Pfahl 140

at Blackwell Inn & Conference Center

Jul 25 @ 09:00 UTC - Jul 25 @ 13:00 UTC

Public session

Bringing Cognitive Modeling Up to Speed with Simulation-Based Inference, A Tutorial Approach Simulation based inference (SBI) is an emerging and transformative paradigm for statistical inference, driving model-based insight in a variety of research domains. Traditional statistics education implicitly relies on likelihood functions as the backbone of parameter inference for computational models. However, deriving analytical likelihoods is often infeasible, even for models that are simple to simulate. SBI circumvents this challenge by enabling parameter inference directly from simulation code, eliminating the need for explicit likelihood derivations. Recent advances in computational statistics, deep learning, and user-friendly software have supercharged SBI. In psychology and cognitive science, these developments have made it possible to infer parameters for a wide range of cognitive models. Many of these models were of long-standing theoretical interest but remained untestable due to the lack of efficient inference methods. As a result, our research community has now gained access to a much larger array of computational models that can be routinely tested against data. This workshop aims to guide participants from principles to practice. It will provide attendees with a comprehensive overview of the landscape of SBI techniques as well as a hands-on path from simulation to parameter estimation, using a large basket of methods and covering multiple practical use cases. Participants will gain a clear understanding of the current ecosystem of tools for SBI, the strengths and limitations of specific SBI techniques for different use cases, and the potential synergies between conceptual approaches and existing software libraries.