CogMath: Closed-Loop Co-Personalized Math Tutor Integrating Executive Functions

CogMath - Project Overview

CogMath is a closed-loop, cognition-aware math tutoring system for Grades 3–5 that co-personalizes (a) math subskills and (b) executive functions (EF)—working memory, inhibitory control, and cognitive flexibility—in real time during normal learning, without extra hardware.

What makes it different is the closed-loop “co-personalization” cycle: as a learner practices math, the system infers momentary cognitive bottlenecks from sensor-free interaction telemetry (e.g., latency, error patterns, strategy switches) and can insert short, curriculum-aligned EF micro-interventions that are predicted to unblock the next math step.

CogMath is designed as a teacher-coachable system: a teacher dashboard explains why a cognitive support was triggered, so educators can interpret, coach, or override decisions—supporting human-in-the-loop instruction.

To be deployable in EU contexts, CogMath also builds privacy-preserving learning at scale using federated learning + differential privacy, keeping telemetry local while still enabling model improvement across sites.

Approach & Methods (Brief)

• Sensor-free cognitive diagnostics embedded in math tasks

• AI adaptive engine combining rule-based logic and machine-learning models

• Contextual bandits to learn individual causal effects of EF micro-interventions

• Teacher dashboards explaining AI decisions and supporting human oversight

• Federated learning + differential privacy for EU-ready deployment

• Classroom pilots in Finland and Armenia with rigorous evaluation

Work Plan Overview (WPs)

• WP1 – Project Management, Governance, Risk, Coordination

• WP2 – Education & Curriculum Mapping

• WP3 – Cognitive Science Framework, Measurement & Validation

• WP4 – AI/ML Personalisation (interpretable + privacy-preserving deployment)

• WP5 – Platform & Mobile Implementation, Gamification, Integration

• WP6 – HCI/UX, Teacher Dashboard, Explainability UX

• WP7 – Pilots & Evaluation (multi-country plan)

• WP8 – Exploitation, Dissemination, Policy Translation

  Key deliverables include validated cognitive diagnostics, AI personalization engine, secure multi-platform system, pilot evaluation reports, FAIR datasets, policy briefs, and exploitation roadmap.

Partners Sought during this Brokerage Opportunity

We are seeking additional consortium partners or advisers, especially:

• AI ethics & child-safety experts (Responsible AI, fairness, explainability)

• GDPR & data-protection specialists (DPIA, federated analytics, compliance)

• HCI / UX / Serious Games partners (child-centred design, teacher dashboards)

• Policy & education-systems advisers (scaling, curriculum alignment, equity)

Target EU Calls / Funding Context

• EIC Pathfinder (Open / Challenge)

• HORIZON-CL2-2027-01-TRANSFO-05 / -06

• HORIZON-CL2-2026-01-TRANSFO-05