CUBICS: Situation-aware performance estimation for safety-relevant ML components

2026-08-17

AI Summary

This study addresses the challenge that safety-critical ML component performance is highly scenario-dependent yet lacks modular field assessment capabilities. We propose CUBICS, a framework that partitions the operational domain into specific contexts and employs subjective logic with Bayesian inference to dynamically update component-level probabilistic assurances, deriving overall risk through context frequency analysis. The core contribution lies in enabling context-aware modular evaluation without requiring a monolithic system model, thereby overcoming the limitations of traditional global failure probabilities. Consequently, this work provides effective building blocks for safety assurance of safety-critical ML systems based on field data, facilitating more granular and adaptive risk quantification in dynamic environments.

Abstract

Machine learning (ML) is a key technology driving innovation today, but ensuring ML safety remains a major challenge for safety-related applications. A promising idea is to build proven-in-use arguments from field data, e.g. by running ML components (MLCs) in shadow mode or within safety envelopes so that their outputs can be monitored as 'safe probes' without affecting safety. These probes can then be used to build a statistical argument about field performance in a Bayesian way. However, many Bayesian field-data approaches in safety engineering model failures as a simple Bernoulli (or binomial) process with a single global failure probability and i.i.d. trials, which is rarely adequate for MLCs whose performance depends strongly on context. Statistical evidence is also about coverage of relevant situations, including edge cases, and building a single integrated statistical model for the entire system is usually not feasible. To address these challenges, this paper introduces CUBICS, a context-modular framework for per-component, situation-aware performance estimation of safety-relevant ML components. CUBICS partitions the operational design domain into situations and, for each safety-relevant component, defines a set of situation-specific assumptions and probabilistic guarantees that are represented and updated in a Bayesian manner using Subjective Logic (SL). By combining these guarantees with beliefs about how often each situation occurs, CUBICS derives an overall risk estimate for each component without requiring a monolithic system-level statistical model, and thus provides a building block for modular, field-data based safety assurance.

Problem

• Safety-relevant ML
• Situation-aware performance estimation
• Bayesian field-data analysis
• Operational Design Domain
• Modular safety assurance

Innovation

• Situation-aware performance estimation
• Context-modular framework
• Subjective Logic
• Bayesian safety assurance
• Operational Design Domain partitioning