Research Papers

PluriHarms: Benchmarking the Full Spectrum of Human Judgments on AI Harm

Jing-Jing Li, Joel Mire, Eve Fleisig +4

Current AI safety frameworks, which often treat harmfulness as binary, lack the flexibility to handle borderline cases where humans meaningfully disagree. To build more pluralistic systems, it is essential to move beyond consensus and instead understand where and why disagreements arise. We introduce PluriHarms, a benchmark designed to systematically study human harm judgments across two key dimensions -- the harm axis (benign to harmful) and the agreement axis (agreement to disagreement). Our scalable framework generates prompts that capture diverse AI harms and human values while targeting cases with high disagreement rates, validated by human data. The benchmark includes 150 prompts with 15,000 ratings from 100 human annotators, enriched with demographic and psychological traits and prompt-level features of harmful actions, effects, and values. Our analyses show that prompts that relate to imminent risks and tangible harms amplify perceived harmfulness, while annotator traits (e.g., toxicity experience, education) and their interactions with prompt content explain systematic disagreement. We benchmark AI safety models and alignment methods on PluriHarms, finding that while personalization significantly improves prediction of human harm judgments, considerable room remains for future progress. By explicitly targeting value diversity and disagreement, our work provides a principled benchmark for moving beyond "one-size-fits-all" safety toward pluralistically safe AI.

Generative AI's humanlike qualities are driving its rapid adoption in professional domains. However, this anthropomorphic appeal raises concerns from HCI and responsible AI scholars about potential hazards and harms, such as overtrust in system outputs. To investigate how technology workers navigate these humanlike qualities and anticipate emergent harms, we conducted focus groups with 30 professionals across six job functions (ML engineering, product policy, UX research and design, product management, technology writing, and communications). Our findings reveal an unsettled knowledge environment surrounding humanlike generative AI, where workers' varying perspectives illuminate a range of potential risks for individuals, knowledge work fields, and society. We argue that workers require comprehensive support, including clearer conceptions of ``humanlikeness'' to effectively mitigate these risks. To aid in mitigation strategies, we provide a conceptual map articulating the identified hazards and their connection to conflated notions of ``humanlikeness.''

Big Data has become central to modern applications in finance, insurance, and cybersecurity, enabling machine learning systems to perform large-scale risk assessments and fraud detection. However, the increasing dependence on automated analytics introduces important concerns about transparency, regulatory compliance, and trust. This paper examines how explainable artificial intelligence (XAI) can be integrated into Big Data analytics pipelines for fraud detection and risk management. We review key Big Data characteristics and survey major analytical tools, including distributed storage systems, streaming platforms, and advanced fraud detection models such as anomaly detectors, graph-based approaches, and ensemble classifiers. We also present a structured review of widely used XAI methods, including LIME, SHAP, counterfactual explanations, and attention mechanisms, and analyze their strengths and limitations when deployed at scale. Based on these findings, we identify key research gaps related to scalability, real-time processing, and explainability for graph and temporal models. To address these challenges, we outline a conceptual framework that integrates scalable Big Data infrastructure with context-aware explanation mechanisms and human feedback. The paper concludes with open research directions in scalable XAI, privacy-aware explanations, and standardized evaluation methods for explainable fraud detection systems.

Reliability and Admissibility of AI-Generated Forensic Evidence in Criminal Trials

Sahibpreet Singh, Lalita Devi

This paper examines the admissibility of AI-generated forensic evidence in criminal trials. The growing adoption of AI presents promising results for investigative efficiency. Despite advancements, significant research gaps persist in practically understanding the legal limits of AI evidence in judicial processes. Existing literature lacks focused assessment of the evidentiary value of AI outputs. The objective of this study is to evaluate whether AI-generated evidence satisfies established legal standards of reliability. The methodology involves a comparative doctrinal legal analysis of evidentiary standards across common law jurisdictions. Preliminary results indicate that AI forensic tools can enhance scale of evidence analysis. However, challenges arise from reproducibility deficits. Courts exhibit variability in acceptance of AI evidence due to limited technical literacy and lack of standardized validation protocols. Liability implications reveal that developers and investigators may bear accountability for flawed outputs. This raises critical concerns related to wrongful conviction. The paper emphasizes the necessity of independent validation and, development of AI-specific admissibility criteria. Findings inform policy development for the responsible AI integration within criminal justice systems. The research advances the objectives of Sustainable Development Goal 16 by reinforcing equitable access to justice. Preliminary results contribute for a foundation for future empirical research in AI deployed criminal forensics.

