Generative AI is increasingly important in many human activities. In this keynote, I describe a series of experiments in generative AI, using a highly-conversational user interface (UI) to large language models (LLMs). I specify multiple strategies through which the conversational paradigm can be used to achieve ethical outcomes that promote AI humility, AI brevity, human control, and co-creativity among human Generative AI has the potential to support human creativity. Over the past 60 years, scholars in Human Centered AI (HCAI) have proposed diverse models of how systems for humancomputer co-creativity can be designed. In 1961, Rhodes proposed a “4Ps” model in which “A Person engages in a computerized Process to make a Product in an environment (Press)” [1]. In this early view, the computer was primarily a tool. Almost 60 years later, Kantosalo and Takala made the most recent update in their “5Cs” model: “A Collective (a Human and an AI) Collaborate to make a Contribution for a Community in a Context” [2].
https://research.ibm.com/people/michael-muller (M. Muller)
© 2023 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). the conversation developed, and which aspects of the conversation would be preserved in an analogy-based design.
In our fith, unpublished experiment, we moved from specialist methods to the more generallyadopted processes of brainstorming. A human was able to guide the UI+LLM in exercises based on divergent-thinking, convergent-thinking, summarization, and structured organization/reorganization of outcomes. The user structured the activity, and the AI provided content. The user could question, critique, and reject certain content, and the AI could (when requested) provide alternatives to that content. The user could also tell the AI to re-organize its proposed higher-level structuring of content.
While these initial experiments were successful, we were only able to implement a dialog
between one human and one AI. Our next projects will use a specialized environment in
which multiple humans can interact with the UI+LLM configuration, with preservation of each
human’s identity, thus adding aspects of Mutual Theory of Mind [
Throughout this work, we have focused on principles of IBM’s Augmented Human Intelligence, in which AI is used to support and extend the work of humans – not to replace humans [21]. Following a recent debate [22], we label all AI conversational turns with an “AI” or “APP” marker – i.e., we explicitly avoid any so-called Turing test confusions about who or what is speaking or acting. We maintain human control of both process and outcomes.
As we showed in a recent CHIWORK paper [23], these are design choices. It is possible to create interactive AI solutions that channel and control the work of humans [24]. Recent work by many researchers have documented the potential and actual harms of such systems (e.g., [25, 26, 27]). We make a diferent choice: We design for AI applications that support, educate, and enable human abilities and human agency.
Michael Muller works at IBM Research in the role of Senior Research Scientist. His work takes place at the intersection of human-computer interaction, AI, and social justice. Michael is known for early work in participatory design, and for co-proposing and co-leading the CHI conference program subcommittee on Critical and Sustainable Computing and Social Justice. His more recent work has explored possible futures for human interactions with generative AI applications. Michael has led and participated in mentorship programs for students and early-career scholars at multiple ACM SIGCHI conferences, including the CHI Early Career Symposium and the CSCW Student Reviewer Mentoring program. ACM recognizes Michael as a Distinguished Scientist. Michael co-chairs the SIGCHI CARES committee, and is a member of the SIGCHI Research Ethics committee. He holds membership in Fempower.tech and AccessSIGCHI, and has begun a term of service for the US National Academies of Science, Engineering and Medicine on the Board on Human-Systems Integration (BOHSI). //mici.codingconduct.cc/. [19] E. Horvitz, Principles of mixed-initiative user interfaces, in: Proceedings of the SIGCHI conference on Human Factors in Computing Systems, 1999, pp. 159–166. [20] R. Farrell, J. Lenchner, J. Kephart, A. Webb, M. Muller, T. Erickson, D. Melville, R. Bellamy,
D. Gruen, J. Connell, et al., Symbiotic cognitive computing, ai Magazine 37 (2016) 81–93. [21] R. Mariani, F. Rossi, R. Cucchiara, M. Pavone, B. Simkin, A. Koene, J. Papenbrock, Trustworthy ai—part 1, Computer 56 (2023) 14–18. [22] B. Shneiderman, M. Muller, On AI anthropomorphism. Medium 4 April (2023). URL: https://medium.com/human-centered-ai/on-ai-anthropomorphism-abff4cecc5ae. [23] M. Muller, J. Weisz, Frameworks for collaborating humans and ais: Sequence and sociality in organizational applications, in: CHIWORK, 2022. [24] M. Muller, J. He, J. D. Weisz, The trouble with ai-based workflows (2023). URL: https://www.researchgate.net/profile/Michael-Muller-12/publication/368242833_ The_Trouble_with_AI-Based_Workflows/links/63dd67fc62d2a24f92f5a04d/ The-Trouble-with-AI-Based-Workflows.pdf. [25] N. Holten Møller, G. Nef, J. G. Simonsen, J. C. Villumsen, P. Bjørn, Can workplace tracking ever empower? collective sensemaking for the responsible use of sensor data at work, Proceedings of the ACM on human-computer interaction 5 (2021) 1–21. [26] C. Lin, S. Margot Lindtner, Techniques of use: Confronting value systems of productivity, progress, and usefulness in computing and design, in: Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems, 2021, pp. 1–16. [27] R. Shelby, S. Rismani, K. Henne, A. Moon, N. Rostamzadeh, P. Nicholas, Y.-A. N’MAH, J. Gallegos, A. Smart, G. VIRK, Identifying sociotechnical harms of algorithmic systems: Scoping a taxonomy for harm reduction, arXiv preprint arXiv:2210.05791 (2022).