The Role of Meta-Induction in Human Reasoning

点击翻译按钮获取中文摘要

The background of this project is a dual account of cognition, in which locally adaptive strategies of prediction or inference and general meta-inductive strategies work together. Based on inductive generalizations of observed success rates, meta-inductive strategies attempt to select or to combine locally adapted reasoning methods in ways which perform optimally in a possibly changing environment. Following from insights and questions resulting from work in the 1st phase of this project, the renewal application (years 4-6) focuses on three major objectives. (1) Objective 1: "Meta-inductive prediction strategies between logical generality and local adaptivity." Different meta-inductive prediction strategies shall be investigated in the framework of intermittent prediction games based on online learning, by means of computer simulations with artificial and real-world data, combined with logico-mathematical analysis. We will study the recurrence of local adaptivity at the meta-level and ways to deal with this challenge within our dual account. (2) Objective 2: "Empirical investigation of human learning strategies." We will perform a social learning experiment designed to study the factual role and weight that meta-induction plays in the social learning behavior of individuals.(3) Objective 3: "Cautious and risky reasoning from uncertain conditionals in environments with varying degrees of regularity." Different methods of uncertain reasoning with conditionals will be investigated in environments with different entropy, based on computer simulations of conditional reasoning in artificial and in real-world environments. We intend to develop a system of meta-inductive rules describing the dependence of the performance of different systems of conditional reasoning on the environment's entropy.