Identifying Latent Dimensions of Temporal Attention Across Individuals via Unsupervised Machine Learning

• Grant number: 25/12185-2
• Support Opportunities: Scholarships abroad - Research Internship - Doctorate
• Start date: December 20, 2025
• End date: June 19, 2026
• Field of knowledge: Humanities - Psychology - Cognitive Psychology
• Principal Investigator: André Mascioli Cravo
• Grantee: Gustavo Brito de Azevedo
• Supervisor: Jennifer Coull
• Host Institution: Centro de Matemática, Computação e Cognição (CMCC). Universidade Federal do ABC (UFABC). Santo André , SP, Brazil
• Institution abroad: Aix-Marseille Université (AMU), France
• Associated to the scholarship: 24/05153-4 - The Effect of Dopamine Depletion and Brain Stimulation on Temporal Binding: Possible Dissociations between Event Timing and Interval Duration., BP.DR

Abstract

Temporal perception is a crucial cognitive function underpinning numerous neurocognitive processes, including decision-making, action planning, and multisensory integration. Experimental paradigms such as temporal orienting and foreperiod tasks have been widely employed to examine how individuals adapt to temporal expectations. However, despite their robust within-subject effects, interindividual variability in temporal attention remains understudied, particularly regarding the reliability and generalisability of behavioural measures across populations and experimental contexts. This project aims to evaluate the psychometric reliability and multivariate structure of behavioural indices derived from temporal attention tasks in diverse samples, including healthy individuals, clinical populations (e.g., schizophrenia, ADHD), and participants exposed to pharmacological interventions (e.g., ketamine, clonidine, dopamine depletion). Using large-scale datasets collected under varied experimental manipulations, we will first assess within-task reliability through intraclass correlation coefficients and explore between-task correlations among reaction-time-based metrics. Subsequently, we will apply unsupervised machine learning techniques, including clustering algorithms and multivariate distance-based methods, to identify latent performance profiles and examine their correspondence with experimental conditions. By integrating reliability assessments, psychophysical paradigms, and data-driven modelling, this project seeks to uncover shared and dissociable features of temporal attention across populations. The findings will advance our understanding of the neurocognitive mechanisms underlying temporal perception and support the development of robust behavioural markers for experimental research.