Dr Matthew J. Crossley is a Senior Lecturer in the Department of Cognitive Science at Macquarie University. He is a computational cognitive neuroscientist who develops mathematical and neural network models to understand how people learn, adapt, and flexibly change behaviour across contexts. His research integrates theory, behavioural experiments, and neuroimaging to explain how multiple brain systems coordinate learning and action. His work spans basic neuroscience and applied machine learning, including prior industry experience as a machine learning engineer and data scientist at SRI International, where he helped translate neuroscience principles into practical AI methods.

Computational neuroscience of learning and unlearning habits and skills

How do people acquire habits and skills, and how do they selectively modify or discard them when circumstances change? Crossley studies context-dependent learning, focusing on how the brain uses information about the current situation to regulate what is learned and when learning updates occur. A central contribution of this work is a computational account of how local basal ganglia microcircuits support context-sensitive associative learning, extending classical reinforcement learning theories of habitual action selection. This research aims to explain both the formation of automatic behaviour and the mechanisms that enable flexibility when habits become maladaptive.

Computational neuroscience of interacting learning systems

Learning and decision making rely on multiple systems rather than a single mechanism. Crossley’s research examines how declarative, hypothesis-driven learning and procedural, trial-and-error learning interact during categorization and decision making. His work shows that procedural learning can proceed even when behaviour appears to be under explicit declarative control, and that switching between learning systems leaves distinct behavioural and neural signatures. This research develops computational models that link behaviour to underlying mechanisms and generate testable predictions for behavioural and neuroimaging studies.

Neuroinspired AI and machine learning

Modern artificial intelligence systems often struggle with context sensitivity, rules, and rapid behavioural switching. Crossley’s research explores how principles from computational neuroscience can inform machine learning systems that exhibit more flexible and human-like behaviour. Drawing on collaborations established in industry, this work focuses on incorporating context dependence and rule-like control into artificial agents. The broader goal is to develop machine learning methods that are both biologically grounded and effective in real-world applications.

Crossley teaches undergraduate courses in cognitive science and neuroscience, with an emphasis on quantitative reasoning, computational modelling, and the neural mechanisms of learning and action.

He convenes and teaches COGS2020 (Experimental Design and Data Analysis for the Cognitive and Brain Sciences), an undergraduate course focused on statistical reasoning, data analysis, and reproducible research practices. He maintains an open-access online book that supports this course and serves as a general resource for statistics and data science in the cognitive sciences.

He also convenes and teaches COGS3020 (Computational Neuroscience), an undergraduate course introducing mathematical and computational approaches to understanding neural systems. The course emphasizes linking models to behaviour and neural data. An accompanying online book provides core material and worked examples.

In addition, Crossley contributes lectures to undergraduate core courses. In COGS1000, he teaches the biophysical foundations of neural computation, with a focus on electrical signalling in nerve cells. In COGS1005, he delivers a more traditional overview of the basal ganglia, covering core circuitry and function including the direct, indirect, and hyperdirect pathways, and their role in action selection.