John Molloy
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
John Molloy has a total of 37 co-authors. Here are the top co-authors by number of shared publications:
- Jie Zou
Department of Computer Science, University of York, York, U.K.
4 shared publications - John McDermid
University of York
4 shared publications - Philippa Ryan
University of York
4 shared publications - Sepeedeh Shahbeigi
University of York, York, United Kingdom
4 shared publications - Colin Paterson
University of York
2 shared publications - Ibrahim Habli
University of York, York, UK
2 shared publications
Latest Publications
- Open Access
- 28 July 2026
- Published by Elsevier in SSRN Electronic Journal
Safety Assurance of AI-enabled Autonomy Using Structural Causal World Models
Safety assurance of autonomous systems requires reasoning about how software components interpret uncertain operating contexts. As autonomy increasingly incorporates machine-learned components, assurance based only on predictive accuracy or task performance offers limited support for traceable safety arguments, runtime monitoring, and post-incident analysis. This paper advances Structural Causal World Models (SCWMs) as assurance-oriented world models that make safety-relevant assumptions, dependencies, uncertainty bounds, and constraints explicit. We argue that internal structural coherence, the requirement that a world model behaves plausibly with respect to known physical and semantic constraints, is a necessary precondition for using world models in safety-critical systems. To make this claim evaluable, we introduce structural-coherence metrics that complement predictive accuracy: interventional directionality checks, SHAP-based structural attribution, and regime-dependent amplification metrics for detecting misaligned explanatory structure near safety-critical boundaries. We instantiate the framework in a longitudinal automatic emergency braking use case with an explicit sensor-degradation model for assurance-relevant perception uncertainty. Results show that the contract-induced stopping-distance model has plausible causal effect directions, assigns influence to meaningful variables, amplifies relevant mechanisms in critical regimes, and exposes deliberately injected structural faults. CARLA consistency checks indicate that these qualitative structural patterns persist within a realistic but constrained operating domain. Closed-loop Monte Carlo and CARLA experiments provide supporting operational evidence, under matched sampled conditions; SCWM-derived guardrails maintain or improve observed safety margins without observed regressions. Overall, SCWMs provide a causal and semantic foundation for integrating modelling, specification, operation, and assurance, and explicit criteria for judging whether a world model is structurally suitable for safety assurance.
- 28 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessSafety Assurance of AI-enabled Autonomy Using Structural Causal World Models
Safety assurance of autonomous systems requires reasoning about how software components interpret uncertain operating contexts. As autonomy increasingly incorporates machine-learned components, assurance based only on predictive accuracy or task performance offers limited support for traceable safety arguments, runtime monitoring, and post-incident analysis. This paper advances Structural Causal World Models (SCWMs) as assurance-oriented world models that make safety-relevant assumptions, dependencies, uncertainty bounds, and constraints explicit. We argue that internal structural coherence, the requirement that a world model behaves plausibly with respect to known physical and semantic constraints, is a necessary precondition for using world models in safety-critical systems. To make this claim evaluable, we introduce structural-coherence metrics that complement predictive accuracy: interventional directionality checks, SHAP-based structural attribution, and regime-dependent amplification metrics for detecting misaligned explanatory structure near safety-critical boundaries. We instantiate the framework in a longitudinal automatic emergency braking use case with an explicit sensor-degradation model for assurance-relevant perception uncertainty. Results show that the contract-induced stopping-distance model has plausible causal effect directions, assigns influence to meaningful variables, amplifies relevant mechanisms in critical regimes, and exposes deliberately injected structural faults. CARLA consistency checks indicate that these qualitative structural patterns persist within a realistic but constrained operating domain. Closed-loop Monte Carlo and CARLA experiments provide supporting operational evidence, under matched sampled conditions; SCWM-derived guardrails maintain or improve observed safety margins without observed regressions. Overall, SCWMs provide a causal and semantic foundation for integrating modelling, specification, operation, and assurance, and explicit criteria for judging whether a world model is structurally suitable for safety assurance.
- 11 June 2026
- Published by Springer Nature
- p. 80-95
Quantitative Verification of In-flight UAV Reconfiguration Under Uncertainty
- 11 June 2026
- Published by Springer Nature
- p. 80-95
Quantitative Verification of In-flight UAV Reconfiguration Under Uncertainty
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 68-73
Formally Guaranteed Control Adaptation for ODD-Resilient Autonomous Systems
Ensuring reliable performance in situations outside the Operational Design Domain (ODD) remains a primary challenge in devising resilient autonomous systems. We explore this challenge by introducing an approach for adapting probabilistic system models to handle out-of-ODD scenarios while, in parallel, providing quantitative guarantees. Our approach dynamically extends the coverage of existing system situation capabilities, supporting the verification and adaptation of the system’s behaviour under unanticipated situations. Preliminary results demonstrate that our approach effectively increases system reliability by adapting its behaviour and providing formal guarantees even under unforeseen out-of-ODD situations.
