Research Papers

The field of AI safety seeks to prevent or reduce the harms caused by AI systems. A simple and appealing account of what is distinctive of AI safety as a field holds that this feature is constitutive: a research project falls within the purview of AI safety just in case it aims to prevent or reduce the harms caused by AI systems. Call this appealingly simple account The Safety Conception of AI safety. Despite its simplicity and appeal, we argue that The Safety Conception is in tension with at least two trends in the ways AI safety researchers and organizations think and talk about AI safety: first, a tendency to characterize the goal of AI safety research in terms of catastrophic risks from future systems; second, the increasingly popular idea that AI safety can be thought of as a branch of safety engineering. Adopting the methodology of conceptual engineering, we argue that these trends are unfortunate: when we consider what concept of AI safety it would be best to have, there are compelling reasons to think that The Safety Conception is the answer. Descriptively, The Safety Conception allows us to see how work on topics that have historically been treated as central to the field of AI safety is continuous with work on topics that have historically been treated as more marginal, like bias, misinformation, and privacy. Normatively, taking The Safety Conception seriously means approaching all efforts to prevent or mitigate harms from AI systems based on their merits rather than drawing arbitrary distinctions between them.

Recent advances in AI research make it increasingly plausible that artificial agents with consequential real-world impact will soon operate beyond tightly controlled environments. Ensuring that these agents are not only safe but that they adhere to broader normative expectations is thus an urgent interdisciplinary challenge. Multiple fields -- notably AI Safety, AI Alignment, and Machine Ethics -- claim to contribute to this task. However, the conceptual boundaries and interrelations among these domains remain vague, leaving researchers without clear guidance in positioning their work. To address this meta-challenge, we develop a structured conceptual framework for understanding AI alignment. Rather than focusing solely on alignment goals, we introduce a taxonomy distinguishing the alignment aim (safety, ethicality, legality, etc.), scope (outcome vs. execution), and constituency (individual vs. collective). This structural approach reveals multiple legitimate alignment configurations, providing a foundation for practical and philosophical integration across domains, and clarifying what it might mean for an agent to be aligned all-things-considered.

Evaluating synthetic tabular data is challenging, since they can differ from the real data in so many ways. There exist numerous metrics of synthetic data quality, ranging from statistical distances to predictive performance, often providing conflicting results. Moreover, they fail to explain or pinpoint the specific weaknesses in the synthetic data. To address this, we apply explainable AI (XAI) techniques to a binary detection classifier trained to distinguish real from synthetic data. While the classifier identifies distributional differences, XAI concepts such as feature importance and feature effects, analyzed through methods like permutation feature importance, partial dependence plots, Shapley values and counterfactual explanations, reveal why synthetic data are distinguishable, highlighting inconsistencies, unrealistic dependencies, or missing patterns. This interpretability increases transparency in synthetic data evaluation and provides deeper insights beyond conventional metrics, helping diagnose and improve synthetic data quality. We apply our approach to two tabular datasets and generative models, showing that it uncovers issues overlooked by standard evaluation techniques.

International cooperation is common in AI research, including between geopolitical rivals. While many experts advocate for greater international cooperation on AI safety to address shared global risks, some view cooperation on AI with suspicion, arguing that it can pose unacceptable risks to national security. However, the extent to which cooperation on AI safety poses such risks, as well as provides benefits, depends on the specific area of cooperation. In this paper, we consider technical factors that impact the risks of international cooperation on AI safety research, focusing on the degree to which such cooperation can advance dangerous capabilities, result in the sharing of sensitive information, or provide opportunities for harm. We begin by why nations historically cooperate on strategic technologies and analyse current US-China cooperation in AI as a case study. We further argue that existing frameworks for managing associated risks can be supplemented with consideration of key risks specific to cooperation on technical AI safety research. Through our analysis, we find that research into AI verification mechanisms and shared protocols may be suitable areas for such cooperation. Through this analysis we aim to help researchers and governments identify and mitigate the risks of international cooperation on AI safety research, so that the benefits of cooperation can be fully realised.

