The AI-DEC: A Card-based Design Method for User-centered AI Explanations

TL;DR

The paper tackles the lack of user-centered AI explanations by introducing AI-DEC, a card-based design method that structures explanations along four dimensions: content, modality, frequency, and direction. It validates AI-DEC through co-design sessions with 16 workers across healthcare, finance, and management, demonstrating that tailored explanations can accommodate varying performance and autonomy needs depending on the AI system’s workplace role. The findings reveal six design themes—adaptation, task-support, collaboration, acceptance, accessibility, and communication—showing how end-users can actively shape explanations to fit their tasks and contexts. The work offers practical guidance for deploying context-adaptive, user-centered AI explanations and suggests AI-DEC as a bridge between end-users and AI designers/engineers, with potential applications beyond workplaces to domain-specific intelligent systems.

Abstract

Increasing evidence suggests that many deployed AI systems do not sufficiently support end-user interaction and information needs. Engaging end-users in the design of these systems can reveal user needs and expectations, yet effective ways of engaging end-users in the AI explanation design remain under-explored. To address this gap, we developed a design method, called AI-DEC, that defines four dimensions of AI explanations that are critical for the integration of AI systems -- communication content, modality, frequency, and direction -- and offers design examples for end-users to design AI explanations that meet their needs. We evaluated this method through co-design sessions with workers in healthcare, finance, and management industries who regularly use AI systems in their daily work. Findings indicate that the AI-DEC effectively supported workers in designing explanations that accommodated diverse levels of performance and autonomy needs, which varied depending on the AI system's workplace role and worker values. We discuss the implications of using the AI-DEC for the user-centered design of AI explanations in real-world systems.

Figures (6)

• Figure 1: Example of AI-DEC Design Card ---Left: The figure depicts a sample design card in the AI-DEC. The card is color-coded to represent its category within the communication dimensions. It includes a comprehensive description of the design element, a visual example, and examples of other use cases. Right: The complete AI-DEC set is displayed. For further details, please see Figure \ref{['fig:cardCollection']} in the Appendix.
• Figure 2: Findings Summary --- The graph shows different goals of the explanations that workers designed using the AI-DEC. The goals varied in the level of support that workers desired for their autonomy and performance across work domains. Healthcare workers focused explanation designs on enhancing their performance. The finance workers aimed to design explanations that improved both their performance and autonomy. In management, workers prioritized explanations that supported their autonomy.
• Figure 3: Worker's AI Explanation Design Using the AI-DEC --- The figure shows an AI explanation crafted by a worker (H6). