Human-centered XAI (HCXAI) advocates that algorithmic transparency alone is not sufficient for making AI explainable. Explainability of AI is more than just “opening” the black box — who opens it matters just as much, if not more, as the ways of opening it. In the era of Large Language Models (LLMs), is “opening the black box” still a realistic goal for XAI? In this fourth CHI workshop on Human-centered XAI (HCXAI), we build on the maturation through the previous three installments to craft the coming-of-age story of HCXAI in the era of Large Language Models (LLMs). We aim towards actionable interventions that recognize both affordances and pitfalls of XAI. The goal of the fourth installment is to question how XAI assumptions fare in the era of LLMs and examine how human-centered perspectives can be operationalized at the conceptual, methodological, and technical levels. Encouraging holistic (historical, sociological, and technical) approaches, we emphasize “operationalizing.” We seek actionable analysis frameworks, concrete design guidelines, transferable evaluation methods, and principles for accountability.

Explainability is an essential pillar of Responsible AI that calls for equitable and ethical Human-AI interaction. Explanations are essential to hold AI systems and their producers accountable, and can serve as a means to ensure humans’ right to understand and contest AI decisions. Human-centered XAI (HCXAI) argues that there is more to making AI explainable than algorithmic transparency. Explainability of AI is more than just “opening” the black box — who opens it matters just as much, if not more, as the ways of opening it. In this third CHI workshop on Human-centered XAI (HCXAI), we build on the maturation through the first two installments to craft the coming-of-age story of HCXAI, which embodies a deeper discourse around operationalizing human-centered perspectives in XAI. We aim towards actionable interventions that recognize both affordances and potential pitfalls of XAI. The goal of the third installment is to go beyond the black box and examine how human-centered perspectives in XAI can be operationalized at the conceptual, methodological, and technical levels. Encouraging holistic (historical, sociological, and technical) approaches, we emphasize “operationalizing.” Within our research agenda for XAI, we seek actionable analysis frameworks, concrete design guidelines, transferable evaluation methods, and principles for accountability.

Currently, a significant gap exists between academic and industrial research in automated driving development. Despite this, there is common sense that cooperative control approaches in automated vehicles will surpass the previously favored takeover paradigm in most driving situations due to enhanced driving performance and user experience. Yet, the application of these concepts in real driving situations remains unclear, and a holistic approach to driving cooperation is missing. Existing research has primarily focused on testing specific interaction scenarios and implementations. To address this gap and offer a contemporary perspective on designing human–vehicle cooperation in automated driving, we have developed a three-part taxonomy with the help of an extensive literature review. The taxonomy broadens the notion of driving cooperation towards a holistic and application-oriented view by encompassing (1) the “Cooperation Use Case”, (2) the “Cooperation Frame”, and (3) the “Human–Machine Interface”. We validate the taxonomy by categorizing related literature and providing a detailed analysis of an exemplar paper. The proposed taxonomy offers designers and researchers a concise overview of the current state of driver cooperation and insights for future work. Further, the taxonomy can guide automotive HMI designers in ideation, communication, comparison, and reflection of cooperative driving interfaces.

Research on automotive user interfaces has highlighted the importance of transparent and explainable user interfaces, especially in safety-critical situations where automation failures can lead to (fatal) accidents. In this work, we investigated the use of augmented reality (AR) windshield displays as a means to improve explainability in safety-critical driving situations. To this end, we conducted a mixed-methods study (VR-based driving simulation combined with a semi-structured interview) with N = 24 participants and evaluated two concepts, containing (1) first-person information about vehicle behavior and (2) additional information about other traffic objects. Results showed that confidence and situational awareness were significantly higher when additional information was presented. The highest scores were achieved when participants were shown "how & why" information.

Cooperation between drivers and automated vehicles requires transparent communication of the automation’s current status. This can be achieved by communicating confidence or certainty in its current perception or decision. We evaluate different sensory modalities for communicating information about how safely an automated driving system can perform the driving task in critical traffic situations to a driver who is present as a cooperator. We aimed to improve communication between the driving system and the human driver, ultimately increasing the overall driving experience and performance. In a virtual reality driving simulation study with 34 participants, we presented confidence information across three modalities: visual, auditory, and vibrotactile, compared to a baseline condition. Our results indicate that communicating automation uncertainty through the auditory and vibrotactile modalities improved user experience, trust in automation, and perceived safety. At the same time, interactions with the Non-Driving Related Task were reduced by communicating confidence information in critical driving situations.