Public defence in Interactive Systems, M.Sc.(Tech.) Aini Putkonen

The title of the thesis: Modeling human decision-making in naturalistic settings

Thesis defender: Aini Putkonen
Opponent: Prof. Adam Sanborn, University of Warwick, UK
Custos: Prof. Antti Oulasvirta, Aalto University School of Electrical Engineering

Making decisions mediated by graphical user interfaces (GUIs) is increasingly common. Such decisions range from booking a flight online to deciding which news article to read. These tasks involve first gathering information about the available alternatives on the visual display, then selecting the best option. Despite the ubiquity of choices in such naturalistic settings, much of previous work has focused on understanding decision-making in controlled environments that significantly differ from their everyday counterparts. Also, the process of information gathering in GUIs has been understudied. 

The purpose of this work is to bridge the gap between controlled and naturalistic settings in modeling decision-making. The case study of this thesis is decision-making mediated by GUIs. 

The results of this work are presented in three claims. Firstly, I argue that applying existing models developed in controlled settings to naturalistic ones needs to be undertaken with caution. Through applying two classic models of decision-making in naturalistic game logs, it is shown that data collected under everyday conditions may not fulfill the requirements that these models have to achieve reliable results. Secondly, I claim that understanding information gathering is critical for modeling naturalistic decision-making. Empirical evidence is presented from two eye-tracking studies focusing on visual search and browsing. Finally, I propose that models of naturalistic decision-making should include information gathering. Such a model based on the notion of computational rationality is presented. 

The current work has both theoretical and practical implications. Modeling decision-making is an area of theoretical interest across several fields, including cognitive science, economics, and artificial intelligence. This thesis contributes to efforts in understanding decision-making in more challenging and fuzzy naturalistic settings. Simultaneously, the work has practical implications as better models of decision-making can be used in developing applications such as decision support systems.

Key words: decision-making, graphical user interface, cognitive model, POMDP

Thesis available for public display 7 days prior to the defence at Aaltodoc. 

Contact: www.ainiputkonen.fi