whether data or services, are discovered based on the semantic similarity, calculated in terms of the Euclidean distance, between the real-world situation and predefined resource descriptions as part of SWS representations Even though we refer to the SWS framework WSMO in this paper, we would like

to highlight, that our approach could be applied

to other SWS reference ontologies such as

OWL-S (OWL-OWL-S Coalition 2004) Consequently, by aligning CSS to established SWS technologies, the expressiveness of symbolic SWS standards is

extended with context information on a conceptual

level described in terms of natural quality

dimen-sions to enable fuzzy context-aware delivery of

information resources at runtime Whereas

cur-rent SWS frameworks address the allocation of

distributed services for a given (semantically)

well-described task, Mobile Situation Spaces

particularly address the similarity-based discovery

of the most appropriate SWS task representation

for a given situation To prove the feasibility of

our approach, a proof-of-concept prototype

ap-plication was presented, which applies the MSS

to enable context-adaptive resource discovery in

a mobile setting

However, although our approach applies CS

to solve SWS-related issues such as the symbol

grounding problem, several criticisms still have

to be taken into account Whereas defining

situ-ational contexts, respectively members within a

given MSS, appears to be a straightforward

pro-cess of assigning specific values to each quality

dimension, the definition of the MSS itself is not

trivial at all and strongly dependent on individual

perspectives and subjective appraisals Whereas

the semantics of an object are grounded to

met-rics in geometrical vector spaces within a MSS,

the quality dimensions itself are subject to ones

perspective and interpretation what may lead to

ambiguity issues With regard to this, MSS do

not appear to solve the symbol grounding issue

but to shift it from the process of describing

instances to the definition of a MSS Moreover,

distinct semantic interpretations and conceptual

groundings of each dimension may be applied by

different individuals Apart from that, whereas

the size and resolution of a MSS is indefinite,

defining a reasonable space for a specific domain

and purpose may become a challenging task

Nevertheless, distance calculation as major

con-tribution of the MSS approach, not only makes

sense for quantifiable parameters but also relies

on the fact, that parameters are described in the

same geometrical space

Consequently, CS-based approaches, such

as MSS, may be perceived as step forward but

do not fully solve the issues related to symbolic Semantic Web (Services)-based knowledge rep-resentations Hence, future work has to deal with the aforementioned issues For instance, we fore-see to enable adjustment of prominence values

to quality dimensions of a specific space to be accomplished by a user him/herself, in order to most appropriately suit his/her specific priorities and preferences regarding the resource allocation process, since the prioritization of dimensions

is a highly individual and subjective process Nevertheless, further research will be concerned with the application of our approach to further domain-specific situation settings

vices based Applications Proceedings of the 5 th International Semantic Web Conference (ISWC),

Cregan, A (2007) Symbol Grounding for the

Semantic Web 4th European Semantic Web

Conference 2007, Innsbruck, Austria.

Devore, J., & Peck, R (2001) Statistics—The

Exploration and Analysis of Data 4th ed Pacific

Grove, CA: Duxbury

Dietze, S., Gugliotta, A., & Domingue, J (2007)

A Semantic Web Services-based Infrastructure for

Context-Adaptive Process Support Proceedings

of IEEE 2007 International Conference on Web

Services (ICWS), Salt Lake City, Utah, USA.

Dietze, S., Gugliotta, A., & Domingue, J (2008)

Towards Context-aware Semantic Web Service

Discovery through Conceptual Situation Spaces

Workshop: International Workshop on Context

enabled Source and Service Selection,

Integra-tion and AdaptaIntegra-tion (CSSSIA), 17th InternaIntegra-tional

World Wide Web Conference (WWW2008),

Bei-jing, China

Fensel, D., Lausen, H., Polleres, A., de Bruijn,

J., Stollberg, M., Roman, D., & Domingue, J

(2006) Enabling Semantic Web Services—The

Web service Modelling Ontology Springer.

