Research Papers

This chapter introduces a conceptual framework for qualitative risk assessment of AI, particularly in the context of the EU AI Act. The framework addresses the complexities of legal compliance and fundamental rights protection by itegrating definitional balancing and defeasible reasoning. Definitional balancing employs proportionality analysis to resolve conflicts between competing rights, while defeasible reasoning accommodates the dynamic nature of legal decision-making. Our approach stresses the need for an analysis of AI deployment scenarios and for identifying potential legal violations and multi-layered impacts on fundamental rights. On the basis of this analysis, we provide philosophical foundations for a logical account of AI risk analysis. In particular, we consider the basic building blocks for conceptually grasping the interaction between AI deployment scenarios and fundamental rights, incorporating in defeasible reasoning definitional balancing and arguments about the contextual promotion or demotion of rights. This layered approach allows for more operative models of assessment of both high-risk AI systems and General Purpose AI (GPAI) systems, emphasizing the broader applicability of the latter. Future work aims to develop a formal model and effective algorithms to enhance AI risk assessment, bridging theoretical insights with practical applications to support responsible AI governance.

This paper provides an overview and critique of the risk based model of artificial intelligence (AI) governance that has become a popular approach to AI regulation across multiple jurisdictions. The 'AI Policy Landscape in Europe, North America and Australia' section summarises the existing AI policy efforts across these jurisdictions, with a focus of the EU AI Act and the Australian Department of Industry, Science and Regulation's (DISR) safe and responsible AI consultation. The 'Analysis' section of this paper proposes several criticisms of the risk based approach to AI governance, arguing that the construction and calculation of risks that they use reproduces existing inequalities. Drawing on the work of Julia Black, it argues that risk and harm should be distinguished clearly and that the notion of risk is problematic as its inherent normativity reproduces dominant and harmful narratives about whose interests matter, and risk categorizations should be subject to deep scrutiny. This paper concludes with the suggestion that existing risk governance scholarship can provide valuable insights toward the improvement of the risk based AI governance, and that the use of multiple regulatory implements and responsive risk regulation should be considered in the continuing development of the model.

Understanding the Impact of Physicians' Legal Considerations on XAI Systems

Gennie Mansi, Mark Riedl

Physicians are--and feel--ethically, professionally, and legally responsible for patient outcomes, buffering patients from harmful AI determinations from medical AI systems. Many have called for explainable AI (XAI) systems to help physicians incorporate medical AI recommendations into their workflows in a way that reduces the potential of harms to patients. While prior work has demonstrated how physicians' legal concerns impact their medical decision making, little work has explored how XAI systems should be designed in light of these concerns. In this study, we conducted interviews with 10 physicians to understand where and how they anticipate errors that may occur with a medical AI system and how these anticipated errors connect to their legal concerns. In our study, physicians anticipated risks associated with using an AI system for patient care, but voiced unknowns around how their legal risk mitigation strategies may change given a new technical system. Based on these findings, we describe the implications for designing XAI systems that can address physicians' legal concerns. Specifically, we identify the need to provide AI recommendations alongside contextual information that guides their risk mitigation strategies, including how non-legally related aspects of their systems, such as medical documentation and auditing requests, might be incorporated into a legal case.

Pneumonia remains one of the leading causes of death among children worldwide, underscoring a critical need for fast and accurate diagnostic tools. In this paper, we propose an interpretable deep learning model on Residual Networks (ResNets) for automatically diagnosing paediatric pneumonia on chest X-rays. We enhance interpretability through Bayesian Gradient-weighted Class Activation Mapping (BayesGrad-CAM), which quantifies uncertainty in visual explanations, and which offers spatial locations accountable for the decision-making process of the model. Our ResNet-50 model, trained on a large paediatric chest X-rays dataset, achieves high classification accuracy (95.94%), AUC-ROC (98.91%), and Cohen's Kappa (0.913), accompanied by clinically meaningful visual explanations. Our findings demonstrate that high performance and interpretability are not only achievable but critical for clinical AI deployment.

