Tackling social loafing:
An academic framework using ChatGPT simulations for group performance optimization

NOTE: A part of this on-going work was presented as a contributed oral paper to the
11th International and 12th National Teachers and Education Students' Conference (TEStCon 2024)
Cebu Normal University, Cebu City, 18-19 November 2024

INTRODUCTION

An academic framework for simulating learners' behaviors while conducting practical tasks within a group will be a valuable tool for educators and students of education  because it addresses a common challenge in group-based learning: social loafing. Understanding how factors like task visibility and workload fairness influence group dynamics can help educators design better collaborative activities. By leveraging AI-driven simulations, the framework offers a new data-driven approach to enhancing student engagement, accountability, and learning outcomes in team projects. It introduces innovative strategies that can be applied across various disciplines, making it highly relevant for both current and future educators.

Why is this interesting to the academia?

This framework will be interesting to educators and students of education for several reasons:

1. Improving Group-Based Learning: Group work is a central part of many educational programs, particularly in project-based learning environments. However, managing group dynamics and ensuring equitable participation is often difficult. Social loafing directly impacts both student learning outcomes and the perceived fairness of group work. Educators are consistently looking for ways to design collaborative tasks that promote accountability and maximize learning. This study’s insights into reducing social loafing can directly inform how educators structure group work, making it more effective and equitable.

2. Practical Application for Educators: The framework and methodology proposed in this study, which uses ChatGPT-driven simulations, provide practical, evidence-based strategies that educators can apply in real-world classroom settings. It offers an innovative tool for anticipating and mitigating the effects of social loafing before they arise in actual student groups. For teachers, having a data-driven method to test different group configurations or task designs would enable better management of group dynamics, improving overall classroom outcomes.

3. Personalized Learning and Fairness: Social loafing often leads to the “sucker effect,” where students who perceive they are doing more work than others reduce their effort to match their peers. By exploring how task visibility, workload fairness, and group size affect loafing, the study offers solutions for educators to foster an environment where all students are motivated to contribute equally. This can lead to a more personalized learning experience, where individual contributions are recognized and rewarded, enhancing fairness and student satisfaction.

4. AI-Driven Pedagogical Innovation: Using ChatGPT as part of a multi-agent simulation system introduces educators to the potential of artificial intelligence in education. This shows how AI can be integrated into curriculum design, classroom management, and pedagogy to simulate and solve complex human behavior problems. This approach allows educators to experiment with different group scenarios in a risk-free environment, empowering them to create better collaborative learning environments.

5. Research-Driven Educational Practices: Educators are increasingly expected to ground their teaching practices in research. This study links artificial intelligence (AI), behavioral science, and education, providing educators with a robust research framework to understand how group behaviors impact learning. Implementing evidence-based strategies that address social loafing would make educators’ teaching practices more effective and supported by research.

6. Future-Proofing Teaching Techniques: For educators and those training to be teachers, learning how to manage group-based learning effectively is a skill that will only become more important as collaborative work continues to grow in prominence. This study equips future teachers with cutting-edge tools to address long-standing issues in group learning, helping them stay ahead of pedagogical challenges.

By focusing on the intersection of AI and collaborative learning, the framework not only addresses a key challenge in education but also offers a forward-thinking approach that can shape future teaching strategies.

What is the background of social loafing research?

Recent academic literature on measuring the perception of social loafing in group-based learning underscores the complexity of evaluating individual contributions within collaborative settings. Studies reveal that perceptions of social loafing are influenced by factors such as task visibility, group size, and the fairness of task distribution. Tools like self-report surveys and observational methods are commonly used to assess students' awareness of uneven effort among peers. Research highlights that social loafing can negatively impact motivation, group cohesion, and overall learning outcomes. To mitigate these effects, scholars recommend strategies such as enhancing task visibility, fostering procedural and distributive justice, and promoting clear communication within teams. This body of literature provides valuable insights for educators aiming to design and manage group activities that balance workload distribution and encourage active participation from all members.

