Arroyo, López & Torres (2026) — Universidad de Granada. 📄 Full text (arXiv)
Addresses a critical tension in CS education: the widespread adoption of generative AI makes it impossible to distinguish authentic student effort from AI code synthesis by evaluating final submissions alone. The paper presents VISMATIC, a rootless containerized framework that pairs robust environment isolation with explicit user-interaction tracking at the API level.
Key innovation: VISMATIC captures macro-level behavioral metrics (keystroke patterns, command execution timing, copy-paste behavior) that successfully flag statistical anomalies indicative of automated or off-platform workflows — while preserving student anonymity. Unlike JupyterHub, which exposes the host system to arbitrary shell commands, VISMATIC's rootless containers prevent lateral movement and host compromise.
The framework provides a scalable blueprint for preserving pedagogical integrity in the AI era, enabling educators to enforce authentic coding discipline through process-oriented monitoring rather than final-output evaluation. Open source on GitHub for institutional adoption.
Related Pages
- academic-integrity — Academic integrity challenges in AI-era CS education
- formative-assessment — Process-oriented assessment and monitoring approaches
- stem-education — CS and computing education
- edtech-platform — Educational technology platforms and infrastructure
- scaffolding — Scaffolding authentic programming practice
- ai-generated-traces-novice-programmers — Detecting AI-generated vs human code in education