This graduate seminar investigates the theory of social computation underlying the interdisciplinary field of computational social science (CSS), with an emphasis on exploring the intersection of cognitive science, computational sociology, and theoretical economics. It aims to identify not only how recent innovations in computational theory can provide new ways of understanding social systems as computational systems, but also how social systems can provide new ways of understanding the structure and nature of computation itself. It covers foundational discoveries in CSS to date that are grounded in well-defined mathematical and/or simulation-based models, as well as successfully validated with empirical data from experiments and observational studies. Each class is organized to examine a potential feature and/or dynamic of social systems that may (or may not, as we shall discuss) be considered a defining constitutive characteristic of uniquely social computation. The features/dynamics covered in the seminar include social network structure, social contagion and diffusion (i.e., the spread of ideas and behaviors through populations), collective intelligence and the wisdom of crowds, conventions and social norms, categories and analogies, cultural markets and path dependent historical evolution, innovation, ranking systems, and search processes, and the macro dynamics of organizations that coordinate and aggregate collective cognition and behavior. The penultimate class reviews current efforts to use large language models (LLMs) and artificial intelligence more generally to imitate human agents in controlled psychological tasks and simulated social network environments, with the aim of examining (i) whether these artificial agents can serve as viable models for predicting, identifying, and explaining the computational dynamics of social systems, and (ii) whether these models can be improved toward this end using the theory of social computation explored in this seminar.
3 units · GSB Letter Graded
This graduate seminar investigates the theory of social computation underlying the interdisciplinary field of computational social science (CSS), with an emphasis on exploring the intersection of cognitive science, computational sociology, and theoretical economics. It aims to identify not only how recent innovations in computational theory can provide new ways of understanding social systems as computational systems, but also how social systems can provide new ways of understanding the structure and nature of computation itself. It covers foundational discoveries in CSS to date that are grounded in well-defined mathematical and/or simulation-based models, as well as successfully validated with empirical data from experiments and observational studies. Each class is organized to examine a potential feature and/or dynamic of social systems that may (or may not, as we shall discuss) be considered a defining constitutive characteristic of uniquely social computation. The features/dynamics covered in the seminar include social network structure, social contagion and diffusion (i.e., the spread of ideas and behaviors through populations), collective intelligence and the wisdom of crowds, conventions and social norms, categories and analogies, cultural markets and path dependent historical evolution, innovation, ranking systems, and search processes, and the macro dynamics of organizations that coordinate and aggregate collective cognition and behavior. The penultimate class reviews current efforts to use large language models (LLMs) and artificial intelligence more generally to imitate human agents in controlled psychological tasks and simulated social network environments, with the aim of examining (i) whether these artificial agents can serve as viable models for predicting, identifying, and explaining the computational dynamics of social systems, and (ii) whether these models can be improved toward this end using the theory of social computation explored in this seminar.
Offered in Spring 2027 at Stanford University.