Researchers are beginning to recognize that leadership is an important part of complex, dynamic systems and that understanding how a complex system func- tions across time is necessary to understand leadership effects. Uhl-Bien and colleagues (Hazy & Uhl-Bien, 2014; Marion & Uhl-Bien, 2001) popularized the examination of complex, dynamic systems in the leadership field, although the idea of a systems approach has its origins in a number of prior theories, including Katz and Kahn’s (1978) influential work on open systems theory, theories that extend outwardly from an organization to its external environ- ment (Hooijberg, 1996), and theories that reach inwardly to include followers (Shamir, 2007; Uhl-Bien, Riggio, Lowe, & Carsten, 2014) and groups (Grand, Braun, Kuljanin, Kozlowski, & Chao, 2016).
The hallmark of complexity is that the interaction of diverse system parts produces complexity (Page, 2007), and this interaction can unfold over time at multiple levels of an organization (Simon, 1962). Furthermore, interactions may be produced by feedback processes in which one part of a system accepts or modi- fies an input originating from a different system component or the environment.
As such processes reverberate across system units, new structures and constructs
can emerge without any central direction – perhaps even without leadership. To illustrate how this interaction process works, consider that complexity resulting from a conscious integration of many local processes within the human brain often creates an entirely new construct (Dehaene, 2014) – a process called com- pilational aggregation (Dinh et al., 2014; Kozlowski & Klein, 2000). This entire process takes only a few hundred milliseconds. At the other end of the time spectrum, the emergence of new organizational strategies or identities from inter- actions within a complex dynamic system can take a decade (Hazy, 2008). Yet leadership does have an impact on complex systems, albeit often indirectly and through its influence on the rules or processes that guide interactions in the system (Hazy & Uhl-Bien, 2014).
It is helpful to recognize that dynamic systems adapt to changes in the exter- nal or internal environments by moving through time to a new point of stability that is called an attractor. This movement, however, can be erratic, as the system changes and explores a new landscape. Thus movement over time is not neces- sarily linear, and the movement to a new point of stability is often preceded by a period of high variability or high entropy (Stephen, Dixon, & Isenhower, 2009).
Leaders should recognize and perhaps contribute to this process, even though we might classically think that a leader’s goal would be to decrease variability and the uncertainty that often accompanies it. At times, high variability may be needed to allow a system to find new solutions, but such variability can come from any level or part of a system, and it may involve processes that operate on many timescales. Typically, emergence from lower levels creates this variability and allows systems to explore, and eventually move to, new attractors.
In short, complex adaptive systems comprise a cacophony of processes and time-dependent cycles, each influencing a part of the system, but often within different time frames. These time-dependent processes all aggregate to affect the way in which a system functions. This richness is missed when researchers take a retrospective approach to understanding leadership, such as by asking about leadership traits or styles in surveys, because surveys typically reflect combined perceptions across unspecified periods of time. Even longitudinal research that captures information about specific points in time often also misses much of this picture because the data collection often will pick up only those processes whose cycles have certain time lags (for example, with measurements at 3, 6, and 9 months) and will miss processes with longer or shorter cycles.
To move from leadership theories focused on entities and antecedents to a more process-oriented understanding, a more comprehensive approach to thinking about how to treat time in the study of leadership processes is needed.
Table 8.1 identifies multiple domains that have been investigated as dynamic systems in the past, and it indicates a relevant time frame for processes to unfold within each domain. The nature of leadership processes may be quite different across these domains, operating automatically at faster time frames (< 500ms), or more slowly at conscious levels (≥ 500ms).
As Dinh, Lord, and Hoffman (2014) noted, the very fastest processing speeds depend on the type of architecture that is guiding information processing, with embodied architectures operating as fast as 5ms; emotional architectures, as fast as 10ms; connectionist architectures completing processing within 200–500ms;
and symbolic architectures requiring at least 500ms. What this means is that because leaders or leadership researchers are often focused on a single level in Table 8.1, they can miss many unrecognized and unintended effects that oper- ate at lower levels, and perhaps also at levels that are much higher and operate much more slowly. For example, when theory is focused at the level of goal emergence, perhaps being concerned with a leader’s effects on the regulatory focus of goals (that is, promotion versus prevention goals), then the effects of much faster-acting emotions or the effects of much slower group processes may be missed.
How leadership occurs, then, naturally depends on the domain in which influences are operating and the time frame of interest. When focusing on intrapersonal processes, leaders can prime various automatic processes to create an influence that seems instantaneous. For example, leaders, either intentionally or unintentionally, can infect others via emotional contagion processes that operate in a few hundred milliseconds, and they can prime different levels of follower self-concepts by their choices of pronouns (Lord
& Brown, 2004).
TABLE 8.1 Dynamic systems domains and characteristic time frames
Domain Time frame Researchers
Emotion 5–50ms Metcalf & Mischel (1999); Kuppins,
Oravecz, & Tuerlinckx (2010)
Information access 10–30ms Newell (1990)
Person construal 200ms Freeman & Ambady (2011)
Consciousness 300ms Dehaene (2014)
Information processing
5ms–10sec Bassett & Gazzaniga (2011); Dinh, Lord, & Hoffman (2014) Goal emergence 250–500ms Carver & Scheier (2002) Goal orientation 250–500ms DeShon & Gillespie (2005) Leadership emergence Minutes–hours Acton, Foti, Lord, & Gladfelter
(2018); DeRue (2011)
Identity construction Minutes–weeks Ashforth & Schinoff (2016); Lord, Gatti, & Chui (2016)
Group process Minutes–days Grand et al. (2016)
Personality Months–years Read et al. (2010); Mischel & Shoda (1998)
Organizational complexity
Months–decades Marion (1999); Hazy (2008) Leadership niches Centuries–millennia Spisak et al. (2015)
Interaction processes and time frames are quite different at the dyadic and group levels. For example, social processes such as claiming and granting lead- ership (DeRue & Ashford, 2010) may provide the building blocks of dyadic interactions. Leadership acts become part of the communication–behavior cycles that Weick (1979) terms “double-interacts,” as an individual’s leadership actions are accepted or rejected by dyadic partners, which in turn feeds back to the original actor, who might adjust their behavior accordingly. The effects of these double-interacts can aggregate, as various dyadic structures influence each other, allowing group-level structures to emerge (Acton et al., 2018; DeRue, 2011). Typically, such processes can occur within a few hours, or perhaps over a few days, but they could develop over longer periods of time – although research suggests that dyadic relations tend to form quickly in newly acquainted leader–follower dyads (Liden, Wayne, & Stilwell, 1993), and we would expect group structures also to stabilize relatively quickly.
When the company hierarchical level is the focus, leadership processes can take much longer than days or weeks – sometimes even requiring years or decades, as was the case with Intel, for example. As described by Hazy (2008), leadership processes at Intel were less obvious because they involved the catalyz- ing of adaptive, dynamic processes, rather than the imposition of organizational strategies in a top-down manner. Specifically, management at Intel created an organizational structure that encouraged and rewarded new products and strate- gies in a bottom-up way. They did this by allocating resources to products based on their success in the marketplace (profit margins), allowing successful products to flourish, while less-successful products gradually declined. Further, because this strategy was outwardly focused, there were many cycles of products and market feedback involved in this decade-long change process, which moved Intel from a memory chip to a microprocessor company. Thus the effects of leadership on system functioning were diffuse and distributed over time. For research purposes, a formal computational system is often used to represent such processes, because they may take years to actually observe (Hazy, 2008; Hazy
& Uhl-Bien, 2014).
Leadership processes at each of the levels in Table 8.1 require unique theo- ries and methodologies, yet there is also a fractal quality in that, at each level, one system component influences another, and this influence plays out through actions and feedback that occur over time. But such process may be opaque to leadership researchers, unless they learn the theories and methodologies appro- priate to each level. This is a daunting task, and many researchers may merely see leadership as a personal, trait-like quality, rather than as part of more com- plex, time-dependent, hierarchical systems. Although such trait views can have merit, they may miss the dynamics that produce effective system outcomes.
Thus a theory of leadership that fully explains phenomena must move beyond person-centered leadership perceptions to embrace the multilevel, multi-time, process-oriented approach of complexity theory.
One heuristic strategy, which is based on an insight from levels-of-analysis research offered by Yammarino, Dionne, Chun, and Dansereau (2005), is to also examine the time frames and leadership processes immediately above and below one’s theoretical focus. As they suggest, this may spur scientific advance- ment. The reason why this strategy makes sense in a dynamic system is that there are exchanges of information and resources across many hierarchical levels in dynamic systems. In general, higher levels set constraints on lower-level systems, such that group cultures or climates may constrain individual group members’
behaviors, while it is the aggregation of lower level inputs that often combines to create a group outcome, such as the individual inputs that occur when a group is problem solving. Similarly, a higher-level structure from companies or divisions may constrain group processes and structures, yet the company level may depend on the combined products from multiple groups. A critical ques- tion in understanding leadership at the group level, then, is whether it is the upward or the downward exchange of information and resources that is most critical. This question cannot be answered without examining three levels at once. This can be done without much trouble in terms of statistical analysis (see Schaubroeck et al., 2012, for an example), but, as we have suggested, it is more challenging from a processing perspective, because individual, group, and organizational levels often involve different time spans. Thus, when aggregating from individuals to groups, it is not only aggregation across individuals that is important, but also aggregation across time.
What is needed to advance our understanding is a theoretical system that maps how time, cycles, social and organizational processes, and leadership processes change at various hierarchical levels. Jaques (1989) discussed such differences in terms of “the time span of discretion,” which means the amount of time it takes workers at various hierarchical levels to have effects, but this work is rarely cited by contemporary leadership research, even though it has continued to have an impact on system design and consulting. In contrast, most multilevel leadership theory applies statistical approaches to examine how a particular process varies across hierarchical levels. This is a good start, but the recommendation here goes beyond that by recognizing that the nature of dynamic systems function- ing varies with hierarchical levels (see Table 8.1). Moreover, these factors may interact with other aspects of the medium of time that were noted at the start of the chapter. One approach to addressing this complexity may be to incorporate time as a focal variable in agent-based modeling (Castillo & Trinh, 2018).