[2506.21744] Federated Item Response Models: A Gradient-driven Privacy-preserving Framework for Distributed Psychometric Estimation

Computer Science > Machine Learning arXiv:2506.21744 (cs) [Submitted on 26 Jun 2025 (v1), last revised 3 Apr 2026 (this version, v2)] Title:Federated Item Response Models: A Gradient-driven Privacy-preserving Framework for Distributed Psychometric Estimation Authors:Biying Zhou, Nanyu Luo, Feng Ji View a PDF of the paper titled Federated Item Response Models: A Gradient-driven Privacy-preserving Framework for Distributed Psychometric Estimation, by Biying Zhou and 2 other authors View PDF HTML (experimental) Abstract:Item Response Theory (IRT) models are widely used to estimate respondents' latent abilities and calibrate item difficulty. Traditional IRT estimation typically requires centralizing all raw responses, raising privacy and governance concerns. We introduce Federated Item Response Theory (FedIRT), a framework that enables distributed calibration of standard IRT models without transferring individual-level data, thereby preserving confidentiality while retaining statistical efficiency. To provide formal protection, we further develop FedIRT-DP, a user-level differentially private extension. Each site computes per-student gradients, clips them to a fixed norm, and shares only masked sums; the server adds calibrated Gaussian noise and performs MAP updates. This yields an auditable $(\varepsilon,\delta)$ guarantee at the student level and a single, tunable privacy-utility trade-off via the clipping bound and noise scale. The same mechanism improves robustness to extr...