Reasoning about Novel Systems of Rules and Reward | LISH

Speaker:

Date and Time: November 2 | 2pm EST | Hybrid

Location: Megablast Room, SEC 6th Floor, Harvard John A. Paulson School of Engineering and Applied Sciences, 150 Western Ave, Boston, MA 02134.

Questions?

Please contact lish@harvard.edu

Zoom: Meeting ID: 947 0154 1280
Password: 251678

Abstract

Systems of rules and reward inform many aspects of human life, whether our jobs, organizations, markets, digital platforms, sports, even board games. In systems, we not only reason about what next action to take – but whether a system is even worth engaging with at all. Yet, many of the systems people engage with are new in some way, whether from a new rule (e.g., tariff), new action (e.g., new diagnostic test), or new players (e.g., a new employee, a startup entering your market, or an AI agent to contend with) — meaning we have no direct experiencewith them before we need to come to judgements and decisions about whether and how to engage. Lack of direct experience grows all the more thorny when you are not only making judgements about a given system, but creating a new system. How then do people flexibly andefficiently reason about and create such a wide range of novel systems of rules and rewards?We take first steps toward explaining these capacities via a computational cognitive model that we call the “Intuitive Gamer”, which engages fast, flat goal-directed probabilistic simulation when reasoning about games. In a series of large-scale behavioral studies with over 1000 participants and 121 two-player strategic board games (almost all novel to our participants), our model quantitatively captures human judgments and decisions varying the amount and kind of experience people have with a game – from no experience at all (“just thinking”), to a single round of play, to indirect experience watching another person and predicting how they should play – and does so significantly better than alternate models that range in sophistication and compute-intensivity. By analyzing a new dataset of over 450 human-created board games, we then explore whether this model can capture some of the cognitive mechanisms that may underlie how people innovate to make new systems that they and others may find enjoyable to engage with. We close by discussing possible implications for game theory, human-AI interaction, and applied computational cognitive science towards problems in strategy and innovation.

About the speaker

Katie is a Postdoctoral Research Fellow in the Computational Cognitive Science Group at MIT and Visiting Postdoc at the Princeton AI Lab. Her research focuses on how people and AI systems reason about and create/innovate in and around new systems of rules and reward (games, broadly construed) – alone and together. Her research is focused then on “applied” computational cognitive science, characterizing peoples’ intuitive theory of games, and putting that knowledge to work to understand how people reason in real-world games (e.g., in work or entrepreneurship) and to inform the design of more human-compatible AI thought partners. She received her MPhil in Machine Learning and Machine Intelligence and PhD in Engineering from the University of Cambridge on a Marshall Scholarship, and holds a BSc in Brain and Cognitive Sciences from MIT.