Identifying User Groups and Their ChallengesThe ''''Cookery'''' AI Chatbot cooking assistant is designed to cater to a diverse user population with varied cookinginterests, skill levels, time
Introduction
Scenario Introduction and Problem Definition
In the wake of the global pandemic, one significant lifestyle transformation has been our relationship with food. Dining habits worldwide have been affected, with a distinctive shift towards home-cooked meals This trend is not merely anecdotal but is supported by hard data According to a Nielsen report (2020), the worldwide home cooking rate rose by 60% during the pandemic, and Vietnam mirrored this trend with a surge in home cooking documented by VNExpress (2020).
The rise in self-cooking, while indicative of a healthier and potentially more satisfying approach to food, has also underscored some considerable challenges The process of self-cooking is often time-consuming and complex, necessitating a myriad of decisions such as what to cook, procuring the ingredients, and following a recipe accurately Moreover, considerations around dietary restrictions, nutritional balance, and personal taste preferences add another layer of complexity to the task This scenario provides a compelling opportunity to investigate how technology, specifically AI-Chatbot technology, could be used to improve this situation,making the process of self-cooking more streamlined, enjoyable, and accessible.
Proposed AI Solution
To address these challenges, we propose a solution: 'Cookery', an AI-Chatbot cooking assistant This tool aims to provide users with personalized recipe recommendations based on their preferences and available ingredients, offer easy-to-follow cooking instructions, and answer any food-related queries in real time In the subsequent sections of this report, we detail the design, development, and evaluation plan of our proposed solution.
The ultimate objective is to improve the overall self-cooking experience, by making it more streamlined,engaging, and less stressful, all of which aligns with our commitment to enhancing daily life through innovative AI technology.
User Analysis and Task Definition
Identifying User Groups and Their Challenges
The 'Cookery' AI Chatbot cooking assistant is designed to cater to a diverse user population with varied cooking interests, skill levels, time constraints, and dietary needs Identifying the different groups within this population, along with their distinct goals, both generic and specific, and challenges can significantly contribute to tailoring the AI-Chatbot's functionalities and features to enhance these users' cooking experiences.
Appendix A provides detailed information about the four primary user groups that are likely to interact with'Cookery': Novice Cooks, Health-Conscious Cooks, Busy Professionals, and Culinary Enthusiasts Their characteristics, values, requirements, goals, and potential challenges that they might face are elaborated inTable
Personas:We have developed comprehensive personas to represent four main user groups: Novice Cooks,Health-Conscious Cooks, Busy Professionals, and Culinary Enthusiasts, providing a deeper understanding of their unique challenges and requirements, as shown inFigure 1,Figure 2,Figure 3, andFigure 4in Appendix B.
Task Analysis and AI Solution Application
To maximize the effectiveness of the 'Cookery' AI-Chatbot cooking assistant across our diverse user base, a comprehensive task analysis is crucial This involves delineating the specific tasks that each user group needs to accomplish and understanding how our AI solution can be employed to facilitate these tasks.
Detailed information on the tasks for each user group, along with the corresponding applications of the'Cookery' AI solution, is provided in Appendix C, as outlined inTable 5.
Usability Requirements and Target Levels
For the prototype stage of the 'Cookery' AI-Chatbot cooking assistant, it is essential to establish usability requirements and target levels These benchmarks cater to each user group and serve as a developmental guide, helping to shape 'Cookery' as it progresses.
Detailed information on the usability requirements and target levels for each user group can be found inAppendix D, as outlined inTable 6.
Prototype Design and Specification
Prototype Overview
The 'Cookery' AI-Chatbot prototype focuses on testing AI's effectiveness in delivering personalized recipe recommendations versus traditional search methods The goal is a tailored, efficient cooking experience based on individual preferences.
The prototype works by offering a selection of cooking preferences Users input their choices, and the AI Chatbot uses this to suggest personalized recipes Users can review and refine these recommendations, leading to a more precise recipe.
Our prototype, built on a static website framework using HTML, CSS, and JavaScript, delivers an interactive and authentic testing experience A minimalist design approach and controlled input options emphasize the Chatbot's efficiency in providing personalized recipes.
Please note, this prototype is a testing phase product for evaluating AI Chatbot performance, not a final product.It's a key step towards user needs comprehension and solution refinement Here is the link to our prototype webpage: https://snitcoding.github.io/Cookerynoserver/ Please refer to Figure 5 in Appendix E for a demonstration of the ‘Cookery’ prototype main page.
Design Specification
The design of our ‘Cookery' AI-Chatbot cooking assistant prototype centers on functionality We aim to provide a user interface that supports easy interaction with the AI tool, focusing on its operation and benefits, while also prioritizing accessibility, readability, and user comfort.
The following provides an overview of the key UI design principles applied to the prototype:
Font Family:We selected Arial, sans-serif, a universally readable font compatible with diverse devices and operating systems, in line with our design objective.
Font Sizes:We use a font size of 1.2rem for general text and 2.5rem for headers, ensuring legibility across various screen sizes and resolutions Headers stand out, guiding user navigation.
Line Height:Set at 1.6 to enhance readability, this spacing ensures clear differentiation between text lines.
Accessibility and Comfort:We've prioritized accessibility, ensuring optimal text-background contrast, and incorporated hover/focus outlines for clear visual cues.
Our design's simplicity allows users to focus on the tool's functionality, while its accessibility and comfort are not compromised As the prototype evolves, these principles will be refined based on user feedback, reinforcing our commitment to enhancing the 'Cookery' user experience. b UX Design Strategy
Our User Experience (UX) design strategy for the 'Cookery' AI-Chatbot aims to provide a streamlined and user- friendly environment The design facilitates simplified interaction and delivers personalized recipe recommendations.
Users are presented with nine predefined cooking reference options, encompassing a variety of cooking preferences, to facilitate easy understanding and accessibility These options include four select boxes, three number inputs, and two free-text inputs, all of which are optional, allowing users to customize their input based on their comfort.
After users define their preferences, the AI Chatbot generates a tailored recipe Users can fine-tune the recipe through additional instructions to the Chatbot, iteratively refining it until it aligns with their preferences.
Our UX design caters to users of varying cooking skills and culinary knowledge, ensuring even those with minimal experience can successfully use the tool Ultimately, the design aims to make recipe discovery and cooking more enjoyable and less intimidating for all users. c AI Integration Methodology
To enhance the authenticity of our prototype during testing, we've integrated it with the reputable ChatGPT AI,
7 selected after thorough analysis OpenAI's comprehensive documentation and resources have been instrumental in our effective use of the ChatGPT API and in crafting bespoke prompts.
We've adopted a pure JavaScript implementation for ChatGPT integration, prioritizing cost-efficiency This method, involving direct API calling and dynamic webpage text addition, removes the need for extra dependencies like NodeJS or Python and their required libraries.
Additionally, this methodology facilitates straightforward deployment on GitHub Pages, bypassing the need for server or domain setup, or complex procedures This choice balances AI integrity and functionality with smooth development, highlighting our dedication to an effective, accessible, user-centric solution.
Interactive Prototype Evaluation
Usability Assessment
In the process of conducting a comprehensive evaluation of our interactive prototype, we have utilized the Heuristic Evaluation method Developed by Jakob Nielsen, this method of usability inspection provides a set of established principles for evaluating the interface design of interactive systems (Nielsen, 1994) Here, we present a detailed Heuristic Evaluation of our prototype based on Nielsen's Heuristics:
Visibility of system status:The Cookery AI Chatbot provides immediate feedback on the users' actions. Whenever a user inputs a command or a request, the chatbot responds promptly, keeping users informed about what's happening.
Match between the system and the real world:The language used by the Cookery AI Chatbot is simple, clear, and familiar The use of culinary terms and the presentation of the recipes align with the real-world cooking experience.
User control and freedom:Users can easily navigate through their conversation with the AI-Chatbot, edit their preferences, or exit the chat at any point There's a strong sense of user control and freedom.
Consistency and standards:The user interface of the Cookery website maintains a consistent design throughout The language and interaction patterns remain uniform across the platform.
Error prevention:The Cookery system is designed with a structured and controlled input format, which reduces the likelihood of user errors Moreover, the AI-Chatbot can handle a variety of user inputs without generating errors.
Recognition rather than recall:The prototype supports recognition over recall by presenting users with selectable options for their cooking preferences, rather than requiring them to remember and type their choices.
Flexibility and efficiency of use:The Cookery prototype caters to both novice and experienced users The system allows users to specify their preferences in detail, leading to a more personalized and efficient experience.
Aesthetic and minimalist design:Cookery follows a minimalist design philosophy The user interface is clean and simple, focusing on the essential features and minimizing potential distractions.
Help users recognize, diagnose, and recover from errors:In case of unexpected user input, the AI- Chatbot provides a polite and clear error message, guiding users to input correctly.
Help and documentation:The Cookery prototype provides a brief guide on how to interact with the AI- Chatbot, ensuring that users have the necessary information to use the system effectively. b Cognitive Walkthrough
The task of finding a personalized recipe using the Cookery AI-Chatbot is the most suitable task for a detailed cognitive walkthrough It is pertinent to a broad user spectrum, and it comprises several steps, potentially revealing a wide array of usability aspects The task will be segmented into a series of steps onTable 7in Appendix F, which will be evaluated using The Four Questions of Cognitive Walkthrough as articulated Wharton et al (1994):
Will the user try to achieve the right effect?
Will the user notice that the correct action is available?
Will the user associate the correct action with the effect they're trying to achieve?
If the correct action is performed, will the user see that progress is being made toward the solution of their task?
Test Scenario:Consider a user named Alex, a 25-year-old novice cook who wants to prepare a meal with the ingredients he has at home He decides to use the Cookery AI-Chatbot to find a recipe that matches his preferences and the ingredients he has on hand.
Alex has access to a computer with an internet connection.
Alex is familiar with using web browsers.
Alex is fluent in English and is comfortable navigating websites for information.
