How should reinforcement learning (RL) agents explain themselves to humans not trained in AI? To gain insights into this question, we conducted a 124-participant, four-treatment experiment to compare participants’ mental models of an RL agent in the context of a simple Real-Time Strategy (RTS) game. The four treatments isolated two types of explanations vs. neither vs. both together. The two types of explanations were as follows: (1) saliency maps (an “Input Intelligibility Type” that explains the AI’s focus of attention) and (2) reward-decomposition bars (an “Output Intelligibility Type” that explains the AI’s predictions of future types of rewards). Our results show that a combined explanation that included saliency and reward bars was needed to achieve a statistically significant difference in participants’ mental model scores over the no-explanation treatment. However, this combined explanation was far from a panacea: It exacted disproportionately high cognitive loads from the participants who received the combined explanation. Further, in some situations, participants who saw both explanations predicted the agent’s next action worse than all other treatments’ participants.
24 September 2024 in ACM Transactions on Interactive Intelligent Systems
Motivations : Recent research has emerged on generally how to improve AI products’ human-AI interaction (HAI) user experience (UX), but relatively little is known about HAI-UX inclusivity. For example, what kinds of users are supported, and who are left out? What product changes would make it more inclusive? Objectives : To help fill this gap, we present an approach to measuring what kinds of diverse users an AI product leaves out and how to act upon that knowledge. To bring actionability to the results, the approach focuses on users’ problem-solving diversity. Thus, our specific objectives were (1) to show how the measure can reveal which participants with diverse problem-solving styles were left behind in a set of AI products and (2) to relate participants’ problem-solving diversity to their demographic diversity, specifically gender and age. Methods : We performed 18 experiments, discarding two that failed manipulation checks. Each experiment was a 2 \(\times\) 2 factorial experiment with online participants, comparing two AI products: one deliberately violating 1 of 18 HAI guidelines and the other applying the same guideline. For our first objective, we used our measure to analyze how much each AI product gained/lost HAI-UX inclusivity compared to its counterpart, where inclusivity meant supportiveness to participants with particular problem-solving styles. For our second objective, we analyzed how participants’ problem-solving styles aligned with their gender identities and ages. Results and Implications : Participants’ diverse problem-solving styles revealed six types of inclusivity results: (1) the AI products that followed an HAI guideline were almost always more inclusive across diversity of problem-solving styles than the products that did not follow that guideline—but “who” got most of the inclusivity varied widely by guideline and by problem-solving style; (2) when an AI product had risk implications, four variables’ values varied in tandem: participants’ feelings of control, their (lack of) suspicion, their trust in the product, and their certainty while using the product; (3) the more control an AI product offered users, the more inclusive it was; (4) whether an AI product was learning from “my” data or other people’s affected how inclusive that product was; (5) participants’ problem-solving styles skewed differently by gender and age group; and (6) almost all of the results suggested actions that HAI practitioners could take to improve their products’ inclusivity further. Together, these results suggest that a key to improving the demographic inclusivity of an AI product (e.g., across a wide range of genders, ages) can often be obtained by improving the product’s support of diverse problem-solving styles.
9 January 2026 in ACM Transactions on Interactive Intelligent Systems
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%.
15 March 2021 in ACM Transactions on Interactive Intelligent Systems
Assessing and understanding intelligent agents is a difficult task for users who lack an AI background. “Explainable AI” (XAI) aims to address this problem, but what should be in an explanation? One route toward answering this question is to turn to theories of how humans try to obtain information they seek. Information Foraging Theory (IFT) is one such theory. In this article, we present a series of studies 1 using IFT: the first investigates how expert explainers supply explanations in the RTS domain, the second investigates what explanations domain experts demand from agents in the RTS domain, and the last focuses on how both populations try to explain a state-of-the-art AI. Our results show that RTS environments like StarCraft offer so many options that change so rapidly, foraging tends to be very costly. Ways foragers attempted to manage such costs included “satisficing” approaches to reduce their cognitive load, such as focusing more on What information than on Why information, strategic use of language to communicate a lot of nuanced information in a few words, and optimizing their environment when possible to make their most valuable information patches readily available. Further, when a real AI entered the picture, even very experienced domain experts had difficulty understanding and judging some of the AI’s unconventional behaviors. Finally, our results reveal ways Information Foraging Theory can inform future XAI interactive explanation environments, and also how XAI can inform IFT.
Software producers are now recognizing the importance of improving their products’ suitability for diverse populations, but little attention has been given to measurements to shed light on products’ suitability to individuals below the median s ocio e conomic s tatus (SES)—who, by definition, make up half the population. To enable software practitioners to attend to both lower- and higher-SES individuals, this paper provides two new surveys that together can facilitate measuring how well a software product serves socioeconomically diverse populations. The first survey (SES-Subjective) is who-oriented: it measures who their potential or current users are in terms of their subjective SES (perceptions of their SES). The second survey (SES-Facets) is why-oriented: it collects individuals’ values for an evidence-based set of facet values (individual traits) that (1) statistically differ by SES and (2) affect how an individual works and problem-solves with software products. The surveys’ design goal is worldwide applicability, but as a first step, here we empirically validated both these surveys with deployments at University A and University B (464 and 522 responses, respectively), which showed reliability of both the surveys in a US context. Our results also statistically agree with both ground truth data on respondents’ socioeconomic statuses and with predictions from foundational literature. Finally, we explain how the pair of surveys can be uniquely actionable by software practitioners, such as in requirements gathering, debugging, quality assurance activities, maintenance activities, and fulfilling legal reporting requirements such as those being drafted by various governments for AI-powered software.