Gangemi, A., Guarino, N., Masolo, C., Oltramari,

A., & Schneider, L (2002) Sweetening

Ontolo-gies with DOLCE In A Gómez-Pérez, V Richard

Benjamins (Eds.), Knowledge Engineering and

Knowledge Management Ontologies and the

Semantic Web: 13th International Conference,

EKAW 2002, Siguenza, Spain, October 1-4

Gangemi, A., & Mika, P (2003) Understanding

the Semantic Web through Descriptions and

Situ-ations In R Meersman, Z Tari, & et al (Eds.),

Proceedings of the On The Move Federated

Con-ferences (OTM’03), LNCS Springer Verlag.

Gärdenfors, P (2000) Conceptual Spaces—The

Geometry of Thought MIT Press.

Gärdenfors, P (2004) How to make the semantic web more semantic In A C Vieu & L Varzi,

(Eds.), Formal Ontology in Information Systems,

Joint US/EU ad hoc Agent Markup Language Committee (2004) OWL-S 1.1 Release http://www.daml.org/services/owl-s/1.1/

Korpipaa, P., Mantyjarvi, J., Kela, J., Keranen, H., & Malm, E (2003, Jul-Sept) Managing

Context Information in Mobile Devices IEEE Pervasive Computing / IEEE Computer Society [and] IEEE Communications Society, 2(3), 42–51

doi:10.1109/MPRV.2003.1228526Krause, E F (1987) Taxicab Geometry Dover.Mika, P., Oberle, D., Gangemi, A., & Sabou,

M (2004) Foundations for Service Ontologies: Aligning OWL-S to DOLCE, WWW04.

Motta, E (1998) An Overview of the OCML

Modelling Language.The 8th Workshop on ods and Languages.

Meth-Nosofsky, R (1992) Similarity, scaling and

cognitive process models Annual Review of Psychology, 43, 25–53 doi:10.1146/annurev.

ps.43.020192.000325

Raubal, M (2004) Formalizing Conceptual

Spaces In A Varzi & L Vieu (Eds.), Formal

Ontology in Information Systems, Proceedings

of the Third International Conference (FOIS

2004).Frontiers in Artificial Intelligence and

Applications, 114, 153-164., Amsterdam, The

Netherlands: IOS Press

Sailesh, S., Pavel, D., & Trossen, D (2006)

Con-text Service Framework for Mobile Internet

Inter-national Worskshop on System Support for Future

Mobile Computing Applications (FUMCA 2006),

September 2006, Irvine, California, USA

Schmidt, A., & Winterhalter, C (2004) User

Context Aware Delivery of E-Learning Material:

Approach and Architecture Journal of

Univer-sal Computer Science (JUCS), 10(1), January

2004

Serafini, L., Giunchiglia, F., Mylopoulos, J., &

ernstein, P (2003) Local relational model: a

logical formalization of database coordination

In P Blackburn, C Ghidini, & R Turner (Eds.),

Context’03.

Toivonen, S., Kolari, J., & Laakko, T (2003)

Facilitating mobile users with contextualized

content In Proc Workshop Artificial Intelligence

in Mobile Systems.

Weißenberg, N., Gartmann, R., & Voisard, A (2006) An Ontology-Based Approach to Person-alized Situation-Aware Mobile Service Supply

Geoinformatica 10, 1 (Mar 2006), 55-90 DOI=

http://dx.doi.org/10.1007/s10707-005-4886-9.World Wide Web Consortium W3C (2004a): Resource Description Framework, W3C Rec-ommendation 10 February 2004, http://www.w3.org/RDF/

World Wide Web Consortium W3C (2004b): Web Ontology Language Reference, W3C Rec-ommendation 10 February 2004, http://www.w3.org/TR/owl-ref/

WSMO Working Group (2004), D2v1.0: Web service Modeling Ontology (WSMO) WSMO Working Draft, (2004) (http://www.wsmo.org/2004/d2/v1.0/)

WSMX Working Group (2007), The Web Service Modelling eXecution environment, http://www.wsmx.org/

Chapter 7.3

Uncertainty Representation and Reasoning in the Semantic Web

Paulo Cesar G Costa

George Mason University, USA

Kathryn Blackmond Laskey

George Mason University, USA

Thomas Lukasiewicz

Oxford University Computing Laboratory, UK

ABSTRACT

This chapter is about uncertainty representation and

reasoning for the Semantic Web (SW) We address the

importance, key issues, state-of-the-art approaches,

and current efforts of both the academic and business

communities in their search for a practical, standard

way of representing and reasoning with incomplete

information in the Semantic Web The focus is on

why uncertainty representation and reasoning are

necessary, its importance to the SW vision, and the

major issues and obstacles to addressing uncertainty

in a principled and standardized way Although

some would argue that uncertainty belongs in the

“rule layer” of the SW, we concentrate especially

on uncertain extensions of ontology languages for

the Semantic Web

WHY CARE ABOUT UNCERTAINTY?