What is the background of our proposed solution?

Large Language Models (LLMs), particularly ChatGPT (OpenAI, 2024), have shown significant potential in simulating human behaviors within multi-agent system setups, contributing to research on team dynamics and social interactions. Studies highlight that LLMs can emulate complex social behaviors such as cooperation, negotiation, and conflict resolution by modeling varied personality traits and cognitive processes. ChatGPT's adaptability allows it to simulate multi-agent interactions that reflect real-world roles and psychological constructs, making it a valuable tool for studying group dynamics like social loafing, social facilitation, and leadership influence. Researchers have used LLMs to explore group task performance, communication patterns, and decision-making, underscoring their applicability in educational and organizational contexts to enhance understanding and management of collaborative behavior.

What is presented in this work?

Presented in this work is the methodology for implementing our proposed academic framework. The methodology involves a multi-step approach that begins with defining a clear research objective and structuring a simulated group task relevant to software engineering. The process includes the identification of agent roles, characterized by key personality traits like conscientiousness, agreeableness, and extraversion, which influence behaviors such as dominance, passivity, fairness, or loafing. The next step involves setting up an environment using a LLM-based Chatbot (Mauldin, 1994; Caldarini, et al., 2022) such as ChatGPT to simulate team interactions, with predefined prompts guiding responses according to the assigned traits and roles. Following the simulation, data is collected and analyzed to evaluate interactions, decision-making, and task outcomes, with attention to social dynamics such as social loafing and facilitation. Finally, results are validated by comparing them with real-world data or established literature, providing insights into group-based behaviors and potential applications in educational settings for understanding collaborative learning dynamics.

REVIEW OF LITERATURE

Definition of Terms

Basing our insights from our prior works (Pabico, 2008a,b; Recario, et al., 2015) and related (possibly seminal) academic sources, the following are brief definitions for each terminology used in this presentation:

• Distributive justice: The perception of fairness in the allocation of rewards or outcomes (Liden, et al., 2004).

• Free riding: When a group member benefits from the collective efforts of the team without contributing their fair share (Albanese & Van Fleet, 1985).

• Procedural justice: The perception of fairness in the processes and procedures used to make decisions and distribute resources in a group (Greenberg, 1990).

• Ringelmann effect: The observation that individual productivity decreases as the size of the group increases due to coordination and motivational losses (Kravitz & Martin, 1986).

• Social facilitation: The tendency for individuals to perform better on simple or well-learned tasks when in the presence of others (Bond & Titus, 1983; Cook, 2001).

• Social loafing: The phenomenon where individuals exert less effort when working in a group compared to when working alone (Latane, et al., 1979; Williams & Karau, 1991; Comer, 1995).

• Sucker effect: The tendency of individuals to reduce their effort in a group setting to avoid being perceived as the "sucker" who over-contributes (Kerr, 1983).

• Sucker role: The perception of a team member who feels they are over-contributing and being taken advantage of by others in the group (Kerr, 1983).

• Task visibility: The degree to which an individual's contribution to the group task can be identified and recognized by others (Kidwell & Bennet, 1993).

Social Loafing Research

The following syntheses from literature illustrate how group dynamics, task design, and accountability mechanisms interact to influence social loafing in educational settings:

1. Team Composition and Group Dynamics

   • The composition of teams greatly impacts the occurrence of social loafing. Self-selected teams, where students form groups with friends, often exhibit more social loafing due to groupthink and diminished accountability (Chapman, et al., 2006; Waletzko, 2021). Conversely, randomly assigned teams tend to foster more diverse ideas and reduce loafing, though they may initially experience dissatisfaction and concerns with group cohesion (Chapman, et al., 2006). 

   • Group cohesiveness, the degree to which students feel connected to their peers, is crucial. When team members feel a strong bond and accountability to the group, they are less likely to engage in social loafing (Karau & Williams, 1997).

   • Pabico, et al., (2008b) explored how agents in cooperative groups, if allowed to self-organize and incentivize cooperation, can effectively reduce the inclination to shirk responsibilities, a phenomenon akin to social loafing. This insight highlights the potential for dynamic group structures to enhance participation.

