Computational models in cognitive science
Pro Research Analysisby 
Searched over 200M research papers
Types of Computational Models in Cognitive Science
Computational models in cognitive science are essential tools for understanding how the mind works. There are several main types of models: descriptive models, which summarize data without strong theoretical assumptions; measurement models, which use structural assumptions to relate experimental variables; explanatory models, which explain both processes and differences between experimental conditions; and cognitive architectures, which aim to explain a wide range of cognitive processes at a higher, often symbolic, level of abstraction. All these models can be linked to neural mechanisms underlying cognition, helping researchers move beyond verbal theories to more precise, testable explanations Lewandowsky2018Farrell2018.
Methodological Advances and Challenges in Computational Modeling
Computational modeling has not always kept pace with advances in technology and human-computer interfaces. Many models are still developed in isolated programming languages, making them hard to access, check, or reuse. There is a call for the modeling community to adopt guidelines that improve accessibility and reproducibility, such as sharing source code and using common standards. Journal editors are seen as key players in enforcing these changes to make models more widely available to researchers .
Parameter inference is another challenge. Traditional fitting methods can be inefficient or misleading, potentially leading to the rejection of valid models or acceptance of flawed ones. Modern methods like Bayesian optimization and approximate Bayesian computation offer more efficient, informative, and reproducible ways to fit models and estimate uncertainty in parameter values .
Mechanistic and Computational Explanations
A strong computational model should bridge multiple mechanistic levels, mapping computational processes onto specific cognitive mechanisms. This approach combines ideas from both computational and mechanistic explanations, aiming to show how different levels of a cognitive system implement computational processes. Such integration is seen as more satisfactory for explaining cognitive phenomena than relying solely on either computational or mechanistic accounts Harbecke2020Brun2025.
Applications: Understanding Cognition and Behavior
Computational models are used to explain a wide range of cognitive functions, including perception, memory, reasoning, decision-making, and communication. For example, neural network models help researchers understand how the structure of the visual system affects perception and learning. Hidden Markov Models and deep neural networks have been used to analyze eye movement strategies, revealing patterns not detectable with traditional methods. These models also help compare human cognition with artificial intelligence, supporting the development of explainable AI systems that can interact more effectively with humans Hsiao2024Farrell2018.
In language and communication, computational models of reference help capture the complexity of how speakers refer to objects, sets, or uncertain situations. These models combine insights from philosophy, psychology, logic, and computer science, and are tested through careful experimentation to improve our understanding of human communication .
Model-Based Inference and Generalization
Computational models allow researchers to generate and test theories that generalize beyond specific experimental data. However, model-based analyses depend on fixed assumptions, which can bias interpretations if not carefully managed. Recent work has focused on developing methods to minimize these risks and ensure that models provide reliable, generalizable insights .
A key insight is that cognitive states are not fixed; behavior results from dynamic trajectories through a "cognitive state space." Understanding the constraints, objectives, and mechanisms that guide these movements is crucial for explaining variation in cognition and behavior across time, environments, and individuals .
Conclusion
Computational models are central to cognitive science, offering rigorous, testable frameworks for understanding the mind and behavior. Advances in methodology, accessibility, and integration with mechanistic explanations are helping to overcome traditional challenges. By combining computational modeling with careful experimentation and modern fitting techniques, researchers can develop more accurate, generalizable, and explanatory theories of cognition and its neural underpinnings Lewandowsky2018Harbecke2020Hsiao2024+3 MORE.
Sources and full results
Most relevant research papers on this topic
Parameter Inference for Computational Cognitive Models with Approximate Bayesian Computation
Approximate Bayesian computation is an efficient, informative, and reproducible method for parameter inference in computational cognitive models, offering a path to reproducible results.
Taming the beast: extracting generalizable knowledge from computational models of cognition.
Computational models of cognition can provide general principles from specific experimental measurements, but can be misleading or biased if assumptions are not met.
Computational Models of Referring: A Study in Cognitive Science
Computational models of reference can enhance our understanding of human communication by capturing the complexity of referring, including reference to sets, approximate descriptions, and descriptions produced under uncertainty.
DOI