Identification of Flaws or Misconceptions
The prototype's current design presents several areas of potential confusion or difficulty for users, particularly those who are new or less experienced with such platforms The task steps within the workflow lack clear visual
9 markers or distinct delineations, such as sequential numbering or explicit indications of progress This absence could lead to users feeling overwhelmed or confused, particularly given the numerous input options available on the screen As a result, users may navigate through the form more slowly, and their understanding of the functionality might be compromised This lack of clarity could contribute to unnecessary time consumption and potentially undermine the user experience.
Furthermore, the prototype does not provide clear indicators of the user's current progress within the form The absence of a visible indication of progress may leave users uncertain about whether their inputs have been successfully registered or if modifications to the recipe have been adequately implemented This uncertainty could lead to user dissatisfaction and a perceived lack of control over the process.
While the aim of providing a variety of cooking references is to enhance user convenience, some of these options might be unclear to users without a certain level of culinary knowledge Even though all fields are optional, the need for users to conduct additional research to understand these cooking-related aspects could introduce an unintended complexity into the system.
Finally, the prototype incorporates the ChatGPT Chatbot, renowned for its flexibility and versatility However,its unrestricted free text input capability might offer users an unintended degree of freedom, potentially allowing them to interact with the system in ways that were not originally envisaged This broad scope of interaction could lead to misuse or exploitation of the system, deviating from its intended purpose of personalized recipe generation.
Empirical Evaluation and Testing
Preparation
Our empirical evaluation and testing process incorporates five participants, representative of the four primary user groups identified in our user analysis This approach allows our evaluation to reflect a diverse set of user experiences, expectations, and behaviors.
Two participants, a student and a young professional aged 16-35, represent the 'Novice Cooks' group Their shared attributes include less confidence in cooking, a desire to learn, and time constraints Their primary challenges involve understanding ingredients, following complex instructions, and efficient meal planning. The 'Health-Conscious Cooks' group is represented by a professional aged 25-65+, interested in maintaining a balanced diet and finding recipes that meet specific dietary needs.
The fourth participant, a 'Busy Professional' aged 25-60, prioritizes efficiency and convenience due to work commitments, and often seeks quick, healthy meal options.
The final participant, aged 18-65+, symbolizes the 'Culinary Enthusiasts' group Passionate about cooking and
10 exploring new recipes, this participant often seeks unique or specific cuisine recipes and appreciates advanced cooking techniques. b Consent Form
Our commitment to ethical principles and guidelines in the context of user research involving human participants is paramount Consequently, we have developed a detailed Consent Form This document outlines the specifics of the study, what participants can expect, and how we will prioritize their confidentiality and anonymity throughout the research process Please follow the attached link to access and review the full content of the Consent Form:Consent Form Document. c Task Guide Documents
To facilitate a systematic and thorough testing process, a comprehensive task guide document has been prepared. This document delineates the step-by-step procedure to be followed during the evaluation of the Cookery AI Chatbot, ensuring consistency across all test sessions For a detailed understanding of the testing protocol, please refer to the link provided:Test Guide Document d Tasks Preparation
Before proceeding with the empirical evaluation of the Cookery AI Chatbot, it's crucial to have a systematic plan for task preparation This helps to ensure that the user testing is effective and that we gather the most valuable and relevant information possible about the user experience Further elaboration on each task, its purpose, and expected outcomes from the participants is available inTable 8in Appendix G. e Roles, Facilitators, and Responsibilities
For the smooth functioning of the user testing process of the Cookery AI Chatbot, each member has been designated a specific role These roles are crucial for systematically and efficiently conducting the evaluation,and each role has distinct responsibilities For a detailed overview of each role, its responsibilities, and the assigned team members, please refer toTable 9in Appendix G.
Evaluation
The evaluation is to be conducted in a controlled environment, free from external interruptions and noise This ensures that the participants can focus entirely on interacting with the Cookery AI Chatbot A computer with a stable internet connection is required, and the participants will be able to access the prototype through the provided URL All necessary preparations including the prototype launch and printing of consent forms are managed by Tin Huynh. b Testing Methodology
Our testing methodology employs a combination of established user testing techniques to gain an insightful and
11 comprehensive understanding of the user experience with the Cookery AI Chatbot Here are the primary methods used:
Think-Aloud Protocol: Participants are encouraged to think aloud as they interact with the Cookery AI Chatbot.
This technique enables us to gain insights into the participant's thought processes, understand their expectations, and identify any confusion or challenges they encounter while performing the tasks.
Task-Based Testing: Participants perform a sequence of 12 interrelated tasks designed to mimic typical user interactions with the AI system Each task evaluates a different functionality of the Cookery AI Chatbot, such as its ability to suggest recipes based on user preferences, manage dietary restrictions, work within budget constraints, and more The purpose of task-based testing is to assess the tool's effectiveness in a practical, real- world context.
Concurrent Observation: As participants interact with the chatbot, observers carefully document their actions, reactions, difficulties, and successes This includes both verbal feedback through the Think-Aloud protocol and non-verbal cues like hesitation or confusion.
