Cognitive Alignment Drives Attention: Modeling and Supporting Socially Shared Regulation in Pair Programming

A new study from researchers Anahita Golrang and Kshitij Sharma investigates how cognitive alignment drives attention in pair programming, using socially shared regulation of learning (SSRL) as a framework. Across three eye-tracking experiments involving 182 dyads performing collaborative debugging tasks, the authors measured joint mental effort (JME) via pupillometry and joint visual attention (JVA) via dual eye-tracking. Study 1 found that high-performing pairs exhibit significantly higher JME and JVA, with a stable causal relationship where JME predicts JVA. Study 2 tested reactive adaptive feedback triggered by real-time deviations in JME, JVA, or both. Combined feedback targeting both dimensions outperformed single-channel feedback, yielding the strongest gains in performance, regulatory coherence, and cognitive-to-attentional causality. Study 3 introduced proactive, forecast-based feedback using machine learning to predict upcoming collaboration states. This anticipatory support further enhanced performance by preventing breakdowns before they manifest.

The findings position AI as an intelligence-augmenting co-regulator that helps learners coordinate effort, attention, and understanding together—rather than as an automated controller. Methodologically, the work integrates dual eye-tracking, pupillometry, episode-based analysis, and causal inference to capture SSRL as a dynamic, emergent process. The causal modeling reveals that cognitive alignment systematically drives attentional coordination in successful collaboration, while mismatches between effort and attention characterize unproductive regulation. This research has direct implications for designing AI-assisted collaborative learning tools in programming education and remote pair programming platforms, where real-time feedback on cognitive and attentional states could improve both learning outcomes and code quality.