2. Group Size and Social Loafing

   • Larger groups are more prone to social loafing, as individuals may hide their lack of contribution in bigger teams. Research shows that members of large groups often exert less effort, whereas smaller teams demand more individual participation (Aggarwal & O’Brien, 2008; Waletzko, 2021). 

   • Team-Based Learning (TBL) supports larger teams (5-7 members) for complex tasks that require diverse perspectives, reducing loafing when the tasks are designed to engage all members. However, larger teams must have structured tasks that make individual efforts indispensable to the group’s success (Waletzko, 2021).

   • Recario, et al. (2015) investigated large-scale collaborations and suggested that certain hierarchical or decentralized structures could influence individuals’ commitment, mitigating the risks of free-riding, which parallels social loafing.

3. Task Design and Accountability

   • Tasks that clearly define each member’s responsibilities help reduce social loafing. When individual contributions are visible and assessable, students are less likely to disengage from group work (Aggarwal & O’Brien, 2008; Pfaff & Huddleston, 2003). 

   • Peer feedback and structured assessments increase accountability by making contributions transparent to both peers and instructors, thereby reducing loafing behaviors (Pfaff & Huddleston, 2003).

   • The dynamic evolution of cooperative behaviors in group settings demonstrates that when tasks require high collaboration and each agent’s contribution is essential, social loafing is minimized. This underscores the importance of task design that necessitates mutual cooperation (Pabico, et al., 2008b).

4. Impact on Student Satisfaction

   • Social loafing negatively impacts student satisfaction with group work, as students perceive uneven contributions and unfairness in grading. This dissatisfaction can lower group performance and individual motivation (Aggarwal & O’Brien, 2008).

   • Structured team assessments and transparent workload distribution improve both team effectiveness and student satisfaction, leading to better learning outcomes (Pfaff & Huddleston, 2003).

   • Recario, et al. (2015) supports the idea that structured, cooperative environments can lead to higher engagement and satisfaction, as individuals feel a greater sense of fairness and responsibility in decentralized collaborations.

5. Reducing Social Loafing

   • Long-term teams, where students work together throughout a semester, can reduce loafing by enabling members to establish defined roles and responsibilities. Over time, these teams can develop stronger cohesion and accountability, reducing the likelihood of social loafing (Waletzko, 2021; Karau & Williams, 1997).

   • Peer evaluations and continuous feedback further reduce social loafing by encouraging personal responsibility and ensuring that all members are contributing meaningfully (Chapman et al., 2006; Aggarwal & O’Brien, 2008).

   • Pabico, et al., (2008a,b) highlights how flexible organizational structures and adaptive incentives can create environments where loafing is less likely, as group members dynamically adjust to ensure mutual benefit.

Simulating Human Behaviors with LLMs

1. Social Simulation Platforms and Frameworks

   •  GenSim (Tang, et al., 2024) and AgentVerse (Chen, et al., 2023) are frameworks for using LLM-based agents to facilitate collaboration and social interaction simulations. These platforms support realistic interactions in multi-agent environments and are particularly effective for modeling group behaviors, social dynamics, and task-based collaboration.

   • Ghaffarzadegan, et al. (2024) introduced generative agent-based modeling (GABM) frameworks, merging LLMs with mechanistic models to study social systems. This framework provides flexibility, combining the predictability of mechanistic models with the adaptability of generative AI to simulate social scenarios.

2. Autonomy and Exploration in Virtual Environments

   • Wang, et al. (2023) presented Voyager as an autonomous agent capable of exploring open-ended virtual environments using GPT-based processing. Voyager’s success in navigating complex, evolving scenarios shows LLMs’ potential for dynamic decision-making, self-guided exploration, and emergent behavior in controlled settings.

   • Generative Agents by Park, et al. (2023) illustrates how LLM-powered agents can independently plan, interact, and form relationships in simulated social spaces, modeling complex social networks and individual autonomy.