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 68-73
Open AccessFormally Guaranteed Control Adaptation for ODD-Resilient Autonomous Systems
Ensuring reliable performance in situations outside the Operational Design Domain (ODD) remains a primary challenge in devising resilient autonomous systems. We explore this challenge by introducing an approach for adapting probabilistic system models to handle out-of-ODD scenarios while, in parallel, providing quantitative guarantees. Our approach dynamically extends the coverage of existing system situation capabilities, supporting the verification and adaptation of the system’s behaviour under unanticipated situations. Preliminary results demonstrate that our approach effectively increases system reliability by adapting its behaviour and providing formal guarantees even under unforeseen out-of-ODD situations.
- Open Access
- 20 August 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Autonomous and Adaptive Systems
INSYTE: A Classification Framework for Traditional to Agentic AI Systems
Existing classification frameworks for artificial intelligence (AI) and autonomous systems are being outpaced by recent advancements in AI technologies. This limits their applicability to modern intelligent systems, particularly agentic AI systems (autonomous systems that leverage foundation models to achieve wide-ranging, multi-layered goals). To address this deficiency, we introduce INSYTE, a multi-faceted framework that supports the classification of AI systems ranging from traditional rule-based systems to cutting-edge embodied AI and agentic systems. To that end, INSYTE considers the essential characteristics of an AI system across eight key dimensions grouped into four categories: system design ( underspecification and adaptiveness ); functionality ( breadth and depth ); operating environment ( diversity and dynamism ); and independence from human operational control ( intervention and oversight ). Different AI systems (or versions of systems) yield different “patterns” on an eight-axis radar chart that INSYTE uses to provide an immediate visual summary of an AI system's overall capability, and a detailed representation of its individual characteristics. The INSYTE framework aligns with OECD's definition of deployed AI systems, which is becoming the standard definition used by legislators and developers worldwide.
- 20 August 2025
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Autonomous and Adaptive Systems
Open AccessINSYTE: A Classification Framework for Traditional to Agentic AI Systems
Existing classification frameworks for artificial intelligence (AI) and autonomous systems are being outpaced by recent advancements in AI technologies. This limits their applicability to modern intelligent systems, particularly agentic AI systems (autonomous systems that leverage foundation models to achieve wide-ranging, multi-layered goals). To address this deficiency, we introduce INSYTE, a multi-faceted framework that supports the classification of AI systems ranging from traditional rule-based systems to cutting-edge embodied AI and agentic systems. To that end, INSYTE considers the essential characteristics of an AI system across eight key dimensions grouped into four categories: system design ( underspecification and adaptiveness ); functionality ( breadth and depth ); operating environment ( diversity and dynamism ); and independence from human operational control ( intervention and oversight ). Different AI systems (or versions of systems) yield different “patterns” on an eight-axis radar chart that INSYTE uses to provide an immediate visual summary of an AI system's overall capability, and a detailed representation of its individual characteristics. The INSYTE framework aligns with OECD's definition of deployed AI systems, which is becoming the standard definition used by legislators and developers worldwide.
- 15 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Computer
- Vol. 57 (11) , 60-70
Hazard and Safety Analysis of Machine-Learning-Based Perception Capabilities in Autonomous Vehicles
This article proposes an adapted safety analysis method tailored to autonomous systems perception systems. It rationalizes the selection of guidewords and demonstrates how results inform design and verification through partial analysis of a road vehicle, and discusses the broader role in a comprehensive safety process.
- 15 October 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Computer
- Vol. 57 (11) , 60-70
Hazard and Safety Analysis of Machine-Learning-Based Perception Capabilities in Autonomous Vehicles
This article proposes an adapted safety analysis method tailored to autonomous systems perception systems. It rationalizes the selection of guidewords and demonstrates how results inform design and verification through partial analysis of a road vehicle, and discusses the broader role in a comprehensive safety process.
- 9 September 2024
- Published by Springer Nature
- p. 285-299
A Dynamic Assurance Framework for an Autonomous Survey Drone
- 9 September 2024
- Published by Springer Nature
- p. 285-299
A Dynamic Assurance Framework for an Autonomous Survey Drone
- Open Access
- 3 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Computer
- Vol. 57 (4) , 16-25
The Safety of Autonomy: A Systematic Approach
Autonomy does not subvert existing safety processes, but they must be supplemented with methods that address autonomy’s challenges, especially where perception and decision-making tasks are implemented with machine learning. We present an approach to address the safety of autonomous systems, building on and complementing established safety engineering methods.
- 3 April 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Computer
- Vol. 57 (4) , 16-25
Open AccessThe Safety of Autonomy: A Systematic Approach
Autonomy does not subvert existing safety processes, but they must be supplemented with methods that address autonomy’s challenges, especially where perception and decision-making tasks are implemented with machine learning. We present an approach to address the safety of autonomous systems, building on and complementing established safety engineering methods.