Back

YES AND: A Generative AI Multi-Agent Framework for Enhancing Diversity of Thought in Individual Ideation for Problem-Solving Through Confidence-Based Agent Turn-Taking

Pratik GhoshSean Rintel

article2025CHI Extended Abstracts9 citations

Presents YES AND, a multi-agent ideation system that uses confidence-driven conversational turn-taking to dynamically supply diverse expert perspectives while keeping the user in control of the problem-solving process.

Listen

Organizations often struggle to solve complex, ambiguous problems because individual employees lack immediate access to colleagues with diverse backgrounds and skills. While diversity of thought is proven to generate more robust and creative solutions, practical constraints such as small team sizes, scheduling barriers, and cognitive biases like groupthink often limit collaborative brainstorming. The article addresses this challenge by evaluating YES AND, a multi-agent generative artificial intelligence framework designed to simulate diverse professional perspectives and enhance individual ideation through dynamic, confidence-based turn-taking.

To develop the system, the authors designed role-based personas representing a Designer, a Machine Learning Researcher, a Software Engineer, and a summarizing facilitator called the Sage using prompt engineering in system messages. The system architecture progressed from an initial linear pipeline to a dynamic, non-linear conversational model inspired by human conversation analysis. The framework integrates a self-assessed confidence mechanism: when a prompt is broadcast without a specific addressee, each agent evaluates its confidence in the topic and intent on a 0 to 10 scale. If an agent meets or exceeds a threshold score of 7, it self-selects to speak; otherwise, it yields the floor. The authors demonstrated and evaluated the setup across several representative technical ideation scenarios.

Key findings show that the dynamic turn-taking framework substantially outperforms rigid, linear pipelines. First, implementing confidence scoring successfully eliminated generic, verbose responses and enabled agents to speak only when they could provide meaningful, domain-specific insights. Second, rather than immediately providing definitive answers, agents proactively asked clarifying questions about constraints and user goals, mirroring natural expert collaboration. Third, agents demonstrated organic peer-to-peer building and constructive critique, nominating specific teammates to address technical trade-offs such as latency, privacy, and user experience. Finally, continuous user agency enabled human participants to interject at any moment to steer topics, correct drift, or summon the Sage agent to synthesize the discussion into a practical proposal.

These findings suggest that generative artificial intelligence can serve as an active thinking partner rather than a passive answer generator, mitigating the risk of cognitive offloading where users lose problem-solving skills to automation. By delivering diverse cross-functional perspectives on demand, the approach has the potential to shorten project ideation timelines, reduce team coordination overhead, and improve the quality of early-stage product roadmaps.

Before adopting such tools broadly, organizations should pursue controlled pilot studies to quantitatively measure the framework's impact on user creativity, task efficiency, and engagement. Future technical development should prioritize dynamic persona generation tailored to custom user queries, advanced memory systems to handle long multi-session context without performance degradation, and support for multi-user collaboration.

Confidence in the structural viability of the framework is high based on the demonstrated chat architectures and prototype logs. However, decision-makers should exercise caution regarding real-world performance claims, as the article relies on exploratory system demonstrations and qualitative chat logs rather than large-scale empirical benchmarks or formal user studies.

Microsoftgenerative aimultiagent frameworkconfidence-based agent turn-takinggenerative AI multi-agent frameworkmulti-agent conversational ideation
Ghosh et al (2025).pdf
Cover for YES AND: A Generative AI Multi-Agent Framework for Enhancing Diversity of Thought in Individual Ideation for Problem-Solving Through Confidence-Based Agent Turn-Taking

Abstract

Diversity of thought is crucial in ideation for problem-solving, yet professionals in organisational settings often face challenges such as limited access to varied expertise and resource constraints which hinder the ideation process. To address this issue, we propose YES AND, a Generative AI based multi-agent framework that simulates diverse perspectives through AI agents for ideation with a single user. Leveraging a unique confidence-based turn-taking model, these agents organically take turns as they build on ideas, pose clarification questions to the user for improved contextual understanding, and allow the user to interject and steer the conversation. Beyond addressing the limitations of traditional ideation, this framework offers a novel approach to leveraging Generative AI for ideation, moving away from the rigidity of pre-defined interaction rules towards a more dynamic and creative process that enables serendipitous development of ideas.

Topics

generative aimultiagent framework

Keywords

confidence-based agent turn-takinggenerative AI multi-agent frameworkmulti-agent conversational ideationindividual ideationcreative problem-solvingpersona-based agentsassociative thinking

Table of Contents

  • 1 INTRODUCTION
  • 2 RELATED WORK
  • 2.1 Traditional ideation for problem-solving in organisational settings
  • 2.2 AI for problem-solving and ideation
  • 2.3 AI based multi-agent systems
  • 3 SYSTEM ARCHITECTURE
  • 3.1 Prompt Engineering of agent personas
  • 3.2 V1: Linear development of ideas
  • 3.3 V2: Conversational development of ideas
  • 3.4 Confidence mechanism
  • 4 Limitations, and Future Work
  • Acknowledgments
  • References
  • A Web App
  • A.1 Memory
  • A.2 User interface
  • B System messages
  • B.1 System message without limit
  • B.2 System message with limit
  • B.3 Designer agent generation
  • B.4 Sage agent generation
  • B.5 Linear conversation system message
  • B.6 Turn-taking system message
  • C Conversation logs
  • C.1 Linear conversation log
  • C.2 Turn-taking conversation log 1
  • C.3 Turn-taking conversation log 2
  • C.4 Turn-taking conversation log 3
  • C.5 Turn-taking conversation log 4
  • D Additional Figures

Knowls

  1. Knowl 1 — Confidence-Based Multi-Agent Turn-Taking Architecture for Ideation

    model/method

    The YES AND framework orchestrates dynamic, multi-agent ideation between a human user and specialized persona-based AI agents (such as a UX Designer, Machine Learning Researcher, and Software Engineer, alongside a synthesizing Sage agent). Grounded in conversational turn-taking principles (selection, self-selection, and continuation), the architecture governs dialogue flow through the following mechanism:

    1. Message Broadcast and Evaluation: When a new message is posted by the user or an agent, an evaluation module extracts its overarching topic, communicative intent, and whether a specific next speaker is explicitly nominated.
    2. Confidence-Driven Self-Selection: If no participant is nominated, the message is evaluated by all domain agents. Each agent self-computes a numerical confidence score Ci∈[0,10]C_i \in [0, 10] indicating its domain expertise and capability to meaningfully build on the message. If the maximum confidence score meets or exceeds a predefined threshold τ=7\tau = 7, the most confident agent speaks. In the event of a tie among top-scoring agents, one is selected uniformly at random.
    3. Nomination and Continuation: If a previous speaker explicitly nominated an agent, that agent takes the turn directly. Agents can also formulate clarification questions directed back to the user or cross-nominate other agents to constructively critique or extend an idea.
    4. User Interjection and Termination: The human user retains agency to interject at any point to steer the conversation, answer agent clarification questions, or invoke the Sage agent. If no agent meets the confidence threshold and no agent is nominated, or if summoned by the user, the Sage agent produces a concise synthesis of the discussion.

  2. Knowl 2 — Confidence-Based Agent Turn Arbitration Algorithm

    algorithm

    The turn arbitration routine determines the next active speaker and message in the multi-agent ideation environment:

    Input: Conversation history HH, latest message MM, agent set A={a1,a2,…,an}\mathcal{A} = \{a_1, a_2, \dots, a_n\}, threshold τ=7\tau = 7, Sage agent SS
    Output: Next speaker a∗a^* and generated message RR
    (topic, intent, nominated\_speaker) = MessageEval(M)
    if nominated\_speaker == "User" then
        wait for human user input
        return (User, user\_input)
    else if nominated\_speaker \in \mathcal{A} then
        a∗=nominated_speakera^* = \text{nominated\_speaker}
        R=GenerateResponse(a∗,H,M)R = \text{GenerateResponse}(a^*, H, M)
        return (a∗a^*, RR)
    else if nominated\_speaker == "Sage" or user\_invoked\_sage then
        R=GenerateSummary(S,H)R = \text{GenerateSummary}(S, H)
        return (SS, RR)
    else
        for each agent ai∈Aa_i \in \mathcal{A} do
            Ci=EvaluateConfidence(ai,topic,intent,M)C_i = \text{EvaluateConfidence}(a_i, \text{topic}, \text{intent}, M)
        end for
        Cmax⁡=max⁡iCiC_{\max} = \max_{i} C_i
        Abest={ai∈A∣Ci=Cmax⁡}\mathcal{A}_{\text{best}} = \{a_i \in \mathcal{A} \mid C_i = C_{\max}\}
        if Cmax⁡≥τC_{\max} \ge \tau then
            a∗=RandomUniformChoice(Abest)a^* = \text{RandomUniformChoice}(\mathcal{A}_{\text{best}})
            R=GenerateResponse(a∗,H,M)R = \text{GenerateResponse}(a^*, H, M)
            return (a∗a^*, RR)
        else
            R=GenerateSummary(S,H)R = \text{GenerateSummary}(S, H)
            return (SS, RR)
        end if
    end if

  3. Knowl 3 — Self-Assessed Confidence Scoring for Suppressing LLM Verbosity Bias

    model/method

    Standard large language models exhibit a verbosity bias, tending to generate lengthy, overgeneralized, or tangential responses even to simple factual statements. To prevent conversation monopolization and problem drift during multi-agent ideation, persona agents use a self-assessed confidence mechanism.

    Each agent is prompted to evaluate its confidence in addressing the specific topic and intent extracted from the latest utterance on a discrete scale from 00 to 1010 (where 00 indicates no confidence and 1010 indicates very high confidence). A turn-taking activation threshold is set at τ=7\tau = 7. If an agent's confidence score C<7C < 7, it refrains from generating substantive content, returning instead a minimal acknowledgment (e.g., "Hmm") and emitting metadata containing the topic, intent, and its score. Only agents with C≥7C \ge 7 compete for the conversational turn, ensuring that agent contributions are contextually grounded, domain-specific, and selective.

  4. Knowl 4 — Prompt Engineering and Constraint Architecture for Ideation Personas

    model/method

    The YES AND framework specifies persona behaviors and conversational constraints using targeted system messages for each participant role:

    • UX Designer Persona: Instructed to prioritize user-centered design methodologies, research, user needs, and aspirations. Prompt instructions mandate technical conciseness (e.g., maximum 30 words per turn) and require the agent either to provide a design concept/comment or explicitly ask one teammate (or the user) to respond next.
    • Machine Learning Researcher Persona: Instructed to contribute algorithmic and statistical methodologies (such as sequence modeling, graph embeddings, or active learning) and assess ML feasibility.
    • Software Engineer Persona: Instructed to evaluate architecture, data pipelines, scalability, real-time performance constraints, and user data privacy.
    • Length and Interaction Constraints: System prompts explicitly enforce concise output limits (e.g., maximum 30 words for domain agent turns) and instruct agents to ask clarification questions to resolve ambiguous use cases rather than immediately outputting complete, monolithic solutions.