3. Multi-agent Collaboration and Communication

   • AgentVerse (Chen, et al., 2023), with its multi-agent LLM framework, offers insights into collaborative dynamics within group-based learning. Agents share and negotiate information, work towards common goals, and showcase how LLMs enhance collaboration by understanding and responding to contextual nuances.

   • S3: Social-Network Simulation System (Gao, et al., 2023) simulates social interactions in online social networks, focusing on communication, negotiation, and emergent social dynamics in a scalable manner.

4. Modeling Human Social Behaviors and Trust

   •  Xie, et al. (2024) explored how LLM agents can replicate human-like trust behaviors, using economic models like Trust Games. They suggested that LLMs can emulate nuanced human interactions, making them applicable for understanding interpersonal dynamics in trust-dependent simulations.

   • Generative Agents (Park, et al., 2023) also demonstrates social interaction modeling, where agents engage in behavior patterns typical of human societies, like forming relationships and coordinating in group settings.

5. Behavioral Patterns and Social Dynamics in Organizational and Economic Contexts

   • Ghaffarzadegan, et al. (2024) integrated mechanistic models with LLMs in a GABM framework to simulate organizational behaviors and social norm diffusion.

   • Xie, et al. (2024) analyzed how LLMs in multi-agent setups can be used to replicate and study trust behaviors relevant to organizational and economic contexts, broadening their applicability in social sciences.

6. Ethical and Methodological Implications of LLM-powered Agents: As discussed by Ghaffarzadegan, et al. (2024), the ethical implications of using LLM-powered agents in social simulations raise questions about authenticity, representational accuracy, and potential biases. This underscores the need for rigorous validation to ensure simulations align with ethical standards, especially in educational or organizational scenarios.

Roles in Software Engineering Teams

In a software engineering team, a minimum of five major roles is typically needed to cover essential functions. These roles help to balance responsibilities for planning, coding, testing, and managing the software development process. The following roles represent a foundational structure for software development teams, especially in agile environments, where roles may also adapt based on project needs (Pressman & Maxim, 2019; Sommerville, 2015):

1. Project Manager (PM): Responsible for planning, organizing, and overseeing the project timeline, resources, and communication. The PM coordinates between team members and stakeholders, ensuring the project aligns with goals and deadlines.

2. Product Owner (PO): Manages the product vision, gathers requirements, and represents the stakeholders' needs. They prioritize tasks and ensure the team builds features that meet user expectations.

3. Software Developer/Engineer: Builds and implements the actual code. This role might include front-end, back-end, or full-stack developers who work on the technical implementation of the project.

4. Quality Assurance (QA) Tester: Focuses on testing the software to find and document bugs. They work closely with developers to verify that the software meets specified requirements and functions correctly.

5. UX/UI Designer: Responsible for creating a user-friendly interface and designing an intuitive experience for the end user. They work closely with developers to ensure the product is visually appealing and easy to navigate.

Personality Traits Influencing Behaviors in Organizations

Personality traits that influence behaviors in group-based tasks, like those of software engineering teams, are often based on the Big Five Personality Traits model, widely recognized in organizational psychology (Barrick & Mount, 1991; Hogan & Holland, 2003; Deyoung, et al., 2007). Studies show that three core traits tend to have a strong influence on group dynamics and productivity. Below lists these three core traits with the behaviors exhibited by these traits: 

1. Conscientiousness: Defined as a tendency toward self-discipline, organization, and dependability. Individuals high in conscientiousness are often reliable, goal-oriented, and motivated to meet group objectives, making them key contributors to task management and quality control in group projects.

   • Fair: Individuals high in conscientiousness are often fair, as they value dependability and following established norms. Their tendency toward reliability promotes fair contribution to group tasks.

   • Loafing Tendency: Those low in conscientiousness may exhibit loafing tendencies, as they may lack the motivation or sense of responsibility to contribute equally to group efforts.

2. Extraversion: Characterized by sociability, assertiveness, and enthusiasm, extraverted individuals contribute to group cohesion and communication. They are often comfortable leading discussions, sharing ideas, and encouraging team interaction, which is beneficial for collaborative tasks like brainstorming or requirements gathering.

   • Dominant: Extraverted individuals are often more assertive and willing to take on leadership roles, which can manifest as dominance in group settings. They are usually comfortable voicing opinions and may steer the direction of discussions.

   • Passive: In contrast, individuals low in extraversion (more introverted) might display passive behaviors, preferring to take a backseat in group discussions and contributing when directly asked or prompted.

3. Agreeableness: Refers to a tendency for cooperation, trust, and empathy. People high in agreeableness are usually team players who help foster a collaborative environment, reduce conflicts, and support peers. This trait is crucial for maintaining harmony and motivating others in a team setting.

   • Fair: High agreeableness is closely associated with fairness, as agreeable individuals value harmony, empathy, and a cooperative approach, often ensuring that everyone’s contributions are acknowledged and valued.

   • Loafing Tendency: Low agreeableness can contribute to loafing behaviors, as individuals who are less cooperative may avoid participating fully, especially if they perceive little personal benefit in the group’s success.

FRAMEWORK FOR SIMULATING SOCIAL LOAFING WITH CHATGPT

This framework integrates multi-agent simulation into a structured research design to test the conclusions of our earlier studies (Pabico, et al., 2008a,b; Recario, et al., 2015) in a controlled environment. By replicating various aspects of these studies and analyzing the results quantitatively and qualitatively, we can confirm whether the findings hold in simulated settings.

1. Simulation Setup

   1. Define Agent Characteristics:

      • Agent Roles: Identify different roles within the group (e.g., Project Manager, Project Owner, Software Developer/Engineer, etc.).

      • Agent Traits: Assign each agent a personality trait based on behaviors such as dominance, passivity, fairness, or loafing tendencies. Traits should influence decision-making and interaction styles.

   2. Task Assignment:

      • Choose a simple software engineering project with clear deliverables.

      • Define tasks with varying complexity, requiring collaboration.

      • For each agent, determine workload and visibility of their contributions.

2. Simulation Phases

   1. Initial Setup:

      • Form Groups: Divide agents into groups, ensuring different mixes of roles and traits.

      • Task Visibility Levels: Assign different levels of task visibility to groups (e.g., low visibility where contributions are anonymous vs. high visibility where everyone can track individual contributions).

   2. Define Group Dynamics:

      • Low Task Visibility Group: Limit the visibility of individual efforts, simulating conditions that lead to social loafing.

      • High Task Visibility Group: Increase transparency by making individual contributions observable to all, creating accountability.

3. Simulate Interactions Using ChatGPT:

   1. Inter-agent Interactions:

      • Run Dialogues: Use ChatGPT to simulate group discussions and behaviors. For each scenario, agents will interact based on assigned roles and personalities.

      • Agent Responses: Program ChatGPT to model different reactions depending on their roles. For example:

        1. Social Loafer: Contributes minimally, providing excuses for lack of effort.

        2. Leader: Motivates the group, assigns tasks, and checks on others’ progress.

        3. Dominant: Criticizes or dominates conversations, pushing for their ideas.

        4. Fair Contributor: Balances input and tries to ensure the project is on track.

   2. Group Performance Simulation:

      • Task Progress: At each stage (e.g., planning, coding, testing), track how much work is completed by each agent.

      • Adjust Behavior: Based on simulated feedback (e.g., leaders pushing loafers to contribute more), adjust the behavior of agents during later phases.

4. Collect and Measure Data

   1. Quantitative Data:

      • Number of tasks completed per agent.

      • Time spent on tasks.

      • Bugs/issues identified and fixed by each agent.

      • Task visibility percentage (how visible each agent’s contributions are).

      • Proportion of work distribution (how evenly tasks are distributed).

   2. Qualitative Data:

      • Simulated agent feedback on perceived workload fairness.

      • Simulated perceptions of loafing or dominance.

