The research community lacks a middle ground between StarCraft II full game and its mini-games. The full-game's sprawling state-action space renders reward signals sparse and noisy, but in mini-games simple agents saturate performance. This complexity gap hinders steady curriculum design and prevents researchers from experimenting with modern Reinforcement Learning algorithms in RTS environments under realistic compute budgets. To fill this gap, we present the Two-Bridge Map Suite, the first entry in an open-source benchmark series we purposely engineered as an intermediate benchmark to sit between these extremes. By disabling economy mechanics such as resource collection, base building, and fog-of-war, the environment isolates two core tactical skills: long-range navigation and micro-combat. Preliminary experiments show that agents learn coherent maneuvering and engagement behaviors without imposing full-game computational costs. Two-Bridge is released as a lightweight, Gym-compatible wrapper on top of PySC2, with maps, wrappers, and reference scripts fully open-sourced to encourage broad adoption as a standard benchmark.

Multi-objective reinforcement learning (MORL) seeks to learn policies that balance multiple, often conflicting objectives. Although a single preference-conditioned policy is the most flexible and scalable solution, existing approaches remain brittle in practice, frequently failing to recover complete Pareto fronts. We show that this failure stems from two structural issues in current methods: destructive gradient interference caused by premature scalarization and representational collapse across the preference space. We introduce D^3PO, a PPO-based framework that reorganizes multi-objective policy optimization to address these issues directly. D^3PO preserves per-objective learning signals through a decomposed optimization pipeline and integrates preferences only after stabilization, enabling reliable credit assignment. In addition, a scaled diversity regularizer enforces sensitivity of policy behavior to preference changes, preventing collapse. Across standard MORL benchmarks, including high-dimensional and many-objective control tasks, D^3PO consistently discovers broader and higher-quality Pareto fronts than prior single- and multi-policy methods, matching or exceeding state-of-the-art hypervolume and expected utility while using a single deployable policy.

Introduction. While Privacy Policies and Terms of Service (ToS) are intended to inform users; they often overwhelm, mislead, and confuse in practice. This work investigates automated techniques for analyzing such legal documents, with the goal of supporting user comprehension and regulatory auditing. Method. We use expert-driven annotations from Terms of Service; Didn’t Read (ToS;DR) to train classification models that assign case labels to individual sentences. We also develop a classifier to distinguish document types: Privacy Policy, ToS, or other legal text. Analysis. Models were evaluated using F1 score, and we compared traditional fine-tuned models (RoBERTa, PrivBERT) against GPT-4 Turbo. We then applied the best-performing models to real-world policies to uncover conceptual overlaps between Privacy Policies and ToS. Results. Our case classifier achieved a 0.73 F1 score, while the document-type classifier reached 0.79. GPT-4 performed worse on case classification (0.58 F1). We found that GDPR-relevant clauses often appear in both Privacy Policies and ToS, blurring distinctions and raising risks for user misinterpretation and regulatory non-compliance. Conclusion. By surfacing hidden structures and overlapping clauses, our system enhances transparency and supports digital literacy by increasing accessibility of complex documents. This work lays the foundation for tools that promote user agency and platform accountability.

Motivations. Explainable AI (XAI) systems aim to improve users' understanding of AI, but XAI research has shown that many XAI explanations serve some users well while failing others. In non-AI systems, software practitioners have used inclusive design approaches to address similar problems, sometimes creating "curb-cut" improvements that benefit both underserved users and everyone else. This raises the possibility that inclusive design approaches can bring similar curb-cut improvements to AI explanations. Objectives. Our objective was to investigate possible curb-cut effects of inclusivity-driven fixes an AI product team made using an inclusive design approach (GenderMag) to improve their XAI prototype. Methods. We ran a between-subject study with 69 participants who had no formal AI background. 34 participants used the original version of the XAI prototype and the rest used the version with the AI team's inclusivity fixes. We then compared the two groups' mental model concepts scores and prediction accuracy, and the two prototypes' inclusivity. Results. Our investigation produced four main results. First, the AI team's inclusivity fixes were overall effective, resulting in overall better conceptual mental models with the new prototype. Further (second), the AI team's inclusivity fixes were particularly beneficial to the underserved population's conceptual mental models-which, together with the first result, constitutes a curb-cut effect. However (third), the inclusivity fixes did not improve participants' prediction accuracy scores. Instead, it appears to have harmed them overall-a "curb-fence" effect (opposite of a curb-cut effect). Finally (fourth), the AI team's fixes improved equity, reducing the gender gap by 45%.

