Titolo della tesi: Bounded rational decision making in human spatial navigation: a computational perspective
The purpose of this dissertation is to investigate the significance and effects of bounded rational decision making in human spatial navigation, using a combination of experimental and computational approaches.
The thesis is divided into two main parts. The first part serves as an introduction to bounded rationality and spatial navigation, followed by a description of the formal methods used in the second part of the thesis; namely, the mathematical framework of Markov Decision Processes (MDPs) and the Info(rmation) Reinforcement Learning (InfoRL) algorithm.
The second part presents novel results from three experiments on navigation in virtual environments and uses computational models of bounded rational decision making to interpret these results. The first experiment investigates the limits imposed by cognitive constraints on human spatial navigation, describing this phenomenon as a trade-off between the accuracy of navigation plans and the cognitive effort required to generate them - for example, the cognitive cost to deviate from a well known, habitual route. The second experiment shows that to fully account for human decision making in sequential choice navigation, it is crucial to jointly consider the maximization of both the utility and the availability of future choices. The third experiment compares the navigation strategies of participants in the Sea Hero Quest (SHQ) virtual navigation task with two classes of computational models, which implement goal-directed and affordance-based navigation strategies, respectively. Furthermore, the experiment addresses the effect of age on navigation strategies and performance in the SHQ dataset.
The thesis concludes with a discussion of the implications of the results for our understanding of human spatial navigation and for the future development of computational models of human decision making.
Taken together, these results show how people balance costs and benefits when making decisions in spatial navigation, and how these results can be interpreted in terms of bounded rationality.