Post-task Interviews:After each task, participants are asked to share their feedback on the task, the chatbot's responses, and the overall interface This feedback is critical for understanding their experiences, identifying potential improvements, and assessing the user-friendliness of the tool. c Execution of Testing
1 Introduction and Consent:The test begins with a brief introduction from the facilitator (Subject Handler), explaining the purpose of the test and reassuring the participant that it's the system being evaluated, not them. The facilitator then presents the consent form and ensures that the participant understands and agrees with its content before signing.
2 Task Explanation and Prototype Interaction:The facilitator (Prototype Manager) hands out the task document to the participant, explaining each of the tasks that they will be performing The participant is then directed to the AI chatbot interface, where they will interact with the system and perform the outlined tasks. Throughout the interaction, participants are encouraged to think aloud and express their thoughts, feelings, and questions.
3 Observation and Data Tracking:During the interaction, the Data Tracker keeps a careful record of the participant's actions, reactions, struggles, and successes They note down any areas where the participant seems to be confused or makes errors This concurrent observation provides critical insights into the system's usability.
4 Post-task Discussion and Debriefing:After each task, the Subject Handler asks the participant questions related to their experience with that task, allowing them to provide immediate feedback At the end of all tasks, the Subject Debriefer conducts a more in-depth interview with the participant, discussing their overall
12 experience, any difficulties they encountered, and any improvements they suggest.
5 Evaluation Completion:Once the interview is completed, the participant is thanked for their time and valuable feedback All collected data, including observation notes, task completion rates and times, participant responses, and video recordings, are then gathered and prepared for the analysis stage.
We've recorded one of our testing sessions, the video provides a detailed and realistic perspective of how the evaluation unfolded, the interaction of the participant with the Cookery AI Chatbot, and how our team handled various responsibilities during the evaluation process Here’s the link to the video: https://www.youtube.com/watch?v=G9eaDRJmMrs d Findings
Following the execution of the evaluation, the next crucial step is to collate, examine, and analyze the collected data This process enables us to gather vital insights about the participants' interactions with the Cookery AI Chatbot, understand their experiences, and identify the strengths and potential areas for improvement of the AI tool The findings section will broadly categorize the results into two types:
Quantitative Findings:This will primarily involve the numerical data obtained during the evaluation process. Metrics such as task completion rate, task completion time, error rate, and other measurable elements fall under this category This data provides concrete evidence of the tool's performance, usability, and efficiency from a numerical standpoint.
Qualitative Findings:This segment of findings will encompass the subjective responses and feedback obtained from the participants Observations made during the 'Think Aloud' protocol, responses to the post-test interviews, and subjective opinions on the AI tool's interface, functionality, and overall experience belong to this category.
Data Analysis and Insights
Quantitative Data Analysis
For an in-depth understanding of the user interactions and performances with our AI Cooking Assistant, refer to Table 10in Appendix H This detailed evaluation showcases essential metrics like task completion time, error rate, and misconception rate, providing valuable insights for optimizing the system's user experience. b Concepts Evaluation
Our analysis of the test data also included a systematic evaluation of three key concepts: completion time, error rate, and misconception rate These parameters were derived from our initial test design and played a significant role in assessing the system's performance across different user categories A summary of our findings is presented inTable 11below in Appendix H Please refer to it for a detailed analysis.
Qualitative Data Analysis
In addition to our performance testing, we employed a qualitative data collection method to gain deeper insights into the user experience Specifically, we utilized an online survey via Google Forms, which enabled us to gather detailed feedback from our users The survey form can be accessed at this link:Survey
The responses we received provided valuable insights into user sentiment and perceptions about our AI cooking assistant Please refer toFigure 6,Figure 7,Figure 8 Figure 9 Figure 10, , ,Figure 11,Figure 12, andFigure 13in Appendix I for an overview of the responses we collected. b Face-to-Face Interview Approach
In order to gather more personal and nuanced information about user impressions and usability, we conducted a series of face-to-face interviews Common responses to our questions underscored both strengths and potential areas for improvement in our design For a summary of the questions asked and the common responses we received, please refer toTable 12in Appendix I.
Post-Testing Heuristic Analysis Findings and Conclusions
Upon concluding our user testing, we conducted a thorough heuristic analysis based on Nielsen's principles. This review incorporated both our perspective as designers and the feedback received from users to provide a holistic evaluation of the AI Cooking Assistant's usability and design.
Aesthetic and Minimalist Design:The interface's minimalist design was well-received by users, who appreciated the low-fidelity approach and the straightforward presentation of information.
Ease of Use:The low technical skill requirement for operating the interface was another strength Users could navigate the interface and utilize its functionalities effectively with just basic mouse and keyboard skills.
Efficiency:The system demonstrated efficiency in remembering old inputs and allowing users to select options and adjust values using keyboard shortcuts, thereby speeding up the input process.
However, we also identified several areas where our interface fell short of Nielsen's heuristic principles:
Visibility of System Status:Users expressed difficulty in identifying when the AI had completed the recipe
14 generation, particularly due to the lack of automatic scrolling or notifications The absence of clear indications about the user's progress also compounded this issue.
Help and Documentation:The need for more detailed explanations and help was apparent, especially regarding specific food references and inputs like 'Umami' for taste and currency for budget Users required additional information to utilize these functionalities effectively.