Large Language Models (LLMs) have developed rapidly in web services, delivering unprecedented capabilities while amplifying societal risks. Existing works tend to focus on either isolated jailbreak attacks or static defenses, neglecting the dynamic interplay between evolving threats and safeguards in real-world web contexts. To mitigate these challenges, we propose ACE-Safety (Adversarial Co-Evolution for LLM Safety), a novel framework that jointly optimize attack and defense models by seamlessly integrating two key innovative procedures: (1) Group-aware Strategy-guided Monte Carlo Tree Search (GS-MCTS), which efficiently explores jailbreak strategies to uncover vulnerabilities and generate diverse adversarial samples; (2) Adversarial Curriculum Tree-aware Group Policy Optimization (AC-TGPO), which jointly trains attack and defense LLMs with challenging samples via curriculum reinforcement learning, enabling robust mutual improvement. Evaluations across multiple benchmarks demonstrate that our method outperforms existing attack and defense approaches, and provides a feasible pathway for developing LLMs that can sustainably support responsible AI ecosystems.

Case-based reasoning networks are machine-learning models that make predictions based on similarity between the input and prototypical parts of training samples, called prototypes. Such models are able to explain each decision by pointing to the prototypes that contributed the most to the final outcome. As the explanation is a core part of the prediction, they are often qualified as ``interpretable by design". While promising, we show that such explanations are sometimes misleading, which hampers their usefulness in safety-critical contexts. In particular, several instances may lead to different predictions and yet have the same explanation. Drawing inspiration from the field of formal eXplainable AI (FXAI), we propose Abductive Latent Explanations (ALEs), a formalism to express sufficient conditions on the intermediate (latent) representation of the instance that imply the prediction. Our approach combines the inherent interpretability of case-based reasoning models and the guarantees provided by formal XAI. We propose a solver-free and scalable algorithm for generating ALEs based on three distinct paradigms, compare them, and present the feasibility of our approach on diverse datasets for both standard and fine-grained image classification. The associated code can be found at https://github.com/julsoria/ale

Survival analysis is critical for cancer prognosis and treatment planning, yet existing methods lack the transparency essential for clinical adoption. While recent pathology agents have demonstrated explainability in diagnostic tasks, they face three limitations for survival prediction: inability to integrate multimodal data, ineffective region-of-interest exploration, and failure to leverage experiential learning from historical cases. We introduce SurvAgent, the first hierarchical chain-of-thought (CoT)-enhanced multi-agent system for multimodal survival prediction. SurvAgent consists of two stages: (1) WSI-Gene CoT-Enhanced Case Bank Construction employs hierarchical analysis through Low-Magnification Screening, Cross-Modal Similarity-Aware Patch Mining, and Confidence-Aware Patch Mining for pathology images, while Gene-Stratified analysis processes six functional gene categories. Both generate structured reports with CoT reasoning, storing complete analytical processes for experiential learning. (2) Dichotomy-Based Multi-Expert Agent Inference retrieves similar cases via RAG and integrates multimodal reports with expert predictions through progressive interval refinement. Extensive experiments on five TCGA cohorts demonstrate SurvAgent's superority over conventional methods, proprietary MLLMs, and medical agents, establishing a new paradigm for explainable AI-driven survival prediction in precision oncology.

From Narrow Unlearning to Emergent Misalignment: Causes, Consequences, and Containment in LLMs

Erum Mushtaq, Anil Ramakrishna, Satyapriya Krishna +5

Recent work has shown that fine-tuning on insecure code data can trigger an emergent misalignment (EMA) phenomenon, where models generate malicious responses even to prompts unrelated to the original insecure code-writing task. Such cross-domain generalization of harmful behavior underscores the need for a deeper understanding of the algorithms, tasks, and datasets that induce emergent misalignment. In this work, we extend this study by demonstrating that emergent misalignment can also arise from narrow refusal unlearning in specific domains. We perform refusal unlearning on Cybersecurity and Safety concept, and evaluate EMA by monitoring refusal scores across seven responsible AI (RAI) domains, Cybersecurity, Safety, Toxicity, Bias, Sensitive Content, Medical/Legal, and Privacy. Our work shows that narrow domain unlearning can yield compliance responses for the targeted concept, however, it may also propagate EMA to unrelated domains. Among the two intervened concepts, Cybersecurity and Safety, we find that the safety concept can have larger EMA impact, i.e, causing lower refusal scores, across other unrelated domains such as bias. We observe this effect consistently across two model families, Mistral-7b-0.3v, and Qwen-7b-2.5. Further, we show that refusal unlearning augmented with cross-entropy loss function on a small set of retain data from the affected domains can largely, if not fully, restore alignment across the impacted domains while having lower refusal rate on the concept we perform unlearning on. To investigate the underlying causes of EMA, we analyze concept entanglements at the representation level via concept vectors. Our analysis reveals that concepts with higher representation similarity in earlier layers are more susceptible to EMA after intervention when the refusal stream is altered through targeted refusal unlearning.

