Humans in Algorithms, Algorithms in Humans: Understanding Cooperation and Creating Social AI with Causal Generative Models
Cooperation is the hallmark human trait which has allowed us to congregate into the vast, continent-sprawling societies we live in today. Yet, the precise social, environmental, and cognitive mechanisms which enable this cooperation are not fully understood. Toward this, lucrative insights have been borne through the use of formal computational models of socio-cognitive phenomena: In simulating our own cooperative behavior, we can better deduce the exact factors which cause it. The combined knowledge of these factors and ability to computationally simulate them allows us to further two goals: First, it empowers us with the knowledge of how to modify our social systems to better human well-being and promote more sustainable, equitable, and compassionate societies. Second, the computational aspect allows us to more directly create artificial, socially competent companions—whether robotic or entirely digital—to cooperate with us in the real world in achieving the first goal. In this thesis, I contribute to the development of artificial social cognition by examining two case studies of cooperation dilemmas: a game of social team cooperation inference known as stag-hunt, and a stylized cooperative irrigation system. Specifically, I show causal, generative models encoding hypotheses on actual mechanisms in the human mind which are able to outperform the extant state-of-the-art models in both of these cases. In the second case, I show how models like this can be automatically discovered through an algorithm known as evolutionary model discovery, greatly expediting the deduction of new models in similar domains. The results have implications not only for understanding the dynamics of human teaming and irrigation systems (the humans in algorithms), but also broader human socio-cognitive mechanisms contributing to cooperation (the algorithms in humans)—all while simultaneously allowing these mechanisms to be encoded into socially competent AI.
If this is your thesis or dissertation, and want to learn how to access it or for more information about readership statistics, contact us at STARS@ucf.edu
Master of Science (M.S.)
College of Engineering and Computer Science
Industrial Engineering and Management Systems
Length of Campus-only Access
Masters Thesis (Open Access)
Miranda, Lux, "Humans in Algorithms, Algorithms in Humans: Understanding Cooperation and Creating Social AI with Causal Generative Models" (2022). Electronic Theses and Dissertations, 2020-. 1054.