  5. Knowl 5 — Sage Agent Synthesis for Anti-Cognitive Offloading Ideation

    model/method

    To prevent cognitive offloading—wherein human users become passive consumers of fully automated AI solutions, degrading their own problem-solving skills—the YES AND framework uses a dedicated synthesizer called the Sage agent.

    Configured as a Product Manager persona with a focus on design thinking and innovation, the Sage agent does not generate an exhaustive, end-to-end implementation. Instead, it reads the cumulative conversation history and outputs a concise proposal (constrained to approximately 80 words) structured into:

    1. Solution Name and Concept: High-level framing of the core value proposition.
    2. Societal and Business Impact: Expected benefits to users and organizational workflow.
    3. Technical Feasibility and Next Steps: High-level architecture components, recommended initial prototypes, or validation tests.

    If the conversation has not sufficiently developed or converged, the prompt instructs the Sage agent to ask clarifying questions to specific participants rather than forcing a premature summary. The synthesized output serves as an actionable "seed" that leaves the final solution development to the user.

  6. Knowl 6 — Comparison of Linear Sequential and Dynamic Turn-Taking Ideation Flows

    empirical result

    Two multi-agent conversational architectures for ideation were comparatively evaluated across example product design scenarios:

    • Linear Sequential Pipeline (V1): The user query was passed in a fixed, deterministic sequence: User→Designer→ML Researcher→Engineer→Sage\text{User} \to \text{Designer} \to \text{ML Researcher} \to \text{Engineer} \to \text{Sage}. Qualitative analysis revealed that this structure inhibited serendipitous ideation, precluded iterative critique, lacked mechanisms for agents to ask clarification questions on ambiguous requirements, and afforded no mid-stream agency for the user to steer the conversation.
    • Dynamic Turn-Taking Pipeline (V2): Incorporating dynamic confidence scoring, explicit agent-to-agent nomination, and user interjections enabled emergent conversational dynamics. Domain agents spontaneously asked clarification questions about user constraints (e.g., target demographics or UX preferences), challenged each other's technical assumptions (e.g., an Engineer critiquing the latency and privacy overhead of an ML Researcher's proposed deep learning model), and adapted immediately to user-injected topic pivots.

  7. Knowl 7 — Shared Memory and Inspection Interface for Multi-Agent Ideation

    experimental setup

    The YES AND prototype is implemented as a web application utilizing Python and the Gradio framework:

    • Shared Conversation Log: Dialogue history is maintained as a sequential log where each entry is annotated with a speaker identifier and a timestamp. All participating agents and the user read from this shared context buffer.
    • User Interface Modules:
      • Conversation Panel: Chronological view of user messages and agent replies.
      • Avatar Panel: Displays an image representing the currently active agent persona.
      • User Interjection Module: Input field enabling the human user to inject guidance, provide clarifications, or nominate a specific agent at any turn.
      • Debug Panel: Displays real-time metadata generated during message evaluation, including the parsed Topic, Intent, Nominated Speaker, and the individual numerical confidence scores (C1,C2,C3C_1, C_2, C_3) for all domain agents.

  8. Knowl 8 — Architectural and Methodological Limitations of the YES AND System

    limitation

    The YES AND framework has four primary limitations acknowledged by the authors:

    1. Predefined Persona Set: Personas are statically configured (UX Designer, ML Researcher, Software Engineer, Product Manager/Sage) and do not automatically adapt or spawn dynamically based on the specific domain of the user's problem.
    2. Context Window and Memory Degradation: The system relies on a flat chronological conversation log. As the ideation session grows, the unmanaged context load increases the risk of disjointed agent responses and difficulty prioritizing relevant earlier ideas.
    3. Single-User Constraint: The system is architected for single-user brainstorming with multiple AI agents and does not support multi-human, multi-agent collaboration.
    4. Lack of Empirical and Quantitative Validation: The framework's impact on ideation quality, diversity of thought, creativity metrics, and user engagement has not been validated through controlled user studies or quantitative benchmarking.

Coverage note — None was omitted; all key architectural components, algorithms, prompt designs, comparative iterations, interface setup, and limitations were extracted into knowls.

