Research Papers

As artificial intelligence becomes increasingly pervasive and powerful, the ability to audit AI-based systems is growing in importance. However, explainability for artificial intelligence systems is not a one-size-fits-all solution; different target audiences have varying requirements and expectations for explanations. While various approaches to explainability have been proposed, most explainable artificial intelligence methods for tabular data focus on explaining the outputs of supervised machine learning models using the input features. However, a user's ability to understand an explanation depends on their understanding of such features. Therefore, it is in the best interest of the system designer to try to pre-select understandable features for producing a global explanation of an ML model. Unfortunately, no measure currently exists to assess the degree to which a user understands a given input feature. This work introduces two psychometrically validated scales that quantitatively seek to assess users' understanding of tabular input features for supervised classification problems. Specifically, these scales, one for numerical and one for categorical data, each with two factors and comprising 8 and 9 items, aim to assign a score to each input feature, effectively producing a rank, and allowing for the quantification of feature prioritisation. A confirmatory factor analysis demonstrates a strong relationship between such items and a good fit of the two-factor structure for each scale. This research presents a novel method for assessing understanding and outlines potential applications in the domain of explainable artificial intelligence.

The Universal Declaration of Human Rights and other international agreements outline numerous inalienable rights that apply across geopolitical boundaries. As generative AI becomes increasingly prevalent, it poses risks to human rights such as non-discrimination, health, and security, which are also central concerns for AI researchers focused on fairness and safety. We contribute to the field of algorithmic auditing by presenting a framework to computationally assess human rights risk. Drawing on the UN Guiding Principles on Business and Human Rights, we develop an approach to evaluating a model to make grounded claims about the level of risk a model poses to particular human rights. Our framework consists of three parts: selecting tasks that are likely to pose human rights risks within a given context, designing metrics to measure the scope, scale, and likelihood of potential risks from that task, and analyzing rights with respect to the values of those metrics. Because a human rights approach centers on real-world harms, it requires evaluating AI systems in the specific contexts in which they are deployed. We present a case study of large language models in political news journalism, demonstrating how our framework helps to design an evaluation and benchmarking different models. We then discuss the implications of the results for the rights of access to information and freedom of thought and broader considerations for adopting this approach.

Möbius inversion and Shapley values are two mathematical tools for characterizing and decomposing higher-order structure in complex systems. The former defines higher-order interactions as discrete derivatives over a partial order; the latter provides a principled way to attribute those interactions back to the `atomic' elements of the system. Both have found wide application, from combinatorics and cooperative game theory to machine learning and explainable AI. We generalize both tools simultaneously in two orthogonal directions: 1) from real-valued functions to functions valued in any abelian group (in particular, vector-valued functions), and 2) from partial orders and lattices to directed acyclic multigraphs (DAMGs) and weighted versions thereof. The classical axioms, linearity, efficiency, null player, and symmetry, which uniquely characterize Shapley values on lattices, are insufficient in this more general setting. We resolve this by introducing projection operators that recursively re-attribute higher-order synergies down to the roots of the graph, and by proposing two natural axioms: weak elements (coalitions with zero synergy can be removed without affecting any attribution) and flat hierarchy (on graphs with no intermediate hierarchy, attributions are distributed proportionally to edge counts). Together with linearity, these three axioms uniquely determine the Shapley values via a simple explicit formula, while automatically implying efficiency, null player, symmetry, and a novel projection property. The resulting framework recovers all existing lattice-based definitions as special cases, and naturally handles settings, such as games on non-lattice partial orders, which were previously out of reach. The extension to vector-valued functions and general DAMG-structured hierarchies opens new application areas in machine learning, natural language processing, and explainable AI.

In real-world applications, computational constraints often require transforming large models into smaller, more efficient versions through model compression. While these techniques aim to reduce size and computational cost without sacrificing performance, their evaluations have traditionally focused on the trade-off between size and accuracy, overlooking the aspect of model faithfulness. This limited view is insufficient for high-stakes domains like healthcare, finance, and criminal justice, where compressed models must remain faithful to the behavior of their original counterparts. This paper presents a novel approach to evaluating faithfulness in compressed models, moving beyond standard metrics. We introduce and demonstrate a set of faithfulness metrics that capture how model behavior changes post-compression. Our contributions include introducing techniques to assess predictive consistency between the original and compressed models using model agreement, and applying chi-squared tests to detect statistically significant changes in predictive patterns across both the overall dataset and demographic subgroups, thereby exposing shifts that aggregate fairness metrics may obscure. We demonstrate our approaches by applying quantization and pruning to artificial neural networks (ANNs) trained on three diverse and socially meaningful datasets. Our findings show that high accuracy does not guarantee faithfulness, and our statistical tests detect subtle yet significant shifts that are missed by standard metrics, such as Accuracy and Equalized Odds. The proposed metrics provide a practical and more direct method for ensuring that efficiency gains through compression do not compromise the fairness or faithfulness essential for trustworthy AI.

