Context Effects

What is “context”, and how can we best test for its influence (or the lack thereof)? A principled theoretical approach requires a unity between the verbal definition, the mathematical structure implied by it, and the empirical data observed in experimental investigations. The diversity in settings covered by the literature on contextual influences, spanning from low-level foraging tasks in amoeboid organisms to real-world market behavior, goes hand-in-hand with an almost equal diversity of testing approaches designed to assess the presence of context effects. The present work critically evaluates existing approaches to testing context effects and identifies a mismatch between verbal definitions and the mathematical structure. We propose a new approach to testing context effects through the lens of random preference models. We show that conclusions drawn from random preference models coincide with existing context-effect definitions in the experimental designs typically used. However, the ability of those designs to detect context effects is very limited. We demonstrate how testability can be established by linking individual choices across different contexts, providing a general-purpose, sensitive, and specific test of context effects. A reanalysis of existing datasets confirms that context effects occur more frequently than traditional approaches suggest, and that there is a large degree of heterogeneity in which context effects occur, often incompatible with commonly used labels such as “attraction effect”. The present work highlights the need for theoretical clarifications related to context effects and establishes a modeling framework that is able to achieve testability for a large class of models.

This is an in-person presentation on July 22, 2024 (11:40 ~ 12:00 CEST).

Additive conjoint measurement (ACM) is a formal theory of measurement that specifies how two variables relate to a third (Debreu, 1960; Luce & Tukey, 1964). ACM plays a foundational role in many utility-based decision theories such as prospect theory (Kahneman & Tversky, 1979). We report the results of a decision making under risk/ambiguity experiment designed to test three axioms necessary for an ACM representation of utility: (1) Monotonicity, (2) transitivity, and (3) double cancellation. Our experiment was designed to induce violations of (1) and (2) via the ambiguity aversion effect. For nearly all participants, we find strong evidence that all three ACM axioms hold, even when participants show ambiguity aversion in their preferences.

This is an in-person presentation on July 22, 2024 (12:00 ~ 12:20 CEST).

The disjunction effect is defined as an empirical violation of the Sure-Thing Principle, which states that if a person is willing to take an action independently of the outcome of some event, then they must be willing to do so even when the outcome of the event is unknown. A large recent literature has claimed that a probabilistic version of this statement follows from the Law of Total Probability and used this to claim that a number of empirical findings are incompatible with classical decision theory. We show here that this probabilistic approach cannot show violation either of the Law of Total Probability or of any classical decision theory prediction. We derive from first principles an alternative probabilistic relation which is both necessary and sufficient for a disjunction effect to be inferable in between-subject experiments, and we show that many past claims of a disjunction effect are unsubstantiated.

This is an in-person presentation on July 22, 2024 (12:20 ~ 12:40 CEST).