Research Papers

Probability Density from Latent Diffusion Models for Out-of-Distribution Detection

Joonas Järve, Karl Kaspar Haavel, Meelis Kull

Despite rapid advances in AI, safety remains the main bottleneck to deploying machine-learning systems. A critical safety component is out-of-distribution detection: given an input, decide whether it comes from the same distribution as the training data. In generative models, the most natural OOD score is the data likelihood. Actually, under the assumption of uniformly distributed OOD data, the likelihood is even the optimal OOD detector, as we show in this work. However, earlier work reported that likelihood often fails in practice, raising doubts about its usefulness. We explore whether, in practice, the representation space also suffers from the inability to learn good density estimation for OOD detection, or if it is merely a problem of the pixel space typically used in generative models. To test this, we trained a Variational Diffusion Model not on images, but on the representation space of a pre-trained ResNet-18 to assess the performance of our likelihood-based detector in comparison to state-of-the-art methods from the OpenOOD suite.

This paper is a collaborative piece between two worlds of expertise in the field of data visualization: accessibility and bias. In particular, the rise of generative models playing a role in accessibility is a worrying trend for data visualization. These models are increasingly used to help author visualizations as well as generate descriptions of existing visualizations for people who are blind, low vision, or use assistive technologies such as screen readers. Sighted human-to-human bias has already been established as an area of concern for theory, research, and design in data visualization. But what happens when someone is unable to verify the model output or adequately interrogate algorithmic bias, such as a context where a blind person asks a model to describe a chart for them? In such scenarios, trust from the user is not earned, rather reliance is compelled by the model-to-human relationship. In this work, we explored the dangers of AI-generated descriptions for accessibility, playing a game of telephone between models, observing bias production in model interpretation, and re-interpretation of a data visualization. We unpack ways that model failure in visualization is especially problematic for users with visual impairments, and suggest directions forward for three distinct readers of this piece: technologists who build model-assisted interfaces for end users, users with disabilities leveraging models for their own purposes, and researchers concerned with bias, accessibility, or visualization.

Social media data presents AI researchers with overlapping obligations under the GDPR, copyright law, and platform terms -- yet existing frameworks fail to integrate these regulatory domains, leaving researchers without unified guidance. We introduce PETLP (Privacy-by-design Extract, Transform, Load, and Present), a compliance framework that embeds legal safeguards directly into extended ETL pipelines. Central to PETLP is treating Data Protection Impact Assessments as living documents that evolve from pre-registration through dissemination. Through systematic Reddit analysis, we demonstrate how extraction rights fundamentally differ between qualifying research organisations (who can invoke DSM Article 3 to override platform restrictions) and commercial entities (bound by terms of service), whilst GDPR obligations apply universally. We demonstrate why true anonymisation remains unachievable for social media data and expose the legal gap between permitted dataset creation and uncertain model distribution. By structuring compliance decisions into practical workflows and simplifying institutional data management plans, PETLP enables researchers to navigate regulatory complexity with confidence, bridging the gap between legal requirements and research practice.

Public-sector bureaucracies seek to reap the benefits of artificial intelligence (AI), but face important concerns about accountability and transparency when using AI systems. In particular, perception or actuality of AI agency might create ethics sinks - constructs that facilitate dissipation of responsibility when AI systems of disputed moral status interface with bureaucratic structures. Here, we reject the notion that ethics sinks are a necessary consequence of introducing AI systems into bureaucracies. Rather, where they appear, they are the product of structural design decisions across both the technology and the institution deploying it. We support this claim via a systematic application of conceptions of moral agency in AI ethics to Weberian bureaucracy. We establish that it is both desirable and feasible to render AI systems as tools for the generation of organizational transparency and legibility, which continue the processes of Weberian rationalization initiated by previous waves of digitalization. We present a three-point Moral Agency Framework for legitimate integration of AI in bureaucratic structures: (a) maintain clear and just human lines of accountability, (b) ensure humans whose work is augmented by AI systems can verify the systems are functioning correctly, and (c) introduce AI only where it doesn't inhibit the capacity of bureaucracies towards either of their twin aims of legitimacy and stewardship. We suggest that AI introduced within this framework can not only improve efficiency and productivity while avoiding ethics sinks, but also improve the transparency and even the legitimacy of a bureaucracy.

