NIBS Round Table: Ambiguity and Learning

IMC Auditorium 0.02 - 14:50-16:30

Information and Decisions: Description and Experience Come Together in Individual and Social Interactions

Cleotilde Gonzalez

Abstract

A theoretical distinction has emerged in the past decade regarding how decisions are made from description (explicit definition of risks, outcomes, probabilities) or experience (implicit collection of past outcomes and probabilities). Explanations of choice under descriptive information often rely on Prospect Theory, while experiential choice has been plagued by highly task-specific models that often predict choice in particular tasks but fail to explain behavior even in closely related tasks. Furthermore, in social interactions the information about others (preferences, beliefs, degree of interdependence) may also influence interactions and choice, but narrow self-interest and complete information is a common assumption in empirical game theory paradigms, limiting our understanding of the types of uncertainty that people face in real-world social interactions. In this talk I will discuss recent research that crosses the borders of traditional descriptive or experiential approaches that attempt to address decision making in situations where many levels of information may be available.

Cleotilde Gonzalez

Professor, Carnegie Mellon University

Incomplete Information Games with Ambiguity Averse Players Part 1

Sujoy Mukerji; Peter Klibanoff

Abstract

We study incomplete information games of perfect recall involving players who perceive ambiguity about the types of others and may be ambiguity averse as modeled through smooth ambiguity preferences (Klibanoff, Marinacci and Mukerji, 2005). Our focus is on equilibrium concepts satisfying sequential optimality – each player’s strategy must be optimal at each stage given the strategies of the other players and the player’s conditional beliefs. We show that for the purpose of identifying strategy profiles that are part of a sequential optimum, it is without loss of generality to restrict attention to beliefs generated using a particular generalization of Bayesian updating. We also propose and analyze strengthenings of sequential optimality. Examples illustrate new strategic behavior that can arise under ambiguity aversion. Our concepts and framework are also suitable for examining the strategic use of ambiguity.

Sujoy Mukerji

Professor, Queen Mary University of London

Incomplete Information Games with Ambiguity Averse Players Part 2

Sujoy Mukerji; Peter Klibanoff

Abstract

We study incomplete information games of perfect recall involving players who perceive ambiguity about the types of others and may be ambiguity averse as modeled through smooth ambiguity preferences (Klibanoff, Marinacci and Mukerji, 2005). Our focus is on equilibrium concepts satisfying sequential optimality – each player’s strategy must be optimal at each stage given the strategies of the other players and the player’s conditional beliefs. We show that for the purpose of identifying strategy profiles that are part of a sequential optimum, it is without loss of generality to restrict attention to beliefs generated using a particular generalization of Bayesian updating. We also propose and analyze strengthenings of sequential optimality. Examples illustrate new strategic behavior that can arise under ambiguity aversion. Our concepts and framework are also suitable for examining the strategic use of ambiguity.

Peter Klibanoff

Professor, Northwestern University

The Effect of Learning on Ambiguity Attitudes

Aurélien Baillon

Abstract

Learning information can affect people’s beliefs but also their attitudes towards ambiguity. We propose a method to separate ambiguity attitudes from subjective probabilities and to decompose ambiguity attitudes into two components. Under models like prospect theory that represent ambiguity through non-additive decision weights these components reflect pessimism and likelihood insensitivity. Under multiple priors models, they reflect ambiguity aversion and perceived ambiguity. We apply our method in an experiment varying the level of information subjects had access to. Ambiguity perception and likelihood insensitivity diminished with more information. Ambiguity aversion and pessimism were largely unaffected by new information. Subjects’ behaviour moved towards expected utility with more information, but substantial deviations remained even in the maximum information condition.

Aurélien Baillon

Professor, Erasmus University Rotterdam