After some years of SW research, the subject mains rife with controversy, and there is still some disagreement on how uncertainty should be handled

re-in SW applications Thus, it is no surprise that little was said on the subject in previous chapters of this book A major reason for the present state of affairs

is that the most popular technologies employed in

SW applications are rooted in traditional knowledge representation formalisms that have historically ig-nored uncertainty The most compelling examples are Frame Systems (Minsky, 1975), and Description Logics, which evolved from the so-called “Structured Inheritance Networks” (Brachman, 1977), and form the logical basis for the ontology language OWL

The spotlight is not on the status quo, but on

what the future holds To answer this question, we

DOI: 10.4018/978-1-60566-112-4.ch013

begin with a comprehensive analysis of the major

challenges to be faced by the SW community,

including what kinds of interactions, scenarios,

demands, and obstacles must be addressed to

make the SW promises a reality Next, we assess

whether protocols that rely only on complete,

deterministic information will suffice to address

these challenges Although much progress has

been made by tackling problems in which

un-certainty is inessential or can be circumvented,

addressing the full range of challenges inherent in

the Semantic Web vision will require optimal use

of all available information In this Chapter, we

argue that a principled framework for

represent-ing and reasonrepresent-ing with incomplete information

is necessary to realizing the SW vision Because

uncertainty is a ubiquitous aspect of most

real-world problems, any representation scheme

intended to model real-world entities, properties

and processes must be able to cope with uncertain

phenomena Current SW technologies’ inability to

represent and reason about uncertainty in a sound

and principled manner raises an unnecessary

bar-rier to the development of new, powerful features

for general knowledge application, a limitation

that threatens to derail the original vision for

the Semantic Web as a whole In other words,

we argue that realizing the SW as envisioned by

Tim Berners-Lee (Berners-Lee & Fischetti, 2000)

requires a principled framework for representing

and reasoning with uncertainty

The Semantic Web envisions effortless

coop-eration between humans and computers,

seam-less interoperability and information exchange

among web applications, and rapid and accurate

identification and invocation of appropriate Web

services While considerable progress has been

achieved toward realization of the Semantic Web

vision, it is increasingly apparent that a sound and

principled technology for handling uncertainty is

an important requirement for continued progress

Uncertainty is an unavoidable factor in knowledge

interchange and application interoperability

Different applications have different ontologies,

different semantics, and different knowledge and data stores Legacy applications are usually only partially documented and may rely on tacit usage conventions that even proficient users do not fully understand or appreciate Furthermore, data that

is exchanged in the context of the semantic web

is often incomplete, inconsistent, and inaccurate This suggests that recent work in the application

of probability, fuzzy logic, and decision theory

to complex, open-world problems could be of vital importance to the success of the Semantic Web Incorporating these new technologies into languages, protocols, and specifications for the Semantic Web is fundamental to realizing the Semantic Web vision

Typical Problems Needing Uncertainty Representation and Reasoning The following

web-relevant reasoning challenges illustrate the kinds of problems for which reasoning under uncertainty is important

Information extracted from large

informa-• tion networks such as the World Wide Web

is typically incomplete The ability to ploit partial information is useful for iden-tifying sources of service or information For example, the fact that an online service deals with greeting cards may be evidence that it also sells stationery It is clear that search tools capable of utilizing proba-bilistic knowledge could increase search effectiveness

ex-Much information on the World Wide Web

•

is likely to be uncertain Common examples include weather forecasts and gambling odds A canonical method for representing and integrating such information and the uncertainty associated with it is necessary for communicating such information in a seamless fashion