Systems analysis students increasingly use Generative AI, yet current pedagogy lacks systematic approaches for teaching responsible AI orchestration that fosters critical thinking whilst meeting educational outcomes. Students risk accepting AI suggestions blindly or uncritically without assessing alignment with user needs or contextual appropriateness. SAGE (Structured AI-Guided Education) addresses this gap by embedding GenAI into curriculum design, training students when to accept, modify, or reject AI contributions. Implementation with 18 student groups across four Australian universities revealed how orchestration skills develop. Most groups (84\%) moved beyond passive acceptance, showing selective judgment, yet none proactively identified gaps overlooked by both human and AI analysis, indicating a competency ceiling. Students strong at explaining decisions also performed well at integrating sources, and those with deep domain understanding consistently considered accessibility considerations. Accessibility awareness proved fragile. When writing requirements, 85\% of groups explicitly considered elderly users and cultural needs. Notably, 55\% of groups struggled identifying when AI misclassified system boundaries (what belongs inside versus outside the system), 45\% missed data management errors (how information is stored and updated), and 55\% overlooked missing exception handling. Three implications emerge for educators: (i) require students to document why they accepted, modified, or rejected each AI suggestion, making reasoning explicit; (ii) embed accessibility prompts at each development stage because awareness collapses without continuous scaffolding; and (iii) have students create their own specifications before using AI, then compare versions, and anchor to research or standards to identify gaps.

Disclosure and Evaluation as Fairness Interventions for General-Purpose AI

Vyoma Raman, Judy Hanwen Shen, Andy K. Zhang +4

Despite conflicting definitions and conceptions of fairness, AI fairness researchers broadly agree that fairness is context-specific. However, when faced with general-purpose AI, which by definition serves a range of contexts, how should we think about fairness? We argue that while we cannot be prescriptive about what constitutes fair outcomes, we can specify the processes that different stakeholders should follow in service of fairness. Specifically, we consider the obligations of two major groups: system providers and system deployers. While system providers are natural candidates for regulatory attention, the current state of AI understanding offers limited insight into how upstream factors translate into downstream fairness impacts. Thus, we recommend that providers invest in evaluative research studying how model development decisions influence fairness and disclose whom they are serving their models to, or at the very least, reveal sufficient information for external researchers to conduct such research. On the other hand, system deployers are closer to real-world contexts and can leverage their proximity to end users to address fairness harms in different ways. Here, we argue they should responsibly disclose information about users and personalization and conduct rigorous evaluations across different levels of fairness. Overall, instead of focusing on enforcing fairness outcomes, we prioritize intentional information-gathering by system providers and deployers that can facilitate later context-aware action. This allows us to be specific and concrete about the processes even while the contexts remain unknown. Ultimately, this approach can sharpen how we distribute fairness responsibilities and inform more fluid, context-sensitive interventions as AI continues to advance.

As Large Language Models (LLMs) gain wider adoption, ensuring their reliable handling of Personally Identifiable Information (PII) across diverse regulatory contexts has become essential. This work introduces a scalable multilingual data curation framework designed for high-quality PII annotation across 13 underrepresented locales, covering approximately 336 locale-specific PII types. Our phased, human-in-the-loop annotation methodology combines linguistic expertise with rigorous quality assurance, leading to substantial improvements in recall and false positive rates from pilot, training, and production phases. By leveraging inter-annotator agreement metrics and root-cause analysis, the framework systematically uncovers and resolves annotation inconsistencies, resulting in high-fidelity datasets suitable for supervised LLM fine-tuning. Beyond reporting empirical gains, we highlight common annotator challenges in multilingual PII labeling and demonstrate how iterative, analytics-driven pipelines can enhance both annotation quality and downstream model reliability.

