The Confusion Fingerprint Index: Cognitive Error Taxonomy as an Adaptive Learning Signal in learners

This pre-registered randomised pilot trial pursues two interrelated aims: (1) to test whether error-taxonomy-aware adaptive learning produces superior 2-week delayed retention compared to correctness-only adaptive learning and standard classroom instruction; and (2) to introduce and validate the Confusion Fingerprint Index (CFI), a three-component vector capturing each learner's characteristic distribution of cognitive error types (Recall Failure, Partial Knowledge, Confabulation, Interference). Ninety students (Classes 6–8; ages 11– 14) enrolled at three government junior high schools in Kanpur Dehat district, Uttar Pradesh, India, will be randomly assigned — stratified by school and class level - to one of three conditions: (A) error-taxonomy-aware Cognivia adaptive learning with differentiated Hindi remediation (n=30); (B) correctness-only Cognivia adaptive learning (n=30); or (C) standard classroom instruction (n=30). Paper-based NCERT Science assessments at three time points (pre-test, immediate post-test, 2-week delayed retention test) constitute the primary and secondary outcome measures. This is the first pre-registered empirical test of cognitive error taxonomy as a real-time adaptive sequencing signal, and the first preregistered randomised adaptive learning trial conducted with rural Hindi-medium government-school learners in India.