Web information is also often incorrect or

• only partially correct, raising issues related

to trust or credibility Uncertainty sentation and reasoning helps to resolve

repre-tensions amongst information sources for

purposes of approximating appropriately

The

• Semantic Web will require

numer-ous distinct but conceptually overlapping

ontologies to co-exist and interoperate

Ontology mapping will benefit from the

ability to represent and reason with

in-formation about partial overlap, such as

likelihoods of membership in Class A of

Ontology 1 given membership in Class B

of Ontology 2

Section 5 below discusses some use cases,

based on the work of the W3C Uncertainty

Rea-soning for the World Wide Web Incubator Group

(URW3-XG) These use cases exhibit the above

characteristics, and are representative of the kinds

of challenges that the SW must address Despite

the potential that a principled framework for

rep-resenting uncertainty would have in contributing

to the development of robust SW solutions, for

historical reasons, research on the Semantic Web

started with little support for representing and

reasoning in the presence of uncertain, incomplete

knowledge As interest in and application of SW

technology grows, there is increasing recognition

of the need for uncertain reasoning technology,

and increasing discussion of the most appropriate

ways to address this need

Should Ontologies Represent Uncertainty?

A major impediment to widespread adoption of

technologies for representing and reasoning with

incomplete information is the dominance of the

classical logic paradigm in the field of

ontologi-cal engineering There is a plethora of definitions

of the term ontology in the field of information

systems Among these, a common underlying

as-sumption is that classical logic would provide the

formal foundation for knowledge representation

and reasoning Until recently, theory and methods

for representing and reasoning with uncertain and

incomplete knowledge have been neglected almost

entirely However, as research on knowledge

en-gineering and applications of ontologies matures,

the ubiquity and importance of uncertainty across

a wide array of application areas has generated consumer demand for ontology formalisms that can capture uncertainty Although recognition of the need for uncertainty reasoning is growing, there is disagreement about its appropriate place

in the Semantic Web architecture We have argued elsewhere (e.g., Costa, 2005; Costa and Laskey, 2006), that there is a need to represent declarative knowledge about likelihood in domain ontologies

In environments in which noisy and incomplete information is the rule, likelihood information is

a key aspect of domain knowledge Furthermore, much of the key semantic content needed to en-able interoperability involves information about plausibility For this reason, we have argued, knowledge about likelihoods should be included

in formal domain ontologies

This viewpoint is not universal A argument to our position is that probability is inherently epistemic, whereas formal ontology should represent phenomena as they exist in the world Carried to its extreme, however, this philosophical stance would preclude the use of virtually every ontology that has yet been devel-oped To explore this idea further, we note that

counter-if computational ontologies had existed in the 17th century, Becher and his followers might well have developed an ontology of phlogiston

We may chuckle now at their nạveté, but who among our 17th century predecessors had the foresight to judge which of the many scientific theories then in circulation would stand the test of time? Researchers in medicine, biology, defense, astronomy, and other communities have developed

a plethora of domain ontologies It is virtually certain that at least some aspects of some of these ontologies will, as human knowledge progresses, turn out in retrospect to be as well founded as the theory of phlogiston Shall we outlaw use of all these ontologies until the day we can prove they contain only that which is ontological, and noth-ing that is mere epistemology? Moreover, many aspects of our common, shared knowledge of