Online surveys are a foundational data collection instrument in a variety of fields, with attention checks serving as critical guardians of response quality. However, the rapid emergence of agentic AI (goal directed systems powered by a large language model (LLM) brain and/or a multimodal processing unit with tool-augmented capabilities) raises new questions about the robustness of these safeguards. We investigate how well agentic AI architectures can complete web-based surveys and pass standard attention checks. We evaluate a single-agent architecture capable of multimodal input processing and tool-based web interaction on a controlled survey sandbox. We analyze the problem from two perspectives. From an attack perspective, we demonstrate how structural vulnerabilities such as exposed DOM metadata and predictable option encoding allow agents to resolve attention checks through structured parsing only. From a defense perspective, we implement a mitigation strategy of DOM metadata obfuscation to remove semantic cues in text-based questions. We evaluate multiple open-source language and multimodal models to study capability and orchestration effectiveness. Based on our evaluations, we offer perspectives on how to simultaneously meet the needs of empiricists and agentic AI researchers.

In general, Terms of Service (ToS) and other policy documents are verbose and full of legal jargon, which poses challenges for users to understand. To improve user accessibility and transparency, the "Terms of Service; Didn't Read" (ToS;DR) project condenses intricate legal terminology into summaries and overall grades for the website's policy documents. Nevertheless, uncertainties remain about whether users could truly grasp the implications of simplified presentations. We conducted an online survey to assess the perceived understandability and severity of randomly chosen cases from the ToS;DR taxonomy. Preliminary results indicate that, although most users report understanding the cases, they find a bias towards service providers in about two-thirds of the cases. The findings of our study emphasize the necessity of prioritizing user-centric policy formulation. This study has the potential to reveal the extent of information imbalance in digital services and promote more well-informed user consent.

With their advanced capabilities, Large Language Models (LLMs) can generate highly convincing and contextually relevant fake news, which can contribute to disseminating misinformation. Though there is much research on fake news detection for human-written text, the field of detecting LLM-generated fake news is still under-explored. This research measures the efficacy of detectors in identifying LLM-paraphrased fake news, in particular, determining whether adding a paraphrase step in the detection pipeline helps or impedes detection. This study contributes: (1) Detectors struggle to detect LLM-paraphrased fake news more than human-written text, (2) We find which models excel at which tasks (evading detection, paraphrasing to evade detection, and paraphrasing for semantic similarity). (3) Via LIME explanations, we discovered a possible reason for detection failures: sentiment shift. (4) We discover a worrisome trend for paraphrase quality measurement: samples that exhibit sentiment shift despite a high BERTSCORE. (5) We provide a pair of datasets augmenting existing datasets with paraphrase outputs and scores. The dataset is available on GitHub

The project's aim is to create an AI agent capable of selecting good actions in a game-playing domain called Battlespace. Sequential domains like Battlespace are important testbeds for planning problems, as such, the Department of Defense uses such domains for wargaming exercises. The agents we developed combine Monte Carlo Tree Search (MCTS) and Deep Q-Network (DQN) techniques in an effort to navigate the game environment, avoid obstacles, interact with adversaries, and capture the flag. This paper will focus on the encoding techniques we explored to present complex structured data stored in a Python class, a necessary precursor to an agent.

Assessing an AI system's behavior-particularly in Explainable AI Systems-is sometimes done empirically, by measuring people's abilities to predict the agent's next move-but how to perform such measurements? In empirical studies with humans, an obvious approach is to frame the task as binary (i.e., prediction is either right or wrong), but this does not scale. As output spaces increase, so do floor effects, because the ratio of right answers to wrong answers quickly becomes very small. The crux of the problem is that the binary framing is failing to capture the nuances of the different degrees of "wrongness." To address this, we begin by proposing three mathematical bases upon which to measure "partial wrongness." We then uses these bases to perform two analyses on sequential decision-making domains: the first is an in-lab study with 86 participants on a size-36 action space; the second is a re-analysis of a prior study on a size-4 action space. Other researchers adopting our operationalization of the prediction task and analysis methodology will improve the rigor of user studies conducted with that task, which is particularly important when the domain features a large output space.

The complexities of legalese in terms and policy documents can bind individuals to contracts they do not fully comprehend, potentially leading to uninformed data sharing. Our work seeks to alleviate this issue by developing language models that provide automated, accessible summaries and scores for such documents, aiming to enhance user understanding and facilitate informed decisions. We compared transformer-based and conventional models during training on our dataset, and RoBERTa performed better overall with a remarkable 0.74 F1-score. Leveraging our best-performing model, RoBERTa, we highlighted redundancies and potential guideline violations by identifying overlaps in GDPR-required documents, underscoring the necessity for stricter GDPR compliance.

We grapple with the question: How, for whom and why should explainable artificial intelligence (XAI) aim to support the user goal of agency? In particular, we analyze the relationship between agency and explanations through a user-centric lens through case studies and thought experiments. We find that explanation serves as one of several possible first steps for agency by allowing the user convert forethought to outcome in a more effective manner in future interactions. Also, we observe that XAI systems might better cater to laypersons, particularly "tinkerers", when combining explanations and user control, so they can make meaningful changes.