      • Simulated feedback on overall group satisfaction and dynamics.

5. Experimental Design

   1. Control and Experimental Groups:

      • Control Group: No task visibility, unequal workload distribution (simulate typical conditions that lead to social loafing).

      • Experimental Group: High task visibility, fair workload distribution (test the conditions identified in the original study).

   2. Replicate with Multiple Groups: Simulate the scenario multiple times with different group compositions to test robustness and validate patterns.

6. Analysis

   1. Quantitative Analysis:

      • Use statistical techniques (e.g., ANOVA, t-tests) to compare task performance and workload distribution across the control and experimental groups.

      • Measure the impact of visibility on social loafing (e.g., fewer tasks completed in low-visibility groups).

   2. Qualitative Analysis:

      • Analyze simulated feedback from agents to confirm the perceived increase in loafing in low-visibility groups.

      • Examine dominant-aggressive behaviors and their influence on team performance.

7. Validate and Refine the Simulation

   1. Compare Results: Compare simulation results with findings from the original study, confirming the correlation between social loafing, task visibility, and workload fairness.

   2. Sensitivity Testing: Vary conditions slightly (e.g., partial task visibility) and run new simulations to see if the results hold consistently.

   3. Generalization: Apply the simulation to different types of projects (other than software engineering) to test whether similar behaviors emerge in different domains.

8. Report Findings

   1. Present a comparison of social loafing indicators under different conditions.

   2. Highlight the importance of transparency and fair workload distribution in mitigating loafing behaviors.

   3. Discuss the implications for educational practices, such as the use of collaboration tools that enhance task visibility.

CONTRIBUTIONS

We believe that our framework contributes significantly to the education discipline in multiple ways, and below are just some of the many:

1. Enhancing Collaborative Learning: By simulating group dynamics and behaviors, we provided a framework for understanding how task visibility, workload fairness, and individual roles affect student collaboration. This can inform educators on structuring more effective group work addressing social loafing and maximizing engagement, not only in software engineering courses, but generally in all courses where students need to take on roles and experience cooperation.

2. Applying Multi-Agent Simulation in Education: Introducing the use of ChatGPT as an agent-based simulation tool highlights an innovative, scalable approach for modeling student behaviors. It can help educators test various scenarios before implementing strategies in real classroom settings.

3. Providing Pedagogical Insights into Group Dynamics: We contribute to educational research by demonstrating how group dynamics affect learning outcomes, providing additional evidence that transparency and fair workload distribution in projects lead to improved participation and learning effectiveness.

4. Bridging Education and AI Research: Our work merges artificial intelligence with education, offering a methodology that brings AI-based simulations to curriculum design, project-based learning, and classroom management. This intersection can lead to new tools and strategies for both teachers and students.

Our approach could also spark future discussions on using AI-driven simulations to model complex educational environments, giving educators more data-driven approaches to designing course activities.

References

1. Aggarwal P & CL O'Brien. 2008. Social loafing on group projects: Structural antecedents and effect on student satisfaction. Journal of Marketing Education 30(3):255-264. DOI: 10.1177/0273475308322283.

2. Albanese R & D Van Fleet. 1985. Rational behavior in groups: The free riding tendency. Academy of Management Review 10(2):244-255, 1985. DOI: 10.2307/257966.

3. Barrick MR & MK Mount. 1991. The Big Five personality dimensions and job performance: A meta-analysis. Personnel Psychology 44(1):1-26. DOI: 10.1111/j.1744-6570.1991.tb00688.x.

4. Bond CF & LJ Titus. 1983. Social facilitation: A meta-analysis of 241 studies. Psychological Bulletin 94(2):265-292. DOI: 10.1037/0033-2909.94.2.265.

5. Bubeck S, V Chandrasekaran, R Eldan, J Gehrke, E Horvitz, E Kamar, P Lee, YT Lee, Y Li, S Lundberg, H Nori, H Palangi, MT Ribeiro & Y Zhang. 2023. Sparks of artificial general intelligence: Early experiments with GPT-4. arXiv [cs.CL]. DOI: 10.48550/arXiv.2303.12712.