these domains are inherently probabilistic

Well-established statistical regularities are a key

ele-ment of expert reasoning A principled means of

representing these probabilistic aspects of domain

knowledge is needed to facilitate interoperability

and knowledge sharing

Similar questions arise with the representation

of vagueness Fuzzy logic has been applied

exten-sively to problems of reasoning with imprecisely

defined terms For example, fuzzy reasoning

might be applied to retrieve and sort responses

to a query for “inexpensive” patio furniture A

fuzzy reasoner would assign each furniture set a

degree of membership in the fuzzy set

“inexpen-sive,” and would sort the retrieved sets by their

membership in the fuzzy set There is an

analo-gous question of whether it is legitimate to extend

ontology formalisms to allow representation of

fuzzy membership values, or whether fuzziness

is inherently epistemological and does not belong

in an ontology

There is a valid, important, and as yet

unre-solved philosophical clash between those who

believe that we live in a deterministic world in

which uncertainty is entirely epistemic, and those

who believe the world includes phenomena that

are ontologically stochastic and/or imprecise and

should be represented as such From an

engineer-ing standpoint, we cannot wait for the debate to be

resolved before we move forward with building

and using ontologies

Although our ultimate scientific objective

is to seek the truth about reality as it is, this

ultimate objective is unattainable in the lifetime

of any human Therefore, no “perfect ontology

of all things” is reachable, regardless of one’s

philosophical view on uncertainty Nevertheless,

from a pragmatic perspective, it is necessary and

desirable to do the best we can with the knowledge

we have, even if this causes the ontology to be

under-specified due to incomplete information

Formal ontology provides a useful means of

com-municating domain knowledge in a precise and

shareable manner, and of extending and revising

our descriptions as human knowledge accrues Accepting only complete knowledge would leave

us with too little information to solve most of the interesting problems that ontologies are capable

of addressing

Not surprisingly, as ontology engineering research has achieved a greater level of maturity, the need for uncertainty representation and reason-ing for the Semantic Web has become more and more clear Correspondingly, interest is increasing

in extending traditional ontology formalisms to include standard mechanisms for representing and reasoning with uncertainty Whether the ultimate consensus is that ontology formalisms should be capable of representing information about un-certainty, or that ontologies should represent the space of possibilities and that information about uncertainty should be conveyed in a different semantic layer, principled means of representing and reasoning with uncertainty are increasingly seen as necessary

Uncertainty in Rule Languages A related

stream of research has focused on augmenting SW

rule languages to handle uncertainty (Damásio et al., 2006; Lukasiewicz, 2005, 2006; Lukasiewicz &

Straccia, 2007) Although there is as yet no standard rule language for the Semantic Web, the W3C’s Rule Interchange Format (RIF) Working Group has recently released working draft documents specify-ing use cases, requirements, and a core design for

a format that allows rules to be translated between rules languagesa The use cases and requirements document does not mention uncertainty, but the core design mentions the need to translate between rule languages that handle uncertainty, and makes brief mention of syntactic and semantic implications of the need to treat uncertainty This brief treatment

is far from sufficient to address the full range of issues that need to be addressed to achieve semantic interoperability between systems that express and reason with uncertainty For space reasons, we do not address rule language research in detail in this chapter We note, however, that augmenting ontolo-gies to express uncertainty generates a requirement

to augment rule languages to take advantage of the

information expressed in uncertainty-enhanced

ontologies

Towards a Pragmatic View Apart from the

historical and philosophical issues, as research

on SW leaves the conceptual level and reaches a

level of maturity in which the actual challenges

are better understood, realization has grown that

many SW applications require a principled means

for representing uncertainty and performing

plausible reasoning with incomplete data As the

interest in uncertainty representation techniques

grows, the focus of SW shifts from

philosophi-cal issues toward “down to earth” engineering

issues Important challenges are to identify the

kinds of information management problems that

would benefit most from mechanisms for

reason-ing with uncertainty, to assess the scalability of

uncertainty representation approaches, to evaluate

the suitability of different forms of

representa-tion and reasoning to solve specific use cases,

and others

This pragmatic, focused view has pushed

researchers from many different domains of

knowledge into an appreciation of the need for a

forum to discuss the ways in which uncertainty

reasoning can contribute to addressing their

re-spective challenges, and to evaluate the strengths

and weaknesses of different approaches to

repre-senting and reasoning under uncertainty Although

uncertainty-related papers were sometimes

pre-sented in other venues, the first forum explicitly

geared towards answering the above issues was

the workshop on Uncertainty Representation for

the Semantic Web (URSW workshop), held in

conjunction with the Fourth International

Seman-tic Web Conference (ISWC 2005) The intention

of the URSW workshop was to provide an open

forum to all forms of uncertainty representation

and reasoning, without being prejudicial in favor

of any particular approach At the second

work-shop (URSW 2006), a consensus was reached

that the most important tasks were (1) to develop

a set of use cases for uncertainty in the SW; and

(2) to assess how each approach (or combination

of approaches) would address appropriate lenges set out in the use cases In the end, a much improved understanding of those issues would led to identification of best practices involving uncertainty reasoning in the SW

chal-The strong interest in the URSW and similar venues prompted the W3C to create, in March

