AI-driven adaptive stealth assessment for socially regulated learning in collaborative environments

Abstract Conventional digital assessments often overlook learners’ real-time progress, peer interactions, and cognitive demands. This study addresses these limitations by introducing GenSolve, an adaptive artificial intelligence (AI)-driven stealth assessment framework designed to adjust scenario complexity, deliver personalized feedback, and monitor group regulation using socially regulated learning (SoRL) principles. A six-week quasi-experimental study involving 250 undergraduates compared GenSolve to traditional instruction. Results showed a 17.6% gain in problem-solving accuracy, a 21% increase in group cohesion, and a 12.7% improvement in delayed retention. AI-generated feedback reduced repeated errors by 23.4% and improved self-regulation, while SoRL mechanisms supported a 36% rise in independent conflict resolution. GenSolve contributes a scalable, ethically guided model for real-time collaborative assessment, offering practical advances in adaptive evaluation for digital learning environments.