The Case for Contextual Copyleft: Licensing Open Source Training Data and Generative AI

Grant Shanklin, Emmie Hine, Claudio Novelli +2

The proliferation of generative AI systems has created new challenges for the Free and Open Source Software (FOSS) community, particularly regarding how traditional copyleft principles should apply when open source code is used to train AI models. This article introduces the Contextual Copyleft AI (CCAI) license, a novel licensing mechanism that extends copyleft requirements from training data to the resulting generative AI models. The CCAI license offers significant advantages, including enhanced developer control, incentivization of open source AI development, and mitigation of openwashing practices. This is demonstrated through a structured three-part evaluation framework that examines (1) legal feasibility under current copyright law, (2) policy justification comparing traditional software and AI contexts, and (3) synthesis of cross-contextual benefits and risks. However, the increased risk profile of open source AI, particularly the potential for direct misuse, necessitates complementary regulatory approaches to achieve an appropriate risk-benefit balance. The paper concludes that when implemented within a robust regulatory environment focused on responsible AI usage, the CCAI license provides a viable mechanism for preserving and adapting core FOSS principles to the evolving landscape of generative AI development.

Explainable intrusion detection systems (IDS) are now recognized as essential for mission-critical networks, yet most "XAI" pipelines still bolt an approximate explainer onto an opaque classifier, leaving analysts with partial and sometimes misleading insights. The Interpretable Generalization (IG) mechanism, published in IEEE Transactions on Information Forensics and Security, eliminates that bottleneck by learning coherent patterns - feature combinations unique to benign or malicious traffic - and turning them into fully auditable rules. IG already delivers outstanding precision, recall, and AUC on NSL-KDD, UNSW-NB15, and UKM-IDS20, even when trained on only 10% of the data. To raise precision further without sacrificing transparency, we introduce Multi-Granular Discretization (IG-MD), which represents every continuous feature at several Gaussian-based resolutions. On UKM-IDS20, IG-MD lifts precision by greater than or equal to 4 percentage points across all nine train-test splits while preserving recall approximately equal to 1.0, demonstrating that a single interpretation-ready model can scale across domains without bespoke tuning.

The integration of Large Language Models (LLMs) in K--12 education offers both transformative opportunities and emerging risks. This study explores how students may Trojanize prompts to elicit unsafe or unintended outputs from LLMs, bypassing established content moderation systems with safety guardrils. Through a systematic experiment involving simulated K--12 queries and multi-turn dialogues, we expose key vulnerabilities in GPT-3.5 and GPT-4. This paper presents our experimental design, detailed findings, and a prototype tool, TrojanPromptGuard (TPG), to automatically detect and mitigate Trojanized educational prompts. These insights aim to inform both AI safety researchers and educational technologists on the safe deployment of LLMs for educators.

Post-hoc model-agnostic local attribution (LA) methods have been widely adopted to explain opaque AI models by quantifying feature-wise contributions. However, many existing methods rely on heuristic or only partially justified attribution mechanisms, while the quality of attribution itself is often shaped by downstream objectives without universally accepted standards. In this work, we propose Taylor exPansion-Originated aDaptive Attribution (TaylorPODA), a new post-hoc model-agnostic LA method grounded in the Taylor expansion framework. We first introduce a set of postulates, which formalize principled requirements for explicitly and exhaustively attributing Taylor terms to the corresponding features. Based on these postulates, we analyze existing post-hoc model-agnostic LA methods and identify a fundamental tension between principled attribution and adaptation toward user-defined utilities. To address this challenge, TaylorPODA introduces a controllable allocation mechanism for Taylor interaction effects, enabling attribution results to adapt to downstream objectives while preserving the proposed postulates. Furthermore, although developed from a Taylor-expansion perspective, TaylorPODA also admits a Harsanyi-dividend interpretation, allowing the attribution mechanism to extend beyond model differentiability. Theoretical analysis demonstrates that TaylorPODA satisfies all the proposed postulates together with an additional adaptation property. Empirical results across multiple datasets and both differentiable and non-differentiable models further show that TaylorPODA achieves consistently improved alignment with user-defined utilities while maintaining the communicability of the resulting explanations. Overall, this work provides a starting point toward more trustworthy XAI systems for the deployment of increasingly powerful yet opaque task models.