Case-based explanation (CBE) goes by many names, and in this paper I will argue why we should tend towards alternative choices when designing XAI systems. My argumentation for the stated claim rests on four broad points: (1) people seem to dislike CBE; (2) CBE relies on weak semantic linkage; (3) CBE is epistemically outmatched; (4) CBE is restrictive. This paper expounds on these arguments and concludes with thoughts about characteristics of possible alternatives.

Would you allow an AI agent to make decisions on your behalf? If the answer is “not always,” the next question becomes “in what circumstances”? Answering this question requires human users to be able to assess an AI agent—and not just with overall pass/fail assessments or statistics. Here users need to be able to localize an agent’s bugs so that they can determine when they are willing to rely on the agent and when they are not. After-Action Review for AI (AAR/AI), a new AI assessment process for integration with Explainable AI systems, aims to support human users in this endeavor, and in this article we empirically investigate AAR/AI’s effectiveness with domain-knowledgeable users. Our results show that AAR/AI participants not only located significantly more bugs than non-AAR/AI participants did (i.e., showed greater recall) but also located them more precisely (i.e., with greater precision). In fact, AAR/AI participants outperformed non-AAR/AI participants on every bug and were, on average, almost six times as likely as non-AAR/AI participants to find any particular bug. Finally, evidence suggests that incorporating labeling into the AAR/AI process may encourage domain-knowledgeable users to abstract above individual instances of bugs; we hypothesize that doing so may have contributed further to AAR/AI participants’ effectiveness.

Faced with several AI-powered sequential decision-making systems, how might someone choose on which to rely? For example, imagine car buyer Blair shopping for a self-driving car, or developer Dillon trying to choose an appropriate ML model to use in their application. Their first choice might be infeasible (i.e., too expensive in money or execution time), so they may need to select their second or third choice. To address this question, this paper presents: 1) Explanation Resolution, a quantifiable direct measurement concept; 2) a new XAI empirical task to measure explanations: "the Ranking Task"; and 3) a new strategy for inducing controllable agent variations—Mutant Agent Generation. In support of those main contributions, it also presents 4) novel explanations for sequential decision-making agents; 5) an adaptation to the AAR/AI assessment process; and 6) a qualitative study around these devices with 10 participants to investigate how they performed the Ranking Task on our mutant agents, using our explanations, and structured by AAR/AI. From an XAI researcher perspective, just as mutation testing can be applied to any code, mutant agent generation can be applied to essentially any neural network for which one wants to evaluate an assessment process or explanation type. As to an XAI user's perspective, the participants ranked the agents well overall, but showed the importance of high explanation resolution for close differences between agents. The participants also revealed the importance of supporting a wide diversity of explanation diets and agent "test selection" strategies.

Recently, researchers have made a number of tremendous advancements in AI capabilities. However, I will show that the fixed financial cost of making such advancements is high, and further, so are the recurring energy costs. As a result we see “AI haves and have nots” with wildly differing amounts of power between these two groups. This means our research community needs to carefully consider the role XAI has in mediating communication between stakeholders in the face of such an important power dynamic. This paper aims to engage that process, examining the current state of affairs through a variety of lenses, and then identifying some promising ideas for the future.

Smart systems, such as decision support or recommender systems, continue to prove challenging for people to understand, but are nonetheless ever more pervasive based on the promise of harnessing rich data sources that are becoming available in every domain. These systems tend to be opaque, raising important concerns about how to discover and account for fairness or bias issues. The workshop on Transparency and Explanations in Smart Systems (TExSS) welcomes researchers and practitioners interested in exchanging ideas for overcoming the design, development, and evaluation issues in intelligent user interfaces. Specifically, we will focus on barriers preventing better reliability, trainability, usability, trustworthiness, fairness, accountability, and transparency. This year's theme is "Responsible, Explainable AI for Inclusivity and Trust", emphasizing the importance of responsibility that tech-industry and developers have towards the design, implementation and evaluation of explainable, inclusive and trustworthy human-AI interaction.

How can we help domain-knowledgeable users who do not have expertise in AI analyze why an AI agent failed? Our research team previously developed a new structured process for such users to assess AI, called After-Action Review for AI (AAR/AI), consisting of a series of steps a human takes to assess an AI agent and formalize their understanding. In this paper, we investigate how the AAR/AI process can scale up to support reinforcement learning (RL) agents that operate in complex environments. We augment the AAR/AI process to be performed at three levels—episode-level, decision-level, and explanation-level—and integrate it into our redesigned visual analytics interface. We illustrate our approach through a usage scenario of analyzing why a RL agent lost in a complex real-time strategy game built with the StarCraft 2 engine. We believe integrating structured processes like AAR/AI into visualization tools can help visualization play a more critical role in AI interpretability.