User Control and Freedom:The interface lacked options in certain contexts, such as taste preferences and complexity of meals This restricted users' freedom and control, potentially affecting their overall satisfaction with the system.
In light of these findings, we can conclude that while our AI Cooking Assistant has many strengths, there are crucial areas of improvement that need to be addressed to enhance the overall user experience These insights will guide our subsequent modifications and improvements to the system.
Modifications Based on Evaluation
Identify Modifications
The identification of usability issues and their subsequent modifications are crucial to improve the overall user experience and performance of our AI Cooking Assistant To ensure a systematic approach, we used the Nielsen Severity Rating Scale This scale ranges from 0 to 4 to rate the severity of usability problems:
0 = I don't agree that this is a usability problem at all
1 = Cosmetic problem only: need not be fixed unless extra time is available on project
2 = Minor usability problem: fixing this should be given low priority
3 = Major usability problem: important to fix, so should be given high priority
4 = Usability catastrophe: imperative to fix this before product can be released
Our analysis of user interactions and responses during the testing phase resulted in the identification of five main issues onTable 13in Appendix J.
Plan for Implementing Modifications
Addressing the identified issues in a systematic and efficient manner is key to optimizing the user experience. Here's our proposed plan for implementing modifications:
Developing Guidance and Documentation:We will augment the system with comprehensive guides and in- line help text for each input field This includes clarifying currency and time units for Budget and PreparationTime, respectively Contextual help icons can be added next to these input fields, providing users with necessary guidance as they interact with the system.
Enhancing Taste Choice Descriptions:To help users make informed choices, we will provide clear and detailed descriptions for each taste option This could be implemented as a hover-over tooltip text or a help icon next to each choice, offering users an understanding of what each taste implies in a cooking context.
Introducing Translation Support:To bridge the language gap, we plan to integrate a translation feature, especially for ingredient names and dietary restrictions This will involve mapping common ingredients and dietary terms to their equivalents in various languages We might also consider auto-suggest features to facilitate user inputs in this context.
Clarifying Complexity Levels:To ensure users understand the complexity levels, we will include examples or definitions for 'simple', 'intermediate', and 'advanced' levels These definitions will be presented in easy-to- understand language and displayed in a way that does not clutter the interface, such as hover-over tooltip text.
Improving Interface Interactions:To make the interface more interactive, we propose to add a loading icon while the system processes the recipe We also plan to introduce automatic scrolling to direct users to the result once it's generated These changes aim to provide visual feedback to users about system operations and ensure they do not miss any critical outputs.
Reflections
Team’s Strategy
Our team decided to divide our process into 4 phases namely project planning, prototype designing,prototype developing, and empirical evaluation phase as detailed inTable 14,Table 15,Table 16, andTable 17respectively in Appendix J
Our Strengths, Weaknesses & Challenges
Our team's strength lies in the diverse skill set and extensive experience of our members Hau, Nam, Tin, Duong, and Nhan all bring unique talents to the table Hau, Nam, and Tin are proficient in website development. Nam and Tin, particularly, have substantial experience with AI technologies, and their insights were instrumental in effectively integrating AI Chatbot technology into our project Tin took the lead on many of the technical tasks, providing crucial input and guidance that shaped the project's success On the analytical side, Duong and Nhan excelled in brainstorming sessions, providing substantial input to the project's structure By utilizing our individual strengths and supporting each other's weaknesses, we achieved success in our project.
Despite our strengths, we faced a range of challenges throughout the project The initial stage was particularly challenging as we struggled to understand the project's requirements and objectives This necessitated numerous drafts and consultations with our lecturer As most of us were more familiar with developing complete products
16 rather than prototypes, defining the concepts of low fidelity and high fidelity and integrating AI Chatbot technology into our project interface was a significant hurdle Practical issues like understanding how to conduct an empirical evaluation test and the unexpected delays from test users added to the challenges.
Communication was another challenge, with missed meetings and occasional misunderstandings, despite our use of Facebook and Messenger for team communication Additionally, the need to make our meetings more efficient became apparent as some of them ran long with little outcomes due to distractions or members' unavailability.
Lessons Learned and Future Applications
Through the project, we gained a deeper understanding of the effective use of Chatbot AI, specifically ChatGPT.
We learned how to elicit the desired responses and how to incorporate such technology into our applications or websites These are valuable skills that we can use in future projects.
We also improved our teamwork skills, with a particular focus on task specialization and efficient communication We grappled with issues like miscommunication and last-minute procrastination, especially during the report preparation phase, which often led to delays These experiences underscored the importance of clear communication, efficient time management, and working cohesively as a team.
Looking ahead, the knowledge and skills gained from this project will undoubtedly be beneficial in future endeavors The insights we gained into AI Chatbot technology present exciting opportunities for developing more interactive and intelligent applications Furthermore, the experiences and lessons we've gained regarding teamwork and project management will serve us well in our future group projects.
Nielsen, J (1994) 'Heuristic evaluation', in Nielsen, J and Mack, R.L (eds.) Usability Inspection Methods. John Wiley & Sons.