6. Caldarini G, S Jaf & K McGarry. 2022. A literature survey of recent advances in Chatbots. Information 13(1):41. DOI: 10.3390/info13010041.

7. Chapman KJ, ML Meuter, D Toy & LK Wright. 2006. Are student groups dysfunctional? Perspectives from both sides of the classroom. Journal of Marketing Education 28(1):38-49. DOI: 10.1177/0273475309335575.

8. Chen W, Y Su, J Zuo, C Yang, C Yuan, C-M Chan, H Yu, Y Lu, Y-H Hung, C Qian, Y Qin, X Cong, R Xie, Z Liu, M Sun & J Zhou. 2023. AgentVerse: Facilitating multi-agent collaboration and exploring emergent behaviors. arXiv [cs.CL]. DOI: 10.48550/arXiv.2308.10848.

9. Chidambaram L & LL Tung. 2005. Is Out of Sight, Out of Mind? An Empirical Study of Social Loafing in Technology-Supported Groups. Information Systems Research 16(2):149-168. DOI: 10.1287/isre.1050.0051.

10. Comer DR. 1995. A model of social loafing in real work groups. Human Relations 48(6):647-667. DOI: 10.1177/001872679504800603.

11. Cook R. 2001. Social psychology in project management. PM Forum: Connecting the World to Project Management (http://wwww.pmforum.org).

12. Deyoung CG, LC Quilty & JB Peterson. 2007. Between facets and domains: 10 aspects of the Big Five. Journal of Personality and Social Psychology 93(5): 880-896. DOI: 10.1037/0022-3514.93.5.880.

13. Gao C, X Lan, Z Lu, J Mao, J Piao, H Wang, D Jin & Y Li. 2023. S3: Social-network simulation system with Large Language Model-empowered agents. arXiv [cs.SI]. DOI: 10.48550/arXiv.2307.14984.

14. Ghaffarzadegan N, A Majumdar, R Williams & N Hosseinichimeh. 2023. Generative agent-based modeling: Unveiling social system dynamics through coupling mechanistic models with generative artificial intelligence. arXiv [cs.]. DOI: 10.48550/arXiv.2309.11456.

15. Ghaffarzadegan N, A Majumdar, R Williams & N Hosseinichimeh. 2024. Generative agent-based modeling: An introduction and tutorial. System Dynamics Review 40(1):e1761. DOI: 10.1002/sdr.1761.

16. Greenberg J. 1990. Organizational justice: Yesterday, today and tomorrow. Journal of Management 16(2):299-342. DOI: 10.1177/014920639001600208.

17. Hogan J & B Holland. 2003. Using theory to evaluate personality and job-performance relations: A socioanalytic perspective. Journal of Applied Psychology 88(1):100-112. DOI: 10.1037/0021-9010.88.1.100.

18. Karau SJ & KD Williams. 1997. The effects of group cohesiveness on social loafing and social compensation. Group Dynamics: Theory, Research, and Practice 1(2):156–168. DOI: 10.1037/1089-2699.1.2.156.

19. Kerr N. 1983. Motivation losses in small groups: A social dilemma analysis. Journal of Personality and Social Psychology 45:819-828. DOI: 10.1037/0022-3514.45.4.819.

20. Kidwell R & N Bennet. 1993. Employee propensity to withhold effort: A conceptual model to intersect three avenues of research. Academy of Management Review 18(3):429-456. DOI: 10.2307/258904.

21. Kravitz D & B Martin. 1986. Ringelmann rediscovered: The original article. Journal of Personality and Social Psychology 50(5):936-941. DOI: 10.1037/0022-3514.50.5.936.

22. Latane B, K Williams & S Harkins. 1979. Many hands make light the work: The causes and consequences of social loafing. Journal of Personality and Social Psychology 37:822-832, 1979. DOI: 10.1037/0022-3514.37.6.822.