As AI systems become increasingly integrated into human lives, endowing them with robust social intelligence has emerged as a critical frontier. A key aspect of this intelligence is discerning truth from deception, a ubiquitous element of human interaction that is conveyed through a complex interplay of verbal language and non-verbal visual cues. However, automatic deception detection in dynamic, multi-party conversations remains a significant challenge. The recent rise of powerful Multimodal Large Language Models (MLLMs), with their impressive abilities in visual and textual understanding, makes them natural candidates for this task. Consequently, their capabilities in this crucial domain are mostly unquantified. To address this gap, we introduce a new task, Multimodal Interactive Veracity Assessment (MIVA), and present a novel multimodal dataset derived from the social deduction game Werewolf. This dataset provides synchronized video, text, with verifiable ground-truth labels for every statement. We establish a comprehensive benchmark evaluating state-of-the-art MLLMs, revealing a significant performance gap: even powerful models like GPT-4o struggle to distinguish truth from falsehood reliably. Our analysis of failure modes indicates that these models fail to ground language in visual social cues effectively and may be overly conservative in their alignment, highlighting the urgent need for novel approaches to building more perceptive and trustworthy AI systems.

Towards Human-AI Synergy in Requirements Engineering: A Framework and Preliminary Study

Mateen Ahmed Abbasi, Petri Ihantola, Tommi Mikkonen +1

The future of Requirements Engineering (RE) is increasingly driven by artificial intelligence (AI), reshaping how we elicit, analyze, and validate requirements. Traditional RE is based on labor-intensive manual processes prone to errors and complexity. AI-powered approaches, specifically large language models (LLMs), natural language processing (NLP), and generative AI, offer transformative solutions and reduce inefficiencies. However, the use of AI in RE also brings challenges like algorithmic bias, lack of explainability, and ethical concerns related to automation. To address these issues, this study introduces the Human-AI RE Synergy Model (HARE-SM), a conceptual framework that integrates AI-driven analysis with human oversight to improve requirements elicitation, analysis, and validation. The model emphasizes ethical AI use through transparency, explainability, and bias mitigation. We outline a multi-phase research methodology focused on preparing RE datasets, fine-tuning AI models, and designing collaborative human-AI workflows. This preliminary study presents the conceptual framework and early-stage prototype implementation, establishing a research agenda and practical design direction for applying intelligent data science techniques to semi-structured and unstructured RE data in collaborative environments.

Taming Silent Failures: A Framework for Verifiable AI Reliability

Guan-Yan Yang, Farn Wang

The integration of Artificial Intelligence (AI) into safety-critical systems introduces a new reliability paradigm: silent failures, where AI produces confident but incorrect outputs that can be dangerous. This paper introduces the Formal Assurance and Monitoring Environment (FAME), a novel framework that confronts this challenge. FAME synergizes the mathematical rigor of offline formal synthesis with the vigilance of online runtime monitoring to create a verifiable safety net around opaque AI components. We demonstrate its efficacy in an autonomous vehicle perception system, where FAME successfully detected 93.5% of critical safety violations that were otherwise silent. By contextualizing our framework within the ISO 26262 and ISO/PAS 8800 standards, we provide reliability engineers with a practical, certifiable pathway for deploying trustworthy AI. FAME represents a crucial shift from accepting probabilistic performance to enforcing provable safety in next-generation systems.

AI-assisted gait analysis holds promise for improving Parkinson's Disease (PD) care, but current clinical dashboards lack transparency and offer no meaningful way for clinicians to interrogate or contest AI decisions. To address this issue, we present Motion2Meaning, a clinician-centered framework that advances Contestable AI through a tightly integrated interface designed for interpretability, oversight, and procedural recourse. Our approach leverages vertical Ground Reaction Force (vGRF) time-series data from wearable sensors as an objective biomarker of PD motor states. The system comprises three key components: a Gait Data Visualization Interface (GDVI), a one-dimensional Convolutional Neural Network (1D-CNN) that predicts Hoehn & Yahr severity stages, and a Contestable Interpretation Interface (CII) that combines our novel Cross-Modal Explanation Discrepancy (XMED) safeguard with a contestable Large Language Model (LLM). Our 1D-CNN achieves 89.0% F1-score on the public PhysioNet gait dataset. XMED successfully identifies model unreliability by detecting a five-fold increase in explanation discrepancies in incorrect predictions (7.45%) compared to correct ones (1.56%), while our LLM-powered interface enables clinicians to validate correct predictions and successfully contest a portion of the model's errors. A human-centered evaluation of this contestable interface reveals a crucial trade-off between the LLM's factual grounding and its readability and responsiveness to clinical feedback. This work demonstrates the feasibility of combining wearable sensor analysis with Explainable AI (XAI) and contestable LLMs to create a transparent, auditable system for PD gait interpretation that maintains clinical oversight while leveraging advanced AI capabilities. Our implementation is publicly available at: https://github.com/hungdothanh/motion2meaning.