Understanding the behavior of black-box data-driven controllers is a key challenge in modern control design. In this work, we propose an eXplainable AI (XAI) methodology based on Inverse Optimal Control (IOC) to obtain local explanations for the behavior of a controller operating around a given region. Specifically, we extract the weights assigned to tracking errors and control effort in the implicit cost function that a black-box controller is optimizing, offering a more transparent and interpretable representation of the controller's underlying objectives. This approach presents connections with well-established XAI techniques, such as Local Interpretable Model-agnostic Explanations (LIME) since it is still based on a local approximation of the control policy. However, rather being limited to a standard sensitivity analysis, the explanation provided by our method relies on the solution of an inverse Linear Quadratic (LQ) problem, offering a structured and more control-relevant perspective. Numerical examples demonstrate that the inferred cost function consistently provides a deeper understanding of the controller's decision-making process, shedding light on otherwise counterintuitive or unexpected phenomena.

Social interactions among classroom peers, represented as social learning networks (SLNs), play a crucial role in enhancing learning outcomes. While SLN analysis has recently garnered attention, most existing approaches rely on centralized training, where data is aggregated and processed on a local/cloud server with direct access to raw data. However, in real-world educational settings, such direct access across multiple classrooms is often restricted due to privacy concerns. Furthermore, training models on isolated classroom data prevents the identification of common interaction patterns that exist across multiple classrooms, thereby limiting model performance. To address these challenges, we propose one of the first frameworks that integrates Federated Learning (FL), a distributed and collaborative machine learning (ML) paradigm, with SLNs derived from students' interactions in multiple classrooms' online forums to predict future link formations (i.e., interactions) among students. By leveraging FL, our approach enables collaborative model training across multiple classrooms while preserving data privacy, as it eliminates the need for raw data centralization. Recognizing that each classroom may exhibit unique student interaction dynamics, we further employ model personalization techniques to adapt the FL model to individual classroom characteristics. Our results demonstrate the effectiveness of our approach in capturing both shared and classroom-specific representations of student interactions in SLNs. Additionally, we utilize explainable AI (XAI) techniques to interpret model predictions, identifying key factors that influence link formation across different classrooms. These insights unveil the drivers of social learning interactions within a privacy-preserving, collaborative, and distributed ML framework -- an aspect that has not been explored before.

Learning in Structured Stackelberg Games

Maria-Florina Balcan, Kiriaki Fragkia, Keegan Harris

We initiate the study of structured Stackelberg games, a novel form of strategic interaction between a leader and a follower where contextual information can be predictive of the follower's (unknown) type. Motivated by applications such as security games and AI safety, we show how this additional structure can help the leader learn a utility-maximizing policy in both the online and distributional settings. In the online setting, we first prove that standard learning-theoretic measures of complexity do not characterize the difficulty of the leader's learning task. Notably, we find that there exists a learning-theoretic measure of complexity, analogous to the Littlestone dimension in online classification, that tightly characterizes the leader's instance-optimal regret. We term this the Stackelberg-Littlestone dimension, and leverage it to provide a provably optimal online learning algorithm. In the distributional setting, we provide analogous results by showing that two new dimensions control the sample complexity upper- and lower-bound.

Cancer remains one of the leading causes of mortality worldwide, necessitating accurate diagnosis and prognosis. Whole Slide Imaging (WSI) has become an integral part of clinical workflows with advancements in digital pathology. While various studies have utilized WSIs, their extracted features may not fully capture the most relevant pathological information, and their lack of interpretability limits clinical adoption. In this paper, we propose PATH-X, a framework that integrates Vision Transformers (ViT) and Autoencoders with SHAP (Shapley Additive Explanations) to enhance model explainability for patient stratification and risk prediction using WSIs from The Cancer Genome Atlas (TCGA). A representative image slice is selected from each WSI, and numerical feature embeddings are extracted using Google's pre-trained ViT. These features are then compressed via an autoencoder and used for unsupervised clustering and classification tasks. Kaplan-Meier survival analysis is applied to evaluate stratification into two and three risk groups. SHAP is used to identify key contributing features, which are mapped onto histopathological slices to provide spatial context. PATH-X demonstrates strong performance in breast and glioma cancers, where a sufficient number of WSIs enabled robust stratification. However, performance in lung cancer was limited due to data availability, emphasizing the need for larger datasets to enhance model reliability and clinical applicability.