Nielsen, J (1994) 'Severity Ratings for Usability Problems', Nielsen Norman Group Available at: https://www.nngroup.com/articles/how-to-rate-the-severity-of-usability-problems/(Accessed: 14 May 2023).
NielsenIQ (2020) 'How has Covid-19 impacted Vietnamese consumers', NielsenIQ Available at: https://nielseniq.com/global/en/insights/analysis/2020/how-has-covid-19-impacted-vietnamese-consumers/ (Accessed: 14 May 2023).
Quy, N (2020) 'Covid-19 impact: Vietnamese rediscover joy of eating at home', VNExpress Available at: https://e.vnexpress.net/news/news/covid-19-impact-vietnamese-rediscover-joy-of-eating-at-home-
Wharton, C., Rieman, J., Lewis, C., & Polson, P (1994) 'The Cognitive Walkthrough Method: A Practitioner'sGuide', in Nielsen, J and Mack, R.L (eds.) Usability Inspection Methods John Wiley & Sons.
Appendix A: User Groups Novice Cooks
Socio-Economic Status All ranges
Education Basic to higher levels
Careers Students, young professionals, individuals living independently
Personal Characteristics May be time-pressed, less confident in cooking, eager to learn
Values Learning new skills, gaining independence, value convenience
Requirements Easy-to-follow recipes, detailed instructions, meal planning assistance
Generic Goal Learning to cook
Specific Goals Preparing a specific dish or cuisine
Lack of cooking skills or knowledge leads to difficulty in preparing meals.
Limited understanding of ingredients and their uses.
Difficulty in understanding complex cooking instructions.
Challenge in planning meals and shopping for groceries efficiently.
Time management issues due to lack of experience.
Table 1 ‘Novice Co oks ’ user group
Socio-Economic Status All ranges
Education Higher levels generally, but can also be varied
Careers Professionals, health enthusiasts, fitness trainers
Personal Characteristics Interested in nutritional balance, might have specific dietary needs, committed
Values Healthy eating, fitness, wellness
Requirements Nutritional information, recipes catered to specific dietary needs, meal planning assistance for balanced diets Generic Goal Maintain or improve health
Specific Goals Find and prepare nutritionally balanced meals.
Adhere to specific dietary needs or restrictions
Difficulty in finding recipes that meet specific dietary requirements.
Struggle to calculate nutritional values and balance meals appropriately.
Limited options or ideas for healthy recipes.
Challenge in understanding the health benefits/risks of certain ingredients or cooking methods.
Ta ble 2 ‘Health-Conscious Cooks ’ use r group
Socio-Economic Status Middle-class or above
Education Higher levels generally, but can also be varied
Careers Professionals across different fields, entrepreneurs
Personal Characteristics Time-pressed, may value convenience, might have limited time for meal preparation
Values Efficiency, time management, convenience
Requirements Quick and easy recipes, meal prep ideas, suggestions for healthy and quick snacks Generic Goal Save time in meal preparation
Specific Goals Prepare meals ahead of time.
Discover quick and easy recipes
Limited time for meal preparation due to work commitments.
Difficulty in finding quick and healthy meal options.
Challenge in efficiently planning and preparing meals in advance.
Lack of time to explore new recipes or cooking techniques.
Ta ble 3 ‘Busy Professionals ’ user grou p
Socio-Economic Status All ranges
Education All levels, may have specific culinary education
Careers Can be varied, including chefs, food bloggers, food critics, or simply individuals who love cooking
Personal Characteristics Passionate about cooking, interested in exploring new recipes and cuisines, may have advanced cooking skills Values Creativity, culinary art, diversity in food
Requirements Wide variety of recipes, unique and authentic recipes, advanced culinary techniques Generic Goal Explore culinary arts
Difficulty in finding unique, authentic, or specific cuisine recipes.
Need for advanced cooking techniques which might not be easily available or understandable.
The challenge of finding rare or specific ingredients required for certain dishes.
Desire for a platform to share and discuss culinary experiences and techniques with a like-minded community.
Table 4 ‘Culinary Enthusiasts ’ user group
Fi gure 1 Pe rsona for ‘Novice Cooks’
Fi gure 2 Pe rsona for ‘Health-Conscio us Cooks ’
Fi gure 3 Pe rsona for ‘Busy Pr ofessionals ’
Fi gure 4 Persona for ‘ Culinary Enthu siasts’
Appendix C: Task Analysis and AI Solution Application
User Groups Tasks AI Solution Application
Finding easy-to-follow recipes suitable for beginners.
'Cookery' can recommend beginner-friendly recipes with detailed, step-by-step instructions. Understanding the basics of meal planning and grocery shopping.
The AI can provide guidance on meal planning and generate shopping lists based on selected recipes
Gaining confidence in executing cooking techniques and following complex instructions.
It can also offer real-time assistance and clarification on cooking techniques and instructions.
Learning to manage time efficiently in the kitchen.
'Cookery' can suggest time estimates for each cooking step to help users manage their cooking process efficiently.
Finding recipes that meet specific dietary requirements or restrictions.