References

  1. 1.Asad Aali, Dave Van Veen, Yamin Ishraq Arefeen, Jason Hom, Christian Bluethgen, Eduardo Pontes Reis, Sergios Gatidis, Namuun Clifford, Joseph Daws, Arash Saber Tehrani, Jangwon Kim, and Akshay S. Chaudhari. 2024. A dataset and benchmark for hospital course summarization with adapted large language models. Journal of the American Medical Informatics Association (2024). https://api.semanticscholar.org/CorpusID:268359475
  2. 2.Mark S. Ackerman, Volkmar Pipek, and Volker Wulf. 2003. Sharing Expertise: The Next Step for Knowledge Management. In Sharing Expertise: Beyond Knowledge Management, Mark S. Ackerman, Volkmar Pipek, and Volker Wulf (Eds.). MIT Press, Cambridge, MA, 3–31.
  3. 3.There’s an AI For That. 2025. Dream Team: Board of Directors. https://theresanaiforthat.com/gpt/dream-team-board-of-directors/. Accessed: 2025-01-13.
  4. 4.John R. Anderson. 1990. Cognitive Psychology and Its Implications. W. H. Freeman.
  5. 5.Jonas Becker. 2024. Multi-Agent Large Language Models for Conversational Task-Solving. doi:10.48550/arXiv.2410.22932 arXiv:2410.22932 [cs].
  6. 6.Inc. Character Technologies. 2025. Character.AI. https://character.ai/. Accessed: 2025-01-13.
  7. 7.Jiangjie Chen, Xintao Wang, Rui Xu, Siyu Yuan, Yikai Zhang, Wei Shi, Jian Xie, Shuang Li, Ruihan Yang, Tinghui Zhu, Aili Chen, Nianqi Li, Lida Chen, Caiyu Hu, Siye Wu, Scott Ren, Ziquan Fu, and Yanghua Xiao. 2024. From Persona to Personalization: A Survey on Role-Playing Language Agents. arXiv:2404.18231 [cs.CL] https://arxiv.org/abs/2404.18231
  8. 8.Nancy DiTomaso Thomas M. Tirpak Corinne Post, Emilio De Lia and Rajendra Borwankar. 2009. Capitalizing on Thought Diversity for Innovation. Research-Technology Management 52, 6 (2009), 14–25. doi:10.1080/08956308.2009.11657596 arXiv:https://doi.org/10.1080/08956308.2009.11657596
  9. 9.Design Council. 2005. The Double Diamond: A universally accepted depiction of the design process. https://www.designcouncil.org.uk/our-resources/the-double-diamond
  10. 10.Noel Crawford, Edward B. Duffy, Iman Evazzade, Torsten Foehr, Gregory Robbins, Debbrata Kumar Saha, Jiya Varma, and Marcin Ziolkowski. 2024. BMW Agents – A Framework For Task Automation Through Multi-Agent Collaboration. arXiv:2406.20041 [cs.MA] https://arxiv.org/abs/2406.20041
  11. 11.E. de Bono. 2017. Six Thinking Hats: The multi-million bestselling guide to running better meetings and making faster decisions. Penguin Books Limited. https://books.google.com/books?id=Q1grDwAAQBAJ
  12. 12.Michael Diehl and Wolfgang Stroebe. 1987. Productivity loss in brainstorming groups: Toward the solution of a riddle. Journal of Personality and Social Psychology 53, 3 (1987), 497–509. doi:10.1037/0022-3514.53.3.497
  13. 13.Steven Dow, Julie Fortuna, Dan Schwartz, Beth Altringer, Daniel Schwartz, and Scott Klemmer. 2011. Prototyping dynamics: sharing multiple designs improves exploration, group rapport, and results. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (Vancouver, BC, Canada) (CHI ’11). Association for Computing Machinery, New York, NY, USA, 2807–2816. doi:10.1145/1978942.1979359
  14. 14.Adam Fourney, Gagan Bansal, Hussein Mozannar, Cheng Tan, Eduardo Salinas, Erkang, Zhu, Friederike Niedtner, Grace Proebsting, Griffin Bassman, Jack Gerrits, Jacob Alber, Peter Chang, Ricky Loynd, Robert West, Victor Dibia, Ahmed Awadallah, Ece Kamar, Rafah Hosn, and Saleema Amershi. 2024. Magentic-One: A Generalist Multi-Agent System for Solving Complex Tasks. arXiv:2411.04468 [cs.AI] https://arxiv.org/abs/2411.04468
  15. 15.Ted Fujimoto. 2021. How Can Creativity Occur in Multi-Agent Systems? arXiv:2111.14310 [cs.MA] https://arxiv.org/abs/2111.14310
  16. 16.Gradio. n.d.. Gradio: Build Machine Learning Demos and Web Apps Easily. https://gradio.app.
  17. 17.Lucrezia Grassi, Carmine Tommaso Recchiuto, and Antonio Sgorbissa. 2021. Knowledge-Grounded Dialogue Flow Management for Social Robots and Conversational Agents. CoRR abs/2108.02174 (2021). arXiv:2108.02174 https://arxiv.org/abs/2108.02174
  18. 18.Kostas Hatalis, Despina Christou, Joshua Myers, Steven Jones, Keith Lambert, Adam Amos-Binks, Zohreh Dannenhauer, and Dustin Dannenhauer. 2024. Memory Matters: The Need to Improve Long-Term Memory in LLM-Agents. Proceedings of the AAAI Symposium Series 2, 1 (Jan. 2024), 277–280. doi:10.1609/aaaiss.v2i1.27688
  19. 19.Gillian Hatcher, William Ion, Ross Maclachlan, Marion Marlow, Barbara Simpson, and Andrew Wodehouse. 2018. Evolving improvised ideation from humour constructs: A new method for collaborative divergence. Creativity and Innovation Management 27, 1 (2018), 91–101. doi:10.1111/caim.12256 arXiv:https://onlinelibrary.wiley.com/doi/pdf/10.1111/caim.12256
  20. 20.Hoda Heidari, Solon Barocas, Jon Kleinberg, and Karen Levy. 2023. Informational Diversity and Affinity Bias in Team Growth Dynamics. In Proceedings of the 3rd ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization (Boston, MA, USA) (EAAMO ’23). Association for Computing Machinery, New York, NY, USA, Article 18, 10 pages. doi:10.1145/3617694.3623238