Connecting Feedback to Choice: Understanding Educator Preferences in GenAI vs. Human-Created Lesson Plans in K-12 Education -- A Comparative Analysis

Shawon Sarkar, Min Sun, Alex Liu +4

As generative AI (GenAI) models are increasingly explored for educational applications, understanding educator preferences for AI-generated lesson plans is critical for their effective integration into K-12 instruction. This exploratory study compares lesson plans authored by human curriculum designers, a fine-tuned LLaMA-2-13b model trained on K-12 content, and a customized GPT-4 model to evaluate their pedagogical quality across multiple instructional measures: warm-up activities, main tasks, cool-down activities, and overall quality. Using a large-scale preference study with K-12 math educators, we examine how preferences vary across grade levels and instructional components. We employ both qualitative and quantitative analyses. The raw preference results indicate that human-authored lesson plans are generally favored, particularly for elementary education, where educators emphasize student engagement, scaffolding, and collaborative learning. However, AI-generated models demonstrate increasing competitiveness in cool-down tasks and structured learning activities, particularly in high school settings. Beyond quantitative results, we conduct thematic analysis using LDA and manual coding to identify key factors influencing educator preferences. Educators value human-authored plans for their nuanced differentiation, real-world contextualization, and student discourse facilitation. Meanwhile, AI-generated lesson plans are often praised for their structure and adaptability for specific instructional tasks. Findings suggest a human-AI collaborative approach to lesson planning, where GenAI can serve as an assistive tool rather than a replacement for educator expertise in lesson planning. This study contributes to the growing discourse on responsible AI integration in education, highlighting both opportunities and challenges in leveraging GenAI for curriculum development.

Concept-based eXplainable AI (C-XAI) is a rapidly growing research field that enhances AI model interpretability by leveraging intermediate, human-understandable concepts. This approach not only enhances model transparency but also enables human intervention, allowing users to interact with these concepts to refine and improve the model's performance. Concept Bottleneck Models (CBMs) explicitly predict concepts before making final decisions, enabling interventions to correct misclassified concepts. While CBMs remain effective in Out-Of-Distribution (OOD) settings with intervention, they struggle to match the performance of black-box models. Concept Embedding Models (CEMs) address this by learning concept embeddings from both concept predictions and input data, enhancing In-Distribution (ID) accuracy but reducing the effectiveness of interventions, especially in OOD scenarios. In this work, we propose the Variational Concept Embedding Model (V-CEM), which leverages variational inference to improve intervention responsiveness in CEMs. We evaluated our model on various textual and visual datasets in terms of ID performance, intervention responsiveness in both ID and OOD settings, and Concept Representation Cohesiveness (CRC), a metric we propose to assess the quality of the concept embedding representations. The results demonstrate that V-CEM retains CEM-level ID performance while achieving intervention effectiveness similar to CBM in OOD settings, effectively reducing the gap between interpretability (intervention) and generalization (performance).

Model documentation plays a crucial role in promoting transparency and responsible development of AI systems. With the rise of Generative AI (GenAI), open-source platforms have increasingly become hubs for hosting and distributing these models, prompting platforms like Hugging Face to develop dedicated model documentation guidelines that align with responsible AI principles. Despite these growing efforts, there remains a lack of understanding of how developers document their GenAI models on open-source platforms. Through interviews with 13 GenAI developers active on open-source platforms, we provide empirical insights into their documentation practices and challenges. Our analysis reveals that despite existing resources, developers of GenAI models still face multiple layers of uncertainties in their model documentation: (1) uncertainties about what specific content should be included; (2) uncertainties about how to effectively report key components of their models; and (3) uncertainties in deciding who should take responsibilities for various aspects of model documentation. Based on our findings, we discuss the implications for policymakers, open-source platforms, and the research community to support meaningful, effective and actionable model documentation in the GenAI era, including cultivating better community norms, building robust evaluation infrastructures, and clarifying roles and responsibilities.

Federated Learning (FL) enables collaborative model training while preserving data privacy; however, balancing privacy preservation (PP) and fairness poses significant challenges. In this paper, we present the first unified large-scale empirical study of privacy-fairness-utility trade-offs in FL, advancing toward responsible AI deployment. Specifically, we systematically compare Differential Privacy (DP), Homomorphic Encryption (HE), and Secure Multi-Party Computation (SMC) with fairness-aware optimizers including q-FedAvg, q-MAML, Ditto, evaluating their performance under IID and non-IID scenarios using benchmark (MNIST, Fashion-MNIST) and real-world datasets (Alzheimer's MRI, credit-card fraud detection). Our analysis reveals HE and SMC significantly outperform DP in achieving equitable outcomes under data skew, although at higher computational costs. Remarkably, we uncover unexpected interactions: DP mechanisms can negatively impact fairness, and fairness-aware optimizers can inadvertently reduce privacy effectiveness. We conclude with practical guidelines for designing robust FL systems that deliver equitable, privacy-preserving, and accurate outcomes.