Published Research

The methodology behind Meridian Labs is grounded in peer-reviewed cognitive science. These papers describe its design and its behavior across domains.

AIED 2026 Late Breaking Results

Confidence-Calibrated Adaptive Learning: An Integrated Adaptive Engine for Professional Exam Preparation

Describes the 6-algorithm adaptive engine that powers all Meridian Labs apps. Covers confidence calibration, spaced repetition, misconception detection, stress inoculation, Bloom's integration, and multi-factor readiness prediction. Deployed across 28 production applications with 72,000+ items.

EDM 2026 Poster/Demo

Cross-Domain Analysis of a Confidence-Calibrated Adaptive Learning Engine

Analyzes how item bank properties interact with confidence-calibrated adaptation across four certification domains. A two-factor predictor ranks misconception concentration from difficulty spread and base accuracy alone.

International Journal of Human-Computer Interaction

The Safety–Agency Inversion: Longitudinal Multi-Method Evidence from Frontier Voice AI Companions

Empirical study of 68 sessions with 4 frontier AI models examining the relationship between safety alignment and relational agency.

Philosophy & Technology

The Epistemic Harm of AI Sycophancy: When Agreement Undermines Justified Belief

Philosophical analysis arguing AI sycophancy degrades users' capacity for justified belief through non-truth-contingent agreement.