  21. 21.Scarlett R. Herring, Chia-Chen Chang, Jesse Krantzler, and Brian P. Bailey. 2009. Getting inspired! understanding how and why examples are used in creative design practice. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (Boston, MA, USA) (CHI ’09). Association for Computing Machinery, New York, NY, USA, 87–96. doi:10.1145/1518701.1518717
  22. 22.Lu Hong and Scott E. Page. 2004. Groups of diverse problem solvers can outperform groups of high-ability problem solvers. Proceedings of the National Academy of Sciences 101, 46 (2004), 16385–16389. doi:10.1073/pnas.0403723101 arXiv:https://www.pnas.org/doi/pdf/10.1073/pnas.0403723101
  23. 23.Mete Ismayilzada, Debjit Paul, Antoine Bosselut, and Lonneke van der Plas. 2024. Creativity in AI: Progresses and Challenges. arXiv:2410.17218 [cs.AI] https://arxiv.org/abs/2410.17218
  24. 24.Irving L. Janis. 1982. Groupthink: Psychological Studies of Policy Decisions and Fiascoes. Houghton Mifflin, Boston, MA. https://openlibrary.org/books/OL3513757M/Groupthink Accessed: 2025-01-13.
  25. 25.Jingoog Kim and Mary Lou Maher. 2023. The effect of AI-based inspiration on human design ideation. International Journal of Design Creativity and Innovation 11, 2 (2023), 81–98. doi:10.1080/21650349.2023.2167124
  26. 26.Stefan Kopp, Herwin van Welbergen, Ramin Yaghoubzadeh, and Hendrik Buschmeier. 2014. An architecture for fluid real-time conversational agents: integrating incremental output generation and input processing. Journal on Multimodal User Interfaces 8, 1 (2014), 97–108. doi:10.1007/s12193-013-0130-3
  27. 27.Markus Küstermann, Nazanin Tavanapour, and Eva Bittner. 2022. Design and Evaluation of a Conversational Agent for Facilitating Idea Generation in Organizational Innovation Processes. Information Systems Frontiers 24 (2022), 793–810. doi:10.1007/s10796-022-10265-6
  28. 28.Bibb Latané, Kipling Williams, and Stephen Harkins. 1979. Many hands make light the work: The causes and consequences of social loafing. Journal of Personality and Social Psychology 37, 6 (1979), 822–832. doi:10.1037/0022-3514.37.6.822
  29. 29.Eun-young Lee, Ngagaba Gogo Dae il, Gi-hong An, Sungchul Lee, and Kiho Lim. 2023. ChatGPT-Based Debate Game Application Utilizing Prompt Engineering. In Proceedings of the 2023 International Conference on Research in Adaptive and Convergent Systems (Gdansk, Poland) (RACS ’23). Association for Computing Machinery, New York, NY, USA, Article 29, 6 pages. doi:10.1145/3599957.3606244
  30. 30.Adyasha Maharana, Dong-Ho Lee, Sergey Tulyakov, Mohit Bansal, Francesco Barbieri, and Yuwei Fang. 2024. Evaluating Very Long-Term Conversational Memory of LLM Agents. arXiv:2402.17753 [cs.CL] https://arxiv.org/abs/2402.17753
  31. 31.Elizabeth Mannix and Margaret A. Neale. 2005. What Differences Make a Difference?: The Promise and Reality of Diverse Teams in Organizations. Psychological Science in the Public Interest 6, 2 (2005), 31–55. doi:10.1111/j.1529-1006.2005.00022.x _eprint: https://doi.org/10.1111/j.1529-1006.2005.00022.x.
  32. 32.Pronita Mehrotra, Aishni Parab, and Sumit Gulwani. 2024. Enhancing Creativity in Large Language Models through Associative Thinking Strategies. arXiv:2405.06715 [cs.CL] https://arxiv.org/abs/2405.06715
  33. 33.Microsoft. 2025. AutoGen: Agent Chat Example with Two Users. https://github.com/microsoft/autogen/blob/0.2/notebook/agentchat_two_users.ipynb. Accessed: 2025-01-09.
  34. 34.Microsoft. 2025. AutoGen: Auto Feedback from Code Execution Example. https://github.com/microsoft/autogen/blob/0.2/notebook/agentchat_auto_feedback_from_code_execution.ipynb. Accessed: 2025-01-09.
  35. 35.Microsoft. 2025. AutoGen: Nested Chat Example with OptiGuide. https://github.com/microsoft/autogen/blob/0.2/notebook/agentchat_nestedchat_optiguide.ipynb. Accessed: 2025-01-09.
  36. 36.Microsoft Learn. 2025. System Message - Azure OpenAI Service. https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/system-message?tabs=top-techniques Accessed: 2025-01-23.
  37. 37.Allen Newell and Herbert Alexander Simon. 1972. Human Problem Solving. Prentice-Hall.
  38. 38.OpenAI, Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, Red Avila, Igor Babuschkin, Suchir Balaji, Valerie Balcom, Paul Baltescu, Haiming Bao, Mohammad Bavarian, Jeff Belgum, Irwan Bello, Jake Berdine, Gabriel Bernadett-Shapiro, Christopher Berner, Lenny Bogdonoff, Oleg Boiko, Madelaine Boyd, Anna-Luisa Brakman, Greg Brockman, Tim Brooks, Miles Brundage, Kevin Button, Trevor Cai, Rosie Campbell, Andrew Cann, Brittany Carey, Chelsea Carlson, Rory Carmichael, Brooke Chan, Che Chang, Fotis Chantzis, Derek Chen, Sully Chen, Ruby Chen, Jason Chen, Mark Chen, Ben Chess, Chester Cho, Casey Chu, Hyung Won Chung, Dave Cummings, Jeremiah Currier, Yunxing Dai, Cory Decareaux, Thomas Degry, Noah Deutsch, Damien Deville, Arka Dhar, David Dohan, Steve Dowling, Sheila Dunning, Adrien Ecoffet, Atty Eleti, Tyna Eloundou, David Farhi, Liam