Artificial intelligence (AI) systems increasingly support decision-making across critical domains, yet current explainable AI (XAI) approaches prioritize algorithmic transparency over human comprehension. While XAI methods reveal computational processes for model validation and audit, end users require explanations integrating domain knowledge, contextual reasoning, and professional frameworks. This disconnect reveals a fundamental design challenge: existing AI explanation approaches fail to address how practitioners actually need to understand and act upon recommendations. This paper introduces Explanatory AI as a complementary paradigm where AI systems leverage generative and multimodal capabilities to serve as explanatory partners for human understanding. Unlike traditional XAI that answers "How did the algorithm decide?" for validation purposes, Explanatory AI addresses "Why does this make sense?" for practitioners making informed decisions. Through theory-informed design, we synthesize multidisciplinary perspectives on explanation from cognitive science, communication research, and education with empirical evidence from healthcare contexts and AI expert interviews. Our analysis identifies five dimensions distinguishing Explanatory AI from traditional XAI: explanatory purpose (from diagnostic to interpretive sense-making), communication mode (from static technical to dynamic narrative interaction), epistemic stance (from algorithmic correspondence to contextual plausibility), adaptivity (from uniform design to personalized accessibility), and cognitive design (from information overload to cognitively aligned delivery). We derive five meta-requirements specifying what systems must achieve and formulate ten design principles prescribing how to build them.

There is a growing discussion about social media feeds being increasingly filled with AI-generated content. Due to its visual plausibility, low cost, and fast production speed, AI-generated content is said to be highly effective in "gaming the algorithm" and going viral. Popularly referred to as "AI slop," this phenomenon arguably leads to the presence of sloppy and potentially deceptive content at a scale unseen before. This investigation offers a systematic analysis of AI-generated content and its labelling in TikTok's and Instagram's search results across 13 hashtags (see Appendix) in three European countries (Spain, Germany, and Poland) over the course of June 2025. We manually annotated and analyzed the 30 top search results on political (#trump, #zelensky, #pope) and broader topics (e.g.,#health, #history) to understand the relation between synthetic (content that is partially or entirely made using generative AI) and non-synthetic content across languages and countries. We then explored the emerging phenomenon of accounts producing generative AI content at scale by analyzing 153 accounts and proposing a new categorization schema of what we termed Agentic AI Accounts. Our main findings are:

We propose a novel segmentation-based explainable artificial intelligence (XAI) method for neural networks working on point cloud classification. As one building block of this method, we propose a novel point-shifting mechanism to introduce perturbations in point cloud data. Recently, AI has seen an exponential growth. Hence, it is important to understand the decision-making process of AI algorithms when they are applied in critical areas. Our work focuses on explaining AI algorithms that classify point cloud data. An important aspect of the methods used for explaining AI algorithms is their ability to produce explanations that are easy for humans to understand. This allows them to analyze the AI algorithms better and make appropriate decisions based on that analysis. Therefore, in this work, we intend to generate meaningful explanations that can be easily interpreted by humans. The point cloud data we consider represents 3D objects such as cars, guitars, and laptops. We make use of point cloud segmentation models to generate explanations for the working of classification models. The segments are used to introduce perturbations into the input point cloud data and generate saliency maps. The perturbations are introduced using the novel point-shifting mechanism proposed in this work which ensures that the shifted points no longer influence the output of the classification algorithm. In contrast to previous methods, the segments used by our method are meaningful, i.e. humans can easily interpret the meaning of the segments. Thus, the benefit of our method over other methods is its ability to produce more meaningful saliency maps. We compare our method with the use of classical clustering algorithms to generate explanations. We also analyze the saliency maps generated for example inputs using our method to demonstrate the usefulness of the method in generating meaningful explanations.

As AI-driven dataspaces become integral to data sharing and collaborative analytics, ensuring privacy, performance, and policy compliance presents significant challenges. This paper provides a comprehensive review of privacy-preserving and policy-aware AI techniques, including Federated Learning, Differential Privacy, Trusted Execution Environments, Homomorphic Encryption, and Secure Multi-Party Computation, alongside strategies for aligning AI with regulatory frameworks such as GDPR and the EU AI Act. We propose a novel taxonomy to classify these techniques based on privacy levels, performance impacts, and compliance complexity, offering a clear framework for practitioners and researchers to navigate trade-offs. Key performance metrics -- latency, throughput, cost overhead, model utility, fairness, and explainability -- are analyzed to highlight the multi-dimensional optimization required in dataspaces. The paper identifies critical research gaps, including the lack of standardized privacy-performance KPIs, challenges in explainable AI for federated ecosystems, and semantic policy enforcement amidst regulatory fragmentation. Future directions are outlined, proposing a conceptual framework for policy-driven alignment, automated compliance validation, standardized benchmarking, and integration with European initiatives like GAIA-X, IDS, and Eclipse EDC. By synthesizing technical, ethical, and regulatory perspectives, this work lays the groundwork for developing trustworthy, efficient, and compliant AI systems in dataspaces, fostering innovation in secure and responsible data-driven ecosystems.

