A Hybrid Approach to Causality Analysis

Abstract

In component-based safety-critical systems, when a system safety property is violated, it is necessary to analyze which components are the cause. Given a system execution trace that exhibits component faults leading to a property violation, our causality analysis formalizes a notion of counterfactual reasoning (\what would the system behavior be if a component had been correct?") and algorithmically derives such alternative system behaviors, without re-executing the system itself. In this paper, we show that we can improve precision of the analysis if 1) we can emulate execution of components instead of relying on their contracts, and 2) take into consideration input/output dependencies between components to avoid blaming components for faults induced by other components. We demonstrate the utility of the extended analysis with a case study for a closed-loop patient-controlled analgesia system. Abstract. In component-based safety-critical systems, when a system safety property is violated, it is necessary to analyze which components are the cause. Given a system execution trace that exhibits component faults leading to a property violation, our causality analysis formalizes a notion of counterfactual reasoning ("what would the system behavior be if a component had been correct?") and algorithmically derives such alternative system behaviors, without re-executing the system itself. In this paper, we show that we can improve precision of the analysis if 1) we can emulate execution of components instead of relying on their contracts, and 2) take into consideration input/output dependencies between components to avoid blaming components for faults induced by other components. We demonstrate the utility of the extended analysis with a case study for a closed-loop patient-controlled analgesia system.

Towards A Logical Account of Epistemic Causality

“…The above notion of (objective) causality [3] has motivation that is similar to [10,30], who also discuss causal analysis relative to traces. However, [10,30] work with less expressive languages, and unlike us they provide a counterfactual definition of causality.…”

“…The above notion of (objective) causality [3] has motivation that is similar to [10,30], who also discuss causal analysis relative to traces. However, [10,30] work with less expressive languages, and unlike us they provide a counterfactual definition of causality. As mentioned earlier and shown in [3], our definition above can correctly compute actual causes even for the more problematic examples with early preemption and overdetermination that create serious difficulties for the structural equations-based approach developed in [25,26,15,14].…”