Mind and Matter Researcher Spotlight: Benjamin Cowley and High Performance Cognition

Understanding skilled performance in humans by modelling learning and decision making

Benjamin Cowley, Associate Professor for AI in Learning and Education at the Faculty of Educational Sciences, leads the HiPerCog group which studies how individuals learn to perform with high skill in demanding tasks and environments. According to Cowley, skill acquisition is not simply conditioning but demands inferential intelligence.  

“Inferential intelligence allows us to grasp meaning beyond the information given, to learn under uncertainty and with sparse data by using abstract isomorphisms such as analogy. Such intelligence is powerful and adaptive in a complex, partially predictable world. It allows us to acquire skills at the limits of our physiological and cognitive capacity,” Cowley explains. 

Cowley’s research often involves skill acquisition in games and simulations. “Because of their structured nature, games are an excellent domain for the study of inference. Games can be formally abstracted to sets of relations in a clear, bounded way, and they are already commonly used in machine learning, which facilitates AI agent modelling of observed play behaviour.”  

Cowley is currently planning a longitudinal study of neurotypical and ADHD adolescents (16-19 years), in which participants will learn to perform a battery of tasks, ranging from simple probabilistic structure-learning tasks to commercial videogames. Lab sessions will measure the neural and physiological concomitants of learning, and artificially-intelligent models will be trained and used to link behaviour and biosignals and capture individual player-behaviour profiles. 

The goal of the study is to elucidate how the interplay of conditioning and inference drives structure learning and consequent skill acquisition. In the long-term, Cowley envisions this work leading to enhanced AI tutoring systems that can model our cognitive capabilities and learning needs.

“Interfacing these systems through cloud-connected devices would further allow them to operate and synchronise across devices for a more complete picture of the learner, while connecting to the vast store of knowledge online.”