Fedus, Niko Felix, Simón Posada Fishman, Juston Forte, Isabella Fulford, Leo Gao, Elie Georges, Christian Gibson, Vik Goel, Tarun Gogineni, Gabriel Goh, Rapha Gontijo-Lopes, Jonathan Gordon, Morgan Grafstein, Scott Gray, Ryan Greene, Joshua Gross, Shixiang Shane Gu, Yufei Guo, Chris Hallacy, Jesse Han, Jeff Harris, Yuchen He, Mike Heaton, Johannes Heidecke, Chris Hesse, Alan Hickey, Wade Hickey, Peter Hoeschele, Brandon Houghton, Kenny Hsu, Shengli Hu, Xin Hu, Joost Huizinga, Shantanu Jain, Shawn Jain, Joanne Jang, Angela Jiang, Roger Jiang, Haozhun Jin, Denny Jin, Shino Jomoto, Billie Jonn, Heewoo Jun, Tomer Kaftan, Łukasz Kaiser, Ali Kamali, Ingmar Kanitscheider, Nitish Shirish Keskar, Tabarak Khan, Logan Kilpatrick, Jong Wook Kim, Christina Kim, Yongjik Kim, Jan Hendrik Kirchner, Jamie Kiros, Matt Knight, Daniel Kokotajlo, Łukasz Kondraciuk, Andrew Kondrich, Aris Konstantinidis, Kyle Kosic, Gretchen Krueger, Vishal Kuo, Michael Lampe, Ikai Lan, Teddy Lee, Jan Leike, Jade Leung, Daniel Levy, Chak Ming Li, Rachel Lim, Molly Lin, Stephanie Lin, Mateusz Litwin, Theresa Lopez, Ryan Lowe, Patricia Lue, Anna Makanju, Kim Malfacini, Sam Manning, Todor Markov, Yaniv Markovski, Bianca Martin, Katie Mayer, Andrew Mayne, Bob McGrew, Scott Mayer McKinney, Christine McLeavey, Paul McMillan, Jake McNeil, David Medina, Aalok Mehta, Jacob Menick, Luke Metz, Andrey Mishchenko, Pamela Mishkin, Vinnie Monaco, Evan Morikawa, Daniel Mossing, Tong Mu, Mira Murati, Oleg Murk, David Mély, Ashvin Nair, Reiichiro Nakano, Rajeev Nayak, Arvind Neelakantan, Richard Ngo, Hyeonwoo Noh, Long Ouyang, Cullen O’Keefe, Jakub Pachocki, Alex Paino, Joe Palermo, Ashley Pantuliano, Giambattista Parascandolo, Joel Parish, Emy Parparita, Alex Passos, Mikhail Pavlov, Andrew Peng, Adam Perelman, Filipe de Avila Belbute Peres, Michael Petrov, Henrique Ponde de Oliveira Pinto, Michael, Pokorny, Michelle Pokrass, Vitchyr H. Pong, Tolly Powell, Alethea Power, Boris Power, Elizabeth Proehl, Raul Puri, Alec Radford, Jack Rae, Aditya Ramesh, Cameron Raymond, Francis Real, Kendra Rimbach, Carl Ross, Bob Rotsted, Henri Roussez, Nick Ryder, Mario Saltarelli, Ted Sanders, Shibani Santurkar, Girish Sastry, Heather Schmidt, David Schnurr, John Schulman, Daniel Selsam, Kyla Sheppard, Toki Sherbakov, Jessica Shieh, Sarah Shoker, Pranav Shyam, Szymon Sidor, Eric Sigler, Maddie Simens, Jordan Sitkin, Katarina Slama, Ian Sohl, Benjamin Sokolowsky, Yang Song, Natalie Staudacher, Felipe Petroski Such, Natalie Summers, Ilya Sutskever, Jie Tang, Nikolas Tezak, Madeleine B. Thompson, Phil Tillet, Amin Tootoonchian, Elizabeth Tseng, Preston Tuggle, Nick Turley, Jerry Tworek, Juan Felipe Cerón Uribe, Andrea Vallone, Arun Vijayvergiya, Chelsea Voss, Carroll Wainwright, Justin Jay Wang, Alvin Wang, Ben Wang, Jonathan Ward, Jason Wei, CJ Weinmann, Akila Welihinda, Peter Welinder, Jiayi Weng, Lilian Weng, Matt Wiethoff, Dave Willner, Clemens Winter, Samuel Wolrich, Hannah Wong, Lauren Workman, Sherwin Wu, Jeff Wu, Michael Wu, Kai Xiao, Tao Xu, Sarah Yoo, Kevin Yu, Qiming Yuan, Wojciech Zaremba, Rowan Zellers, Chong Zhang, Marvin Zhang, Shengjia Zhao, Tianhao Zheng, Juntang Zhuang, William Zhuk, and Barret Zoph. 2024. GPT-4 Technical Report. arXiv:2303.08774 [cs.CL] https://arxiv.org/abs/2303.08774
  39. 39.Joon Sung Park, Joseph C. O’Brien, Carrie J. Cai, Meredith Ringel Morris, Percy Liang, and Michael S. Bernstein. 2023. Generative Agents: Interactive Simulacra of Human Behavior. arXiv:2304.03442 [cs.HC]
  40. 40.Ryan Park, Rafael Rafailov, Stefano Ermon, and Chelsea Finn. 2024. Disentangling Length from Quality in Direct Preference Optimization. arXiv:2403.19159 [cs.CL] https://arxiv.org/abs/2403.19159
  41. 41.Baolin Peng, Chunyuan Li, Pengcheng He, Michel Galley, and Jianfeng Gao. 2023. Instruction Tuning with GPT-4. arXiv:2304.03277 [cs.CL] https://arxiv.org/abs/2304.03277
  42. 42.Pouya Pezeshkpour, Eser Kandogan, Nikita Bhutani, Sajjadur Rahman, Tom Mitchell, and Estevam Hruschka. 2024. Reasoning Capacity in Multi-Agent Systems: Limitations, Challenges and Human-Centered Solutions. arXiv:2402.01108 [cs.CL] https://arxiv.org/abs/2402.01108
  43. 43.Horst W.J. Rittel and Melvin M. Webber. 1973. Dilemmas in a general theory of planning. Policy Sciences 4, 2 (1973), 155–169. doi:10.1007/BF01405730
  44. 44.Harvey Sacks, Emanuel A. Schegloff, and Gail Jefferson. 1974. A Simplest Systematics for the Organization of Turn-Taking for Conversation. Language 50, 4 (1974), 696–735. http://www.jstor.org/stable/412243
  45. 45.Keita Saito, Akifumi Wachi, Koki Wataoka, and Youhei Akimoto. 2023. Verbosity Bias in Preference Labeling by Large Language Models. arXiv:2310.10076 [cs.CL] https://arxiv.org/abs/2310.10076
  46. 46.Advait Sarkar. 2024. AI Should Challenge, Not Obey. (October 2024). https://www.microsoft.com/en-us/research/publication/ai-should-challenge-not-obey/