With the intervention of machine vision in our crucial day to day necessities including healthcare and automated power plants, attention has been drawn to the internal mechanisms of convolutional neural networks, and the reason why the network provides specific inferences. This paper proposes a novel post-hoc visual explanation method called MI CAM based on activation mapping. Differing from previous class activation mapping based approaches, MI CAM produces saliency visualizations by weighing each feature map through its mutual information with the input image and the final result is generated by a linear combination of weights and activation maps. It also adheres to producing causal interpretations as validated with the help of counterfactual analysis. We aim to exhibit the visual performance and unbiased justifications for the model inferencing procedure achieved by MI CAM. Our approach works at par with all state-of-the-art methods but particularly outperforms some in terms of qualitative and quantitative measures.

Along with accurate prediction, understanding the contribution of each feature to the making of the prediction, i.e., the importance of the feature, is a desirable and arguably necessary component of a machine learning model. For a complex model such as a random forest, such importances are not innate -- as they are, e.g., with linear regression. Efficient methods have been created to provide such capabilities, with one of the most popular among them being permutation feature importance due to its efficiency, model-agnostic nature, and perceived intuitiveness. However, permutation feature importance has been shown to be misleading in the presence of dependent features as a result of the creation of unrealistic observations when permuting the dependent features. In this work, we develop TRIP (Test for Reliable Interpretation via Permutation), a test requiring minimal assumptions that is able to detect unreliable permutation feature importance scores that are the result of model extrapolation. To build on this, we demonstrate how the test can be complemented in order to allow its use in high dimensional settings. Through testing on simulated data and applications, our results show that the test can be used to reliably detect when permutation feature importance scores are unreliable.

Mass-Scale Analysis of In-the-Wild Conversations Reveals Complexity Bounds on LLM Jailbreaking

Aldan Creo, Raul Castro Fernandez, Manuel Cebrian

As large language models (LLMs) become increasingly deployed, understanding the complexity and evolution of jailbreaking strategies is critical for AI safety. We present a mass-scale empirical analysis of jailbreak complexity across over 2 million real-world conversations from diverse platforms, including dedicated jailbreaking communities and general-purpose chatbots. Using a range of complexity metrics spanning probabilistic measures, lexical diversity, compression ratios, and cognitive load indicators, we find that jailbreak attempts do not exhibit significantly higher complexity than normal conversations. This pattern holds consistently across specialized jailbreaking communities and general user populations, suggesting practical bounds on attack sophistication. Temporal analysis reveals that while user attack toxicity and complexity remains stable over time, assistant response toxicity has decreased, indicating improving safety mechanisms. The absence of power-law scaling in complexity distributions further points to natural limits on jailbreak development. Our findings challenge the prevailing narrative of an escalating arms race between attackers and defenders, instead suggesting that LLM safety evolution is bounded by human ingenuity constraints while defensive measures continue advancing. Our results highlight critical information hazards in academic jailbreak disclosure, as sophisticated attacks exceeding current complexity baselines could disrupt the observed equilibrium and enable widespread harm before defensive adaptation.

SymbolicThought: Integrating Language Models and Symbolic Reasoning for Consistent and Interpretable Human Relationship Understanding

Runcong Zhao, Qinglin Zhu, Hainiu Xu +3

Understanding character relationships is essential for interpreting complex narratives and conducting socially grounded AI research. However, manual annotation is time-consuming and low in coverage, while large language models (LLMs) often produce hallucinated or logically inconsistent outputs. We present SymbolicThought, a human-in-the-loop framework that combines LLM-based extraction with symbolic reasoning. The system constructs editable character relationship graphs, refines them using seven types of logical constraints, and enables real-time validation and conflict resolution through an interactive interface. To support logical supervision and explainable social analysis, we release a dataset of 160 interpersonal relationships with corresponding logical structures. Experiments show that SymbolicThought improves annotation accuracy and consistency while significantly reducing time cost, offering a practical tool for narrative understanding, explainable AI, and LLM evaluation.