Human-Centred Computing Group

Overview

The Human-Centred Computing Group asks a question that machine learning research often defers: what happens when real people encounter these systems? We study usability, comprehensibility, and equity across the full lifecycle of intelligent tools — from initial design through deployment and ongoing governance. Our methods draw equally from human-computer interaction, cognitive science, and machine learning.

A core strand of our work concerns explainability. We develop and evaluate techniques for generating explanations that are not just technically faithful to a model's behavior, but genuinely useful to the specific person receiving them. This requires understanding how prior knowledge, cognitive load, and domain expertise affect explanation comprehension — and designing accordingly. We work with practitioners in medical imaging, financial risk assessment, and public-sector decision systems to test our explanations in realistic deployments.

We are also active in algorithmic fairness research — specifically in understanding how definitions of fairness interact with one another, how fairness criteria degrade under distribution shift, and how affected communities can be meaningfully included in system design. The group collaborates closely with the Probabilistic Inference Lab on calibration questions and with the Computational Neuroscience Group on perception and decision-making in human–AI teams.

Research themes

• Explainability and interpretability of machine learning models
• Human evaluation of AI explanations: methodology and metrics
• Algorithmic fairness and equity in deployed systems
• User interface design for human-AI collaboration
• Participatory and value-sensitive design in AI development
• Human factors in automated decision support
• Transparency, accountability, and audit frameworks for ML systems