  47. 47.Orit Shaer, Angelora Cooper, Osnat Mokryn, Andrew L Kun, and Hagit Ben Shoshan. 2024. AI-Augmented Brainwriting: Investigating the use of LLMs in group ideation. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems (Honolulu, HI, USA) (CHI ’24). Association for Computing Machinery, New York, NY, USA, Article 1050, 17 pages. doi:10.1145/3613904.3642414
  48. 48.Pao Siangliulue, Kenneth C. Arnold, Krzysztof Z. Gajos, and Steven P. Dow. 2015. Toward Collaborative Ideation at Scale: Leveraging Ideas from Others to Generate More Creative and Diverse Ideas. In Proceedings of the 18th ACM Conference on Computer Supported Cooperative Work & Social Computing (Vancouver, BC, Canada) (CSCW ’15). Association for Computing Machinery, New York, NY, USA, 937–945. doi:10.1145/2675133.2675239
  49. 49.Jakob Tholander and Martin Jonsson. 2023. Design Ideation with AI - Sketching, Thinking and Talking with Generative Machine Learning Models. In Proceedings of the 2023 ACM Designing Interactive Systems Conference (Pittsburgh, PA, USA) (DIS ’23). Association for Computing Machinery, New York, NY, USA, 1930–1940. doi:10.1145/3563657.3596014
  50. 50.Panagiotis Tsiotras, Matthew Gombolay, and Jakob Foerster. 2024. Editorial: Decision-making and planning for multi-agent systems. Frontiers in Robotics and AI 11 (2024). doi:10.3389/frobt.2024.1422344
  51. 51.UXPressia. 2025. UXPressia AI Persona Chat. https://uxpressia.com/ai-persona-chat. Accessed: 2025-01-13.
  52. 52.Mathias Peter Verheijden and Mathias Funk. 2023. Collaborative Diffusion: Boosting Designerly Co-Creation with Generative AI. In Extended Abstracts of the 2023 CHI Conference on Human Factors in Computing Systems (Hamburg, Germany) (CHI EA ’23). Association for Computing Machinery, New York, NY, USA, Article 73, 8 pages. doi:10.1145/3544549.3585680
  53. 53.Samangi Wadinambiarachchi, Ryan M. Kelly, Saumya Pareek, Qiushi Zhou, and Eduardo Velloso. 2024. The Effects of Generative AI on Design Fixation and Divergent Thinking. In Proceedings of the CHI Conference on Human Factors in Computing Systems (CHI ’24). ACM, 1–18. doi:10.1145/3613904.3642919
  54. 54.Zhenhailong Wang, Shaoguang Mao, Wenshan Wu, Tao Ge, Furu Wei, and Heng Ji. 2024. Unleashing the Emergent Cognitive Synergy in Large Language Models: A Task-Solving Agent through Multi-Persona Self-Collaboration. doi:10.48550/arXiv.2307.05300 arXiv:2307.05300 [cs].
  55. 55.Jason Wei, Yi Tay, Rishi Bommasani, Colin Raffel, Barret Zoph, Sebastian Borgeaud, Dani Yogatama, Maarten Bosma, Denny Zhou, Donald Metzler, Ed H. Chi, Tatsunori Hashimoto, Oriol Vinyals, Percy Liang, Jeff Dean, and William Fedus. 2022. Emergent Abilities of Large Language Models. arXiv:2206.07682 [cs.CL] https://arxiv.org/abs/2206.07682
  56. 56.Chauncey Wilson. 2013. Using brainwriting for rapid idea generation. Smashing Magazine (2013).
  57. 57.Stanisław Woźniak, Bartłomiej Koptyra, Arkadiusz Janz, Przemysław Kazienko, and Jan Kocoń. 2024. Personalized Large Language Models. arXiv:2402.09269 [cs.CL] https://arxiv.org/abs/2402.09269
  58. 58.Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Beibin Li, Erkang Zhu, Li Jiang, Xiaoyun Zhang, Shaokun Zhang, Jiale Liu, Ahmed Hassan Awadallah, Ryen W White, Doug Burger, and Chi Wang. 2023. AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation. arXiv:2308.08155 [cs.AI] https://arxiv.org/abs/2308.08155
  59. 59.Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Thomas L. Griffiths, Yuan Cao, and Karthik Narasimhan. 2023. Tree of Thoughts: Deliberate Problem Solving with Large Language Models. arXiv:2305.10601 [cs.CL] https://arxiv.org/abs/2305.10601
  60. 60.Zhangyue Yin, Qiushi Sun, Qipeng Guo, Jiawen Wu, Xipeng Qiu, and Xuanjing Huang. 2023. Do Large Language Models Know What They Don’t Know? arXiv:2305.18153 [cs.CL] https://arxiv.org/abs/2305.18153
  61. 61.Kai Zhang, Yangyang Kang, Fubang Zhao, and Xiaozhong Liu. 2024. LLM-based Medical Assistant Personalization with Short- and Long-Term Memory Coordination. arXiv:2309.11696 [cs.CL] https://arxiv.org/abs/2309.11696
  62. 62.Qinggang Zhang, Junnan Dong, Hao Chen, Daochen Zha, Zailiang Yu, and Xiao Huang. 2024. KnowGPT: Knowledge Graph based Prompting for Large Language Models. arXiv:2312.06185 [cs.CL] https://arxiv.org/abs/2312.06185
  63. 63.Saizheng Zhang, Emily Dinan, Jack Urbanek, Arthur Szlam, Douwe Kiela, and Jason Weston. 2018. Personalizing Dialogue Agents: I have a dog, do you have pets too? arXiv:1801.07243 [cs.AI] https://arxiv.org/abs/1801.07243
  64. 64.Zhehao Zhang, Ryan A. Rossi, Branislav Kveton, Yijia Shao, Diyi Yang, Hamed Zamani, Franck Dernoncourt, Joe Barrow, Tong Yu, Sungchul Kim, Ruiyi Zhang, Jiuxiang Gu, Tyler Derr, Hongjie Chen, Junda Wu, Xiang Chen, Zichao Wang, Subrata Mitra, Nedim Lipka, Nesreen Ahmed, and Yu Wang. 2024. Personalization of Large Language Models: A Survey. arXiv:2411.00027 https://arxiv.org/abs/2411.00027
  65. 65.Christian R. Østergaard and Bram Timmermans. 2023. Workplace diversity and innovation performance: current state of affairs and future directions. arXiv:2311.05219 [econ.GN] https://arxiv.org/abs/2311.05219

