Computational modeling
To better understand human behaviors and their neural bases, we use descriptive and normative computational models which aim to capture the mechanisms underlying measured behaviors.
A computational model is a mathematical formulation of a hypothesis regarding a cognitive or neural mechanism.
The lab reference person for computational modeling is Alizée Lopez-Persem.
Observed behaviors can be captured by different types of models and algorithms, that correspond to distinct mechanistic hypotheses. To untangle those different possibilities, we typically conduct model comparisons, mainly using Variational Bayesian Analyses (VBA).
Computational modelling of decision-making processes
Subjective value can be easily measured with rating tasks, choice tasks and effort tasks. We use various utility functions to understand how several dimensions can be combined into a unique subjective value. For instance, when choosing an appartement, we need to consider several dimensions, such as geographical localization or surface area. Individuals weigh those dimensions differently to make their final choices. This variability can be accounted for by utility functions. Example of study using utility functions: Lopez-Persem et al, 2017.
Choices can be described by softmax functions and logistic regressions, they can also be modelled as evidence accumulation processes such as in the Drift Diffusion Model (DDM). We use this kind of models to understand how biases can be expressed in choices, and to understand the relationship between response time and choices. Example of study using both softmax and DDM: Lopez-Persem et al, 2016.
Computational modelling of processes underlying creativity
Creativity relies on idea generation and idea evaluation/selection. To understand how ideas are produced, we combine generation and decision functions into models that capture how individuals search into their semantic memory (via pseudo-random walks) and decide to select one idea or another (via decision-making functions). Example of study modelling creative idea production: Lopez-Persem et al, 2022.
Machine learning
Machine learning is a subfield of artificial intelligence that involves training algorithms to learn patterns and make predictions from data. In other words, it's a way to teach computers to make decisions or predictions without being explicitly programmed to do so.
In cognitive neuroscience, machine learning can be used to analyze large datasets of brain imaging or behavioral data to uncover patterns or relationships that may be difficult or impossible to detect with traditional statistical methods. For example, machine learning algorithms can be used to classify brain activity patterns or predict behavior based on brain activity.
Example of study in the lab using machine learning: Mrah et al, 2022 (Emmanuel Mandonnet).