LIRa session: Salvador Mascarenhas

Speaker: Salvador Mascarenhas (Institut Jean-Nicod, Department of Cognitive Studies, Ecole Normale Supérieure)

Date and Time: Thursday, February 10th 2022, 16:30-18:00, Amsterdam time.

Venue: online.

Title: Question-answer dynamics and confirmation theory in reasoning with alternatives

Abstract.

Informally, confirmation-theoretic approaches to reasoning propose that humans often do not compute posterior probabilities, as the usual norms for rationality recommend, but instead ask themselves to what extent a piece of information (the evidence) supports a particular conclusion (the hypothesis). This family of theories yields our best accounts of the conjunction fallacy and related reasoning problems. Yet from an empirical standpoint, its full scope is still unknown: just how pervasive is this decision-making strategy in human inference making? Additionally, the computational-level justification for confirmation-theoretic reasoning is still an open question: why would humans deploy a mode of reasoning that so often fails to maximize the probability that the conclusions we draw are true?

This talk tries to contribute to answering these two questions. I show that illusory inferences with alternatives, a broad paradigm of deductive fallacies, require confirmation-theoretic tools and cannot be accounted for within traditional deductive theories or with probabilistic theories that exclusively use posterior probabilities as the standard for rationality. Additionally, I present two experiments on the lawyers-and-engineers paradigm from Kahneman and Tversky (1973), showing that errors in variants of this probabilistic task also require confirmation-theoretic notions. One of the experiments applies a simplified version of Kahneman and Tversky’s original paradigm to carefully normed materials, allowing for more fine-grained theory testing. The other is a visual version of the task, using ad hoc concepts where the underlying probabilistic distributions are completely transparent to participants. I then outline a theory that combines confirmation-theoretic tools with the erotetic theory of reasoning (Koralus and Mascarenhas, 2013, 2018), an account of naive deduction where question-answer dynamics are at the root of failures of reasoning in the cases at hand. I argue that confirmation-theoretic reasoning plausibly arises from the semantics and pragmatics of questions and answers, and I propose that this fact offers the beginning of an explanation for the pervasiveness of confirmation-theoretic reasoning.