2007, the Uncertainty Reasoning for the World Wide Web Incubator Group (URW3 XG), with the objective of better defining the challenge of working with incomplete knowledge The URW3 adopted the same “approach-independent” stance

as the URSW, with an initial focus on the problem itself rather than a particular approach to solving

it At the time of this writing, the URW3 is tively pursuing its development of use cases, and planning for a third URSW is underway The next two sections present a brief view of the major ap-proaches for uncertainty in the SW being discussed

ac-in fora such as the URW3 and URSW

PROBABILISTIC APPROACHES

TO UNCERTAINTY IN THE SEMANTIC WEB

Bayesian probability provides a mathematically sound representation language and formal cal-culus for rational degrees of belief, which gives different agents the freedom to have different beliefs about a given hypothesis This provides a compelling framework for representing uncertain, incomplete knowledge that can come from diverse agents Not surprisingly, there are many distinct approaches using Bayesian probability for the Semantic Web

Bayesian knowledge representation and reasoning systems have their formal basis in the axioms of probability theory (e.g., Ramsey, 1931; Kolmogorov, 1960/1933) Probability theory al-lows propositions to be assigned truth-values in the range from zero, meaning certain falsehood,

to one, meaning certain truth Values intermediate

between zero and one reflect degrees of likelihood

of a proposition that may be either true or false

Bayes Rule, a theorem that can be derived from the

axioms of probability theory, provides a method

of updating the probability of a proposition when

information is acquired about a related proposition

The standard format of Bayes rule is:

On the right side of the formula, P(A) is called

the prior probability of A, and represents our belief

in event A before obtaining information on event

B Likewise, P(B) is called the prior probability

of B There is also P(A|B), which is the likelihood

of event A given that event B has happened On

the left side of the formula there is P(B|A), which

is the posterior probability of B, and represents

our new belief in event B after applying Bayes

rule with the information collected from event A

Bayes rule provides the formal basis for the active

and rapidly evolving field of Bayesian probability

and statistics In the Bayesian view, inference is a

problem of belief dynamics Bayes rule provides

a principled methodology for belief change in the

light of new information

Bayesian Networks (BNs) BNs provide

a means of parsimoniously expressing joint

probability distributions over many interrelated

hypotheses A Bayesian network consists of a

directed acyclic graph (DAG) and a set of local

distributions Each node in the graph represents

a random variable A random variable denotes

an attribute, feature, or set of hypotheses about

which we may be uncertain Each random variable

has a set of mutually exclusive and collectively

exhaustive possible values That is, exactly one of

the possible values is or will be the actual value,

and we are uncertain about which one it is The

graph represents direct qualitative dependence

relationships; the local distributions represent

quantitative information about the strength of

those dependencies The graph and the local

dis-tributions together represent a joint probability distribution over the random variables denoted

by the nodes of the graph

Bayesian networks have been successfully applied to create consistent probabilistic represen-tations of uncertain knowledge in diverse fields

Heckerman et al (1995) provide a detailed list

of recent applications of Bayesian Networks The prospective reader will also find comprehensive coverage of Bayesian Networks in a large and growing literature on this subject, such as Pearl (1988), Neapolitan (1990, 2003), and others Figure 1 shows an example of a BN representing part of a highly simplified ontology for wines and pizzas

In this toy exampleb, we assume that domain knowledge about gastronomy was gathered from sources such as statistical data collected among restaurants and expertise from sommeliers and pizzaiolos Moreover, the resulting ontology also considered incomplete knowledge to establish a probability distribution among features of the pizzas ordered by customers (i.e type of base and topping) and characteristics of the wines ordered

to accompany the pizzas

Consider a customer who enters a restaurant and requests a pizza with cheese topping and

a thin and crispy base Using the probability distribution stored in the BN of Figure 1, the waiter can apply Bayes rule to infer the best type of wine to offer the customer given his pizza preferences the body of statistical and expert information previously linking features

of pizza to wines Such computation would be difficult when there are very many features Bayesian networks provide a parsimonious way

to express the joint distribution and a tationally efficient way to implement Bayes rule This inferential process is shown in Figure

compu-2, where evidence (i.e., the customer’s order) was entered in the BN and its result points to Beaujolais as the most likely wine the customer would order, followed by Cabernet Sauvignon, and so on

P(B)