Session | ||
Cognitive modeling
| ||
Presentations | ||
Cognitive-Affective Maps Extended Logic (C.A.M.E.L.): A Toolset for the Visualization and Analysis of Cognitive-Affective Structures in Psychological Research 1Institute of Psychology, University of Freiburg; 2Cluster of Excellence livMatS @ FIT – Freiburg Center for Interactive Materials and Bioinspired Technologies, University of Freiburg A comprehensive set of tools designed to facilitate the use of Cognitive-Affective Maps (CAMs) in psychological research is presented. CAMs are an innovative, mind-map like approach for visually representing individuals’ belief systems, integrating semantic content and affective evaluations. This method enables the systematic elicitation of cognitive-affective structures, providing insight into how individuals mentally and emotionally organize their understanding of complex issues. As such, CAMs extend traditional methodologies by combining free associative mapping with quantifiable network analysis. The software suite Cognitive-Affective Maps Extended Logic (C.A.M.E.L.) offers an integrated research infrastructure comprising three core components: (a) A web-based administrative panel supports the flexible configuration of CAM studies, allowing researchers to set up studies without programming expertise. (b) The data collection tool enables participants to construct CAMs by freely placing or selecting predefined concepts, assigning affective valence, and establishing conceptual connections. (c) The data analysis tool supports semi-automated preprocessing and evaluation of CAM datasets, offering modules for both qualitative and quantitative analyses. The analytic pipeline ensures transparency through documenting all analyses steps. By lowering technical barriers and ensuring methodological rigor, this open-source toolkit allows researchers to efficiently implement CAMs across diverse study designs – ranging from exploratory phases to motivate later survey studies to longitudinal and intervention-based research. Crucially, by modeling the interaction between cognition and emotion, CAMs open new perspectives for studying mental models, attitude formation, and affect-driven decision-making. The tools are freely available under the MIT license, with a detailed online documentation at https://drawyourminds.de Multinomial Processing Trees with Diffusion Model Kernels for Response Time Integration University of Freiburg, Germany We have enhanced the R package "rtmpt" by introducing a newly developed approach for integrating response times within multinomial processing tree (MPT) models. Similar to the method in the previous version, this new approach allows for the estimation of both process-completion times and process-outcome probabilities. However, unlike the previous method, which assumed that each process-completion time followed an exponential distribution, the new approach models these times based on the outcome of a Wiener diffusion process. As a result, process-completion times no longer exhibit the questionable memoryless property. Additionally, the new method can accommodate non-monotonic hazard rates for a single processing branch. These improvements make the new method more representative of actual response time dynamics. Moreover, comparing both approaches provides a valuable way to conduct robustness checks concerning the auxiliary assumptions about process kernels. We demonstrate how to use the new method and provide results of our validation study of the hierarchical Bayesian MCMC algorithm it relies on. Extreme-Value Signal Detection Theory for Recognition Memory: The Parametric Road Not Taken 1University of Freiburg, Germany; 2Syracuse University, NY, USA; 3University College London, UK; 4University of South Carolin, SC, USA Signal Detection Theory (SDT) has been a cornerstone of psychological research, particularly in recognition memory. However, its conventional application relies predominantly on the Gaussian assumption—a reliance driven more by historical precedent than theoretical necessity, with notable drawbacks. This talk critically examines these limitations and introduces an alternative based on extreme-value distributions, specifically event minima—the Gumbel_min SDT model. A key feature distinguishing this model from the Gaussian default is its foundation in a behavioral principle: invariance under choice-set expansions, akin to Yellott's (1977) seminal work on Luce’s Choice Axiom. We present a novel recognition-memory experiment that directly supports this principle and, by extension, the Gumbel_min model. Furthermore, we benchmark its performance against traditional Gaussian SDT across various recognition-memory tasks, including ranking, forced-choice, and simultaneous detection-identification paradigms. Our findings underscore the advantages of Gumbel_min-based modeling, particularly its robust sensitivity index, g′. The Statistical Costs of Correcting Empty Cells in Two-Step Signal Detection Analyses: A Case for a Mixed-Effects Approach University of Freiburg, Germany How people recognize familiar stimuli, decide whether a faint visual cue is present or absent, or judge information as real or fake news and many other psychological research topics revolve around decisions under uncertainty. When investigating such topics, researchers often rely on signal detection theory (SDT), an influential framework that allows one to disentangle two central components within these judgments: Sensitivity, the ability to accurately distinguish signal from noise, and response bias, the tendency to make signal over noise judgments. A common procedure for SDT analyses involves two steps: first estimating individual response bias and sensitivity parameters for each participant and condition, and then subjecting the resulting estimates to inferential statistical analyses. This procedure often requires the use of correction methods for empty cells. Here, we show in three simulation studies that this approach consistently underestimates effects on sensitivity and response bias, and can lead to inflated Type I error rates that depend in complex ways on the true population sensitivity and response bias, as well as the magnitude of correlations between individual parameters. As an alternative, we recommend the use of mixed-effects SDT, which imposes a population distribution on the individual parameters, thus avoiding the need to correct for empty cells. Our simulations show that mixed-effects SDT does not suffer from similar estimation biases and keeps the Type I error rate close to the nominal level. |