ZK-WAGON: Imperceptible Watermark for Image Generation Models using ZK-SNARKs

Aadarsh Anantha Ramakrishnan, Shubham Agarwal, Selvanayagam S +1

As image generation models grow increasingly powerful and accessible, concerns around authenticity, ownership, and misuse of synthetic media have become critical. The ability to generate lifelike images indistinguishable from real ones introduces risks such as misinformation, deepfakes, and intellectual property violations. Traditional watermarking methods either degrade image quality, are easily removed, or require access to confidential model internals - making them unsuitable for secure and scalable deployment. We are the first to introduce ZK-WAGON, a novel system for watermarking image generation models using the Zero-Knowledge Succinct Non Interactive Argument of Knowledge (ZK-SNARKs). Our approach enables verifiable proof of origin without exposing model weights, generation prompts, or any sensitive internal information. We propose Selective Layer ZK-Circuit Creation (SL-ZKCC), a method to selectively convert key layers of an image generation model into a circuit, reducing proof generation time significantly. Generated ZK-SNARK proofs are imperceptibly embedded into a generated image via Least Significant Bit (LSB) steganography. We demonstrate this system on both GAN and Diffusion models, providing a secure, model-agnostic pipeline for trustworthy AI image generation.

This study explores the integration of AI, particularly large language models (LLMs) like ChatGPT, into educational settings, focusing on the implications for teaching and learning. Through interviews with course coordinators from data science courses at Wageningen University, this research identifies both the benefits and challenges associated with AI in the classroom. While AI tools can streamline tasks and enhance learning, concerns arise regarding students' overreliance on these technologies, potentially hindering the development of essential cognitive and problem solving skills. The study highlights the importance of responsible AI usage, ethical considerations, and the need for adapting assessment methods to ensure educational outcomes are met. With careful integration, AI can be a valuable asset in education, provided it is used to complement rather than replace fundamental learning processes.

The European Union's AI Act represents a crucial step towards regulating ethical and responsible AI systems. However, we find an absence of quantifiable fairness metrics and the ambiguity in terminology, particularly the interchangeable use of the keywords transparency, explainability, and interpretability in the new EU AI Act and no reference of transparency of ethical compliance. We argue that this ambiguity creates substantial liability risk that would deter investment. Fairness transparency is strategically important. We recommend a more tailored regulatory framework to enhance the new EU AI regulation. Further-more, we propose a public system framework to assess the fairness and transparency of AI systems. Drawing from past work, we advocate for the standardization of industry best practices as a necessary addition to broad regulations to achieve the level of details required in industry, while preventing stifling innovation and investment in the AI sector. The proposals are exemplified with the case of ASR and speech synthesizers.

Prominent AI companies are producing 'safety frameworks' as a type of voluntary self-governance. These statements purport to establish risk thresholds and safety procedures for the development and deployment of highly capable AI. Understanding which AI risks are covered and what actions are allowed, refused, demanded, encouraged, or discouraged by these statements is vital for assessing how these frameworks actually govern AI development and deployment. We draw on affordance theory to analyse the OpenAI 'Preparedness Framework Version 2' (April 2025) using the Mechanisms & Conditions model of affordances and the MIT AI Risk Repository. We find that this safety policy requests evaluation of a small minority of AI risks, encourages deployment of systems with 'Medium' capabilities for unintentionally enabling 'severe harm' (which OpenAI defines as >1000 deaths or >$100B in damages), and allows OpenAI's CEO to deploy even more dangerous capabilities. These findings suggest that effective mitigation of AI risks requires more robust governance interventions beyond current industry self-regulation. Our affordance analysis provides a replicable method for evaluating what safety frameworks actually permit versus what they claim.

Clean First, Align Later: Benchmarking Preference Data Cleaning for Reliable LLM Alignment

Samuel Yeh, Sharon Li

Human feedback plays a pivotal role in aligning large language models (LLMs) with human preferences. However, such feedback is often noisy or inconsistent, which can degrade the quality of reward models and hinder alignment. While various automated data cleaning methods have been proposed to mitigate this issue, a systematic evaluation of their effectiveness and generalizability remains lacking. To bridge this gap, we introduce the first comprehensive benchmark for evaluating 13 preference data cleaning methods in the context of LLM alignment. PrefCleanBench offers a standardized protocol to assess cleaning strategies in terms of alignment performance and generalizability across diverse datasets, model architectures, and optimization algorithms. By unifying disparate methods and rigorously comparing them, we uncover key factors that determine the success of data cleaning in alignment tasks. This benchmark lays the groundwork for principled and reproducible approaches to improving LLM alignment through better data quality-highlighting the crucial but underexplored role of data preprocessing in responsible AI development. We release modular implementations of all methods to catalyze further research: https://github.com/deeplearning-wisc/PrefCleanBench.