'Cookery' can provide personalized recipe recommendations based on dietary needs or restrictions. Calculating nutritional values and ensuring balanced meals.
It can also calculate and display nutritional information for each recipe to help users track their dietary intake.
Discovering new and varied healthy recipes to prevent meal monotony.
By accessing a vast database of recipes, 'Cookery' ensures users have a wide array of healthy recipes to choose from
Finding quick and healthy meal options suitable for their hectic schedules.
'Cookery' can recommend quick, easy, and healthy recipes tailored to busy lifestyles.
Planning and preparing meals in advance to save time.
It can assist in meal prep planning,allowing users to prepare multiple meals in advance.
24 Exploring new recipes that can be prepared quickly and easily.
The AI can offer a variety of new recipes that require minimal prep time and ingredients, keeping mealtime exciting even for those with limited time.
Finding unique, authentic, or specific cuisine recipes.
'Cookery' can provide a wide range of unique and authentic recipes from various cuisines.
It can demonstrate advanced cooking techniques through detailed instructions.
Finding rare or specific ingredients required for certain dishes.
'Cookery' can also provide information about specific ingredients, including where to find them and possible substitutions.Table 5 Table for Ta sk Analysis and AI Solution Application
Appendix D: Usability Requirements and Target Levels
User Groups Usability Requirements Target Levels
Simplicity: The prototype should be intuitive and easy to navigate.
Clarity: Recipes and instructions need to be clear and easy to understand.
Support: The prototype should provide some level of assistance for any queries or issues.
During testing, novice cooks should be able to navigate the prototype and find a suitable recipe within a reasonable time.
Users should be able to follow the instructions without excessive confusion.
Personalization: The prototype should be able to provide some level of recipe recommendations based on dietary needs.
Information: Basic nutritional information for each recipe should be readily available.
Variety: A reasonable variety of healthy recipes should be available.
During testing, health- conscious cooks should find the personalization feature somewhat helpful in choosing recipes.
Nutritional information should be reasonably accurate and available for the majority of the recipes.
Efficiency: The prototype should aim to provide quick and easy recipe recommendations.
Planning: Basic meal prep planning and time-saving features should be available.
Accessibility: The prototype should be accessible on various devices and platforms for on-the- go usage.
During testing, busy professionals should be able to find a suitable recipe within a reasonable time.
Users should find the meal prep planning and time- saving features somewhat useful.
Diversity: The prototype should provide a variety of unique and authentic recipes.
During testing, culinary enthusiasts should be reasonably satisfied with the variety and authenticity of
26 of advanced cooking techniques and detailed instructions should be available. recipes.
Users should find the advanced features and community sharing options somewhat useful and engaging.
Table 6 Table for Usability Requirements and Tar get Levels
Fi gure 5 Screenshot of ‘Cookery’ prot otype ma in pag e
Will the user try to achieve the right effect?
Will the user notice that the correct action is available?
Will the user associate the correct action with the effect they're trying to achieve?
If the correct action is performed, will the user see that progress is being made toward the solution of their task?
Yes, the user will attempt to do this action as it's the first step to use the service.
Yes, the website URL is clearly visible and accessible.
Yes, the URL is connected to the website.
Yes, the website's homepage will be displayed.
Select cooking preferences and input available ingredients
Yes, the user will try this as it's needed to generate a personalized recipe.
Yes, the selection boxes and input fields are visible on the website.
Yes, the fields are labeled appropriately indicating what should be entered.
Yes, upon input, the entered information is displayed, indicating progress.
Yes, the user will try this to get the recipe based on the inputs provided.
Yes, the 'Get Recipe' button is visible and clearly labeled.
Yes, the label 'Get Recipe' is indicative of its function.
Yes, the user will see a loading indicator or a similar sign to indicate the process is ongoing.
Provide feedback or request modifications to the recipe
Yes, if the user wants to adjust the recipe, they will attempt to do this action.
Yes, the prototype provides a way to communicate with the AI Chatbot.
Yes, it's a common interaction pattern to provide feedback or ask for changes in chatbots.
Yes, the user will see the AI Chatbot's response, indicating the progress.
Yes, this is the end goal of the user.
Yes, after the user's feedback, the AI Chatbot
Yes, it's the expected outcome after providing
Yes, the final recipe is displayed,indicating
28 will provide the final recipe. feedback completion of the task.
Table 7 Table for Cognitive Walkthrough
Appendix G: Empirical Evaluation Preparation Task
1 The interface asks about the user's cooking expertise level.
To allow the AI to suggest meals with appropriate difficulty levels (skill requirements, pre-requisite knowledge in cooking).
All subjects are expected to comprehend this task, despite the term "Expertise" not being explicitly stated as "Cooking Expertise".
2 The interface asks about the desired complexity of the meal.
To enable the AI to suggest meals of suitable complexity (variance of ingredients, number of steps).
Although it is anticipated that all users will understand this task, there may be some ambiguity about the term "complexity", as perceptions of complexity may differ among users.
3 The interface prompts users to include their preferred ingredients.
To enable users to incorporate preferred ingredients into their meal.