Fair and dynamic energy allocation in community microgrids remains a critical challenge, particularly when serving socio-economically diverse participants. Static optimization and cost-sharing methods often fail to adapt to evolving inequities, leading to participant dissatisfaction and unsustainable cooperation. This paper proposes a novel framework that integrates multi-objective mixed-integer linear programming (MILP), cooperative game theory, and a dynamic equity-adjustment mechanism driven by reinforcement learning (RL). At its core, the framework utilizes a bi-level optimization model grounded in Equity-regarding Welfare Maximization (EqWM) principles, which incorporate Rawlsian fairness to prioritize the welfare of the least advantaged participants. We introduce a Proximal Policy Optimization (PPO) agent that dynamically adjusts socio-economic weights in the optimization objective based on observed inequities in cost and renewable energy access. This RL-powered feedback loop enables the system to learn and adapt, continuously striving for a more equitable state. To ensure transparency, Explainable AI (XAI) is used to interpret the benefit allocations derived from a weighted Shapley value. Validated across six realistic scenarios, the framework demonstrates peak demand reductions of up to 72.6%, and significant cooperative gains. The adaptive RL mechanism further reduces the Gini coefficient over time, showcasing a pathway to truly sustainable and fair energy communities.

A Principled Approach for Defining and Comparing Variable Importance Measures

Angel Reyero-Lobo, Pierre Neuvial, Bertrand Thirion

Variable importance measures (VIMs) aim to quantify the contribution of each input covariate to the predictability of a given output. With the growing interest in explainable AI, numerous VIMs have been proposed, many of which are heuristic in nature. This is often justified by the inherent subjectivity of the notion of importance. This raises important questions regarding usage: What makes a good VIM? How can we compare different VIMs? In this paper, we address these questions by: (1) proposing an axiomatic framework that bridges the gap between variable importance and variable selection. This allows VIM methods to leverage the extensive literature on statistical guarantees developed for conditional independence testing. Our framework formalizes the intuitive principle that features providing no additional information should not be assigned importance, helping avoid false positives due to spurious correlations, which can arise with popular methods such as Shapley values; and (2) advocating for a general pipeline for constructing VIMs based on the definition of a theoretical index, its estimation, and the derivation of associated statistical guarantees. This pipeline clarifies the objective of various VIMs and thus facilitates meaningful comparisons. While this approach is natural from a statistical perspective, much of the literature has diverged from it, as we illustrate using the popular Total Sobol' Index. Finally, we apply this framework to an extensive set of VIMs, showing how practitioners can select and estimate indices aligned with their specific goals. Our results are supported by numerical experiments on simulated and real data.

Large Language Models (LLMs) excel in text classification, but their complexity hinders interpretability, making it difficult to understand the reasoning behind their predictions. Explainable AI (XAI) methods like LIME and SHAP offer local explanations by identifying influential words, but they rely on computationally expensive perturbations. These methods typically generate thousands of perturbed sentences and perform inferences on each, incurring a substantial computational burden, especially with LLMs. To address this, we propose \underline{P}erturbation-free \underline{L}ocal \underline{Ex}planation (PLEX), a novel method that leverages the contextual embeddings extracted from the LLM and a ``Siamese network" style neural network trained to align with feature importance scores. This one-off training eliminates the need for subsequent perturbations, enabling efficient explanations for any new sentence. We demonstrate PLEX's effectiveness on four different classification tasks (sentiment, fake news, fake COVID-19 news and depression), showing more than 92\% agreement with LIME and SHAP. Our evaluation using a ``stress test" reveals that PLEX accurately identifies influential words, leading to a similar decline in classification accuracy as observed with LIME and SHAP when these words are removed. Notably, in some cases, PLEX demonstrates superior performance in capturing the impact of key features. PLEX dramatically accelerates explanation, reducing time and computational overhead by two and four orders of magnitude, respectively. This work offers a promising solution for explainable LLM-based text classification.

The accelerating growth of photographic collections has outpaced manual cataloguing, motivating the use of vision language models (VLMs) to automate metadata generation. This study examines whether Al-generated catalogue descriptions can approximate human-written quality and how generative Al might integrate into cataloguing workflows in archival and museum collections. A VLM (InternVL2) generated catalogue descriptions for photographic prints on labelled cardboard mounts with archaeological content, evaluated by archive and archaeology experts and non-experts in a human-centered, experimental framework. Participants classified descriptions as AI-generated or expert-written, rated quality, and reported willingness to use and trust in AI tools. Classification performance was above chance level, with both groups underestimating their ability to detect Al-generated descriptions. OCR errors and hallucinations limited perceived quality, yet descriptions rated higher in accuracy and usefulness were harder to classify, suggesting that human review is necessary to ensure the accuracy and quality of catalogue descriptions generated by the out-of-the-box model, particularly in specialized domains like archaeological cataloguing. Experts showed lower willingness to adopt AI tools, emphasizing concerns on preservation responsibility over technical performance. These findings advocate for a collaborative approach where AI supports draft generation but remains subordinate to human verification, ensuring alignment with curatorial values (e.g., provenance, transparency). The successful integration of this approach depends not only on technical advancements, such as domain-specific fine-tuning, but even more on establishing trust among professionals, which could both be fostered through a transparent and explainable AI pipeline.