Measurement Fidelity in AI-Assisted Experimental Measurement Systems

Abstrak

Artificial intelligence (AI) is increasingly integrated into experimental physics for detector reconstruction, signal interpretation, and automated measurement processing. However, how AI-assisted systems preserve measurement fidelity during automated experimental interpretation remains insufficiently understood. This study systematically reviewed measurement fidelity in AI-assisted physics experimentation using the PRISMA 2020 framework. Literature published between 2020 and 2025 was retrieved from Scopus, Web of Science, and IEEE Xplore, resulting in thirteen studies for qualitative synthesis. The analysis identified that measurement fidelity was primarily evaluated through detector benchmarking, uncertainty calibration, simulation-to-experiment comparison, and validation against independent reference measurements. Recurring challenges included simulation dependence, incomplete physical representation, detector-boundary effects, and uncertainty instability, indicating that computational consistency does not necessarily ensure experimentally reliable interpretation. An integrated conceptual framework is proposed to connect fidelity dimensions, validation approaches, and validation limitations, providing a conceptual basis for more reliable validation strategies in AI-assisted physics experimentation.

Identifier

• OpenAlex: https://openalex.org/W7167054647
• DOI: https://doi.org/10.1051/epjconf/202637702005