While users should understand the task, there may be challenges with users typing ingredient names incorrectly or inputting too many ingredients for the AI to process.
4 The interface asks about the user's budget for each portion of the meal.
To ensure the cost of the meal suggested by the AI aligns with the user's financial situation.
Users may encounter difficulties due to the interface only accepting numeric input, currency differences, and possible typing errors, such as including "k" for thousand or "$" for US dollars.
5 The interface asks about the desired meal preparation time.
To enable the AI to suggest recipes with appropriate cooking methods and dish types.
Potential issues may arise from users inputting time in hours instead of minutes, and variations in perceived cooking time due to factors like cooking skill and familiarity with the recipe.
The interface requests information on the serving scale (number of people or meals).
To allow the AI to suggest the appropriate quantity of ingredients.
While most subjects should be able to complete this task, there may be minor issues in differentiating adult and child portions.
7 The interface asks about the user's nutritional goals and preferences.
To enable the AI to suggest recipes that align with their fitness goals.
All subjects are expected to complete this task effectively.
8 The interface asks about any dietary restrictions.
To allow the AI to exclude any harmful or unwanted ingredients.
All subjects are expected to efficiently complete this task.
9 The interface prompts the user to submit the form.
To enable the AI to process user requirements.
All subjects are expected to effectively complete this task.
The user evaluates the AI-generated recipe and may suggest changes if necessary.
To facilitate user modifications if required.
While most subjects should complete this task effectively, more careful or indecisive users may require more time to evaluate the recipe.
11 The user may suggest changes to the final recipe if necessary.
To ensure the final recipe meets the user's needs.
All subjects are expected to successfully complete this task, although the variety of potential responses could be wide and challenging to predict.
Ta ble 8 Tab le for Tasks Prep aration
1 Sets up the testing prototype
2 Distributes task documents to participants Tin
1 Provides consent forms and test policies to participants
2 Introduces the purpose of the test
3 Asks about the participant's background
4 Guides participants through the different stages of testing
1 Takes detailed notes on participants' responses and behaviors
2 Records key performance indicators such as completion rates and time
Subject Debriefer 1 Handles the distribution and collection of post-test surveys
2 Conducts face-to-face interviews with participants HauTable 9 Table for Roles, Facilitato rs, and Re sponsibilities
Table 10 Tab le for Use rs’ Pe rformance Test
Concept Measuring method Methods Subject Average test result
Target level for each task
Duration from when the subject starts and finishes a task
The subject failed to finish a subtask, so they require explanations or guidance
How many times does the final recipe result dissatisfies the user based on the categories listed in the sub-tasks? Or the subject wrongly understands the text description over the input field
Table 11 Tab le for Con cepts Eva luation
Qualitative Data Analysis
Fig ure 6: Re sp on ses on descri ption s of input fields
Figu re 7: Responses on misunderstanding of the tasks
Figure 8: Responses on available options
35 Figur e 9: Responses on tasks’ options
Figure 10: Responses on the improvement of options
Figur e 11: Responses on AI ch atbot
36 Fig ure 12: Respo ns es on translation an d language barrier
Fig ure 13: Respo ns es on overall experience
What is your first impression of our interface?
It is a simple, low-fidelity interface that is simple to use, but the design is a little bit simple.
Do you have any issues when trying to understand the task?
Misunderstood the currency of “Budget” of VND instead of US dollars. Misunderstood the input field of “Preparation time” to be hours instead of minutes.
Does not understand what “Umami” means as a choice in “TastePreference”
37 Does our website require a high level of fluent use of English for
Most user does not know how to translate some common tropical vegetables or fruits in Vietnamese into English to type in the “Ingredient” field, so they need the interface to support this feature.
Does our website friendly to users with low computer skills?
Most of the website functionality only require basic mouse click and typing skill
Would you use our products or suggest to your friend our product when it is completed and launched on the market?
All users love the idea of Cookery.
But most users expect better visualization, design, and color.
Table 12 Tab le for Face- to-face Intervi ew Approach
Appendix J: Modifications Based on Evaluation
Users were unclear about the currency of the budget or the units of preparation time (e.g., minutes or hours)
Users were unsure about what certain options, like ‘umami’ or
‘sour’, precisely mean This could lead to incorrect input and consequently, unsatisfactory outputs from the AI system
Absence of Translation Support for Ingredients and Dietary
Users needed translation support, especially for tropical fruits, vegetables, and dietary restrictions
Users needed clear examples or interpretations of complexity levels like ‘simple’,
Lack of modifications to enhance the user’s experience like a loading icon or automatic scrolling once the AI finishes processing
Ta ble 13 Table for Modifications Identification
Propose the project idea Duong
Team management (task allocation, set up communication channel) Duong
Write the design specification Tin, Nam
Table 15 Pr ototype De signing phase
Develop the prototype using HTML, CSS, JavaScript Tin, Nam, Hau
Manage the Github repository Nam, Tin
Integrate Chat GPT Open AI’s API Tin
Table 16 Pr ototype De signing phase
Conduct the Empirical Evaluation All members
Analyze the Empirical Evaluation All members
Write the report All members