Project Details
Projekt
Reinforcement-learning models of voluntary task choices
Subject Area
General, Cognitive and Mathematical Psychology
Term
since 2026
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 564829065
In this project, we aim to better understand the interplay of decision making and task performance in voluntary task switching. In detail we investigate how decisions impact task performance and how task performance (in terms of expected effort) informs future (task) decision processes based on experimental research and cognitive modeling of voluntary task switching. To assess strategic task choices, we apply the self-organized task switching paradigm (Mittelstädt, et al., 2018) which requires participants to balance their switch costs with experimentally induced waiting times. To model task decisions, we propose a Reinforcement Learning (RL) model as an algorithmic-level account of the processing involved in task-switch decisions and as a measurement model allowing one to assess the contributions of the involved processes in a more valid and process-pure way than indicated by ad-hoc indices (such as, e.g., the switch SOA, switch rates, reaction-time switch costs). These modeling goals require i) demonstrating the model’s capability to describe the data well and to do so better than alternative and simpler models as well as ii) demonstrating the model parameters’ stability and . via selective-influence studies and correlational relationships - their construct validity. In pursuit of these goals, work packages are devoted to the formal development of the model (WP 1), evaluating the test-retest stability of indices of task-switch performance and the model parameters (WP 2), the empirical validation of the model and its parameters by means of selective-influence studies (WP3 - WP5). Finally, we wish to extend the validated model to take errors into account, to describe tasks that differ in difficulty, and, collaborating with other projects in the research unit, to account for data from related paradigms.
DFG Programme
Research Units
