Fitness Assessment of AI-based Systems
conference paper
AI-based systems differ strongly from other systems – and it is yet uncertain how this affects governing them, i.e., to which extent and how known system health management methodologies and processes need to be upgraded and partially reinvented to be fully suitable for AI-based systems. Our analysis indicates that the differences between AI-based cyber-physical systems and traditional systems impact health prognostics and management mainly due to the AI-based systems’ foundation in information flows, their novel system architectures that become necessary to enable system-internal awareness to ensure that decision-making is based on accurate information, and their unique abilities, as systems that learn and thus change their behavior invalidate known health and performance indicators, often in unpredictable ways, and lifecycle management actions like updates, upgrades, or maintenance might no longer fit systems that adapted to their operational context. Based on this understanding, we propose to complement Health and Lifecycle Management methodologies for AI-based systems with a probabilistic model-based analysis of the system’s information flows, giving insight into the expected timeliness and quality of information for various conditions, and thus whetherthe system is fit for its purpose. We see that estimating this fitness-for-purpose shows promise for various system health purposes, especially those ensuring trust in safety-critical AI applications.
Topics
TNO Identifier
1000103
Repository link
Publisher
IEEE
Source title
2024 Prognostics and System Health Management Conference (PHM 2024), Stockholm, Sweden, May 28-31, 2024
Files
To receive the publication files, please send an e-mail request to TNO Repository.