23. Liden RC, SJ Wayne, RA Jaworski & N Bennet. 2004. Social loafing: A field investigation. Journal of Management 30(2):285-304. DOI: 10.1016/j.jm.2003.02.002.

24. Mauldin M. 1994. ChatterBots, TinyMuds, and the Turing Test: Entering the Loebner Prize Competition. In Proceedings of the Eleventh National Conference on Artificial Intelligence. American Association of Artificial Intelligence Press: Washington DC, USA, pp16-21. See PDF from Wayback Machine.

25. OpenAI. 2024. ChatGPT (August 19 version) [Large language model]. https://openai.com.

26. Pabico JP, CD De Sagun, JPC Galang & JAC Hermocilla. 2008a. Perceptions of social loafing and its measurements in undergraduate computer programming groups. In Proceedings (CDROM) of the 4th Network of CALABARZON Educational Institutions, Inc. (NOCEI) Research Forum, Batangas State University, Batangas City, 21 November 2008.

27. Pabico JP, JAC Hermocilla, JPC Galang & CD De Sagun. 2008b. Perceived social loafing in undergraduate software engineering teams. Philippine Information Technology Journal 1(2):22-28, October 2008. DOI (Preprint): 10.48550/arXiv.1507.08345.

28. Park JS, JC O'Brien, CJ Cai, MR Morris, P Liang & MS Bernstein. 2023. Generative agents: Interactive simulacra of human behavior. arXiv [cs.HC]. DOI: 10.48550/arXiv.2304.03442.

29. Pfaff E & P Huddleston. 2003. Does it matter if I hate teamwork? What impacts student attitudes towards teamwork. Journal of Marketing Education 25(1):37-45. DOI: 10.1177/0273475302250571.

30. Pressman RS & BR Maxim. 2019. Software Engineering: A Practitioner's Approach, 9th Edition. McGraw Hill Education: New York, NY, pp 704. ISBN: 978-1259872976.

31. Rajaguru R, R Narendran & G Rajesh, G. 2020. Social loafing in group-based learning: Student-created and instructor-created group perspectives. Education + Training  62(4):483-501. DOI: 10.1108/ET-01-2019-0018.

32. Recario RNC, MBB de Robles, KEP Bautista & JP Pabico. 2015. Social Loafing Among Members of Undergraduate Software Engineering Groups: Persistence of Perception Seven Years After. DOI (Preprint): 10.48550/arXiv.1509.05338.

33. Sommerville I. 2015. Software Engineering, 10th Edition. Pearson: London, England, pp 816. ISBN: 978-0133943030.

34. Tang J, H Gao, X Pan, L Wang, H Tan, D Gao, Y Chen, X Chen, Y Lin, Yg Li, B Ding, J Zhou, J Wang & J-R Wen. 2024. GenSim: A general social simulation platform with Large Language Model based agents. arXiv [cs.MA]. DOI: 10.48550/arXiv.2410.04360.

35. Waletzko JM. 2021. Social Loafing and Team-Based Learning: Assessing Whether Team-Based Learning Practices are Related to Reduced Perceived Social Loafing. Master's Thesis, University of Wisconsin-Stout. Retrieved from http://digital.library.wisc.edu/1793/83457. 

36. Wang G, Y Xie, Y Jiang, A Mandlekar, C Xiao, Y Zhu, L Fan & A Anandkumar. 2023. Voyager: An open-ended embodied agent with large language models. arXiv [cs.AI]. DOI: 10.48550/arXiv.2305.16291.

37. Williams K & S Karau. Social loafing and social compensation: The effects of expectations of co-worker performance. Journal of Personality and Social Psychology 61(4):570-581. DOI: 10.1037/0022-3514.61.4.570.

38. Xie C, C Chen, F Jia, Z Ye, K Shu, A Bibi, Z Hu, P Torr, B Ghanem & Guohao Li. 2024. Can Large Language Model agents simulate human trust behaviors? arXiv [cs.AI]. DOI: 10.48550/arXiv.2402.04559.