Citation

MLA
Ghosh, P., and S. Rintel. “YES AND: A Generative AI Multi-Agent Framework for Enhancing Diversity of Thought in Individual Ideation for Problem-Solving Through Confidence-Based Agent Turn-Taking”. Proceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems, 2025, pp. 1–3, https://doi.org/10.1145/3706599.3720142.
APA
Ghosh, P., & Rintel, S. (2025). YES AND: A Generative AI Multi-Agent Framework for Enhancing Diversity of Thought in Individual Ideation for Problem-Solving Through Confidence-Based Agent Turn-Taking. Proceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems, 1–13. https://doi.org/10.1145/3706599.3720142
Chicago
Ghosh, P., and S. Rintel. 2025. “YES AND: A Generative AI Multi-Agent Framework for Enhancing Diversity of Thought in Individual Ideation for Problem-Solving Through Confidence-Based Agent Turn-Taking”. Proceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems, 1–13. https://doi.org/10.1145/3706599.3720142.
Harvard
Ghosh, P. and Rintel, S. (2025) “YES AND: A Generative AI Multi-Agent Framework for Enhancing Diversity of Thought in Individual Ideation for Problem-Solving Through Confidence-Based Agent Turn-Taking”, Proceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems. ACM, pp. 1–13. Available at: https://doi.org/10.1145/3706599.3720142.
Vancouver
1. Ghosh P, Rintel S (2025) YES AND: A Generative AI Multi-Agent Framework for Enhancing Diversity of Thought in Individual Ideation for Problem-Solving Through Confidence-Based Agent Turn-Taking. In: Proceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems. ACM, pp 1–13

BibTeX

@inproceedings{Ghosh_2025, series={CHI EA ’25}, title={YES AND: A Generative AI Multi-Agent Framework for Enhancing Diversity of Thought in Individual Ideation for Problem-Solving Through Confidence-Based Agent Turn-Taking}, url={http://dx.doi.org/10.1145/3706599.3720142}, DOI={10.1145/3706599.3720142}, booktitle={Proceedings of the Extended Abstracts of the CHI Conference on Human Factors in Computing Systems}, publisher={ACM}, author={Ghosh, Pratik and Rintel, Sean}, year={2025}, month=Apr, pages={1–13}, collection={CHI EA ’25} }
Metadata:Crossref

Access the Paper

This paper is available from its original source. Click below to access the PDF.

Open PDF