Output Type Before Quality: A Standards-Derived XAI Admissibility Rubric for Autonomous-Driving Safety
arXiv:2606.05461v1 Announce Type: new Abstract: Safety standards for ML-based autonomous driving specify the kind of evidence an assurance case must contain (directed cause-and-effect chains, quantified interventional effects, named root-cause variables), yet the XAI literature is organised by output type and technique family (saliency maps, feature attribution, counterfactuals, causal graphs, language traces). SHAP, the most-recommended ADS XAI method, returns a ranked feature list that no implementation effort can convert into a directed chain (Fig.1). We name this mismatch the evidence-type
This paper highlights an accelerating concern regarding the mismatch between AI explainability (XAI) techniques and the stringent evidential requirements for safety-critical applications like autonomous driving, a gap that is becoming more acute as these systems approach real-world deployment.
For a strategic reader, this signals a fundamental challenge in AI assurance, indicating that current XAI methods may be insufficient to meet regulatory and safety standards, which could significantly delay or complicate the broad adoption of autonomous systems.
The focus for XAI development shifts from merely demonstrating interpretability to generating evidence that directly addresses safety assurance requirements (e.g., directed cause-and-effect chains), rather than just output types, forcing a re-evaluation of current research priorities.
- · AI Safety Researchers
- · Developers of new XAI techniques meeting regulatory standards
- · Certification bodies and regulators
- · Companies relying on current XAI tools like SHAP for safety cases
- · Autonomous driving developers with products nearing deployment
- · General-purpose XAI frameworks
The automotive industry faces increased pressure to adopt or develop XAI techniques that produce actionable, standards-compliant evidence for autonomous driving safety.
This could lead to a bifurcation of XAI research, with one track focused on regulatory compliance and another on general interpretability, potentially slowing the integration of advanced AI into other safety-critical domains.
The heightened scrutiny on AI explainability in autonomous driving may set a precedent for other regulated industries, fostering demand for and investment in 'safety-first' XAI solutions across the board.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI