Formal models of bounded rationality for autonomous agents

Sammanfattning: A key challenge when developing intelligent agents is to instill behavior into computing systems that can be considered as intelligent from a social and "common-sense" perspective. Such behavior requires agents to diverge from typical decision-making algorithms that strive to maximize simple, often one-dimensional metrics. A striking parallel to this research problem can be found in the design of formal models of human decision-making in micro-economic theory. Traditionally, mathematical models of human decision-making also reflect the ambition to maximize a utility or preference function, which economists refer to as the rational man paradigm. However, evidence suggest that these models are flawed, not only because human decision-making is subject to systematic fallacies, but also because the models depend on assumptions that do not hold in reality. Consequently, the research domain of formally modeling bounded rationality emerged, which attempts to account for these shortcomings. By drawing from these developments in micro-economic theory, this thesis explores different novel approaches to instill common sense-based, socially intelligent decision-making abilities into autonomous agents. In particular, the works collected in this thesis i) present formal models of boundedly altruistic decision-making and consensus-finding, ii) introduce a library for implementing web-based autonomous agents, iii) explore the effects of explanations on the human intelligibility of machine concessions in economic games, and iv) analyze economic rationality as a non-monotonic reasoning property in the context of abstract argumentation.