Teorie Preferencí: Metodologie, Modelování a Analýza

This dissertation revisits two core themes in the theory of preferences, each with direct relevance for welfare reasoning and for applied decision work: higher-order preferences and relationships between aspects of actions. The thesis is explicitly interdisciplinary, with a guiding aim of conceptual clarification at the level of foundational principles and definitions. Regarding higher-order preferences, the dissertation clarifies what it means to want something once preferences may vary across time and current selves oppose these variations, and how it relates to welfare claims. Regarding relationships between aspects of actions, it clarifies what it means that aspects of action are related -- in particular, what it means for products to be substitutes or complements -- given that intuitive notions are often preference-based while most operational definitions are market-based, and these need not coincide. In doing so, the thesis connects (i) economics and philosophy, where the central task is to delimit welfare-relevant preference structure under preference variation; (ii) econometrics, where the focus is on observability and identification of such structure; and (iii) policy studies and operations research, where the objective is to make these objects usable in applied design and decision work. The thesis consists of five papers. Two papers build welfare-relevant structure for higher-order preferences and subsequent paternalistic claims: (i) a goal-based framework distinguishing terminal and instrumental goals, where misalignments between goals can generate systematic departures from the agent’s “true” preferences, and where terminal goals provide a disciplined basis for welfare and paternalistic reasoning; and (ii) a model of precommitments that represents an agent as a sequence of selves and clarifies when restricting future choice sets can be welfare-improving under the “As-Judged-By-Themselves” criterion, because it can substitute for self-restrictions that are costly for present and future selves. Two papers operationalize product relationships: (iii) an agent-based framework for simulating behavior conditional on substitutability and complementarity between aspects of choice, using CES as a benchmark and showing that AIDS admits a stable relationship interpretation in CES terms whereas QUIT does not; and (iv) a new empirical measure -- product relation correlation -- based on shared purchasing patterns, shown to align with cross-price elasticity and usable for clustering and shelf organization. The final paper (v) provides first evidence on team composition and performance using professional CS:GO data and serves as an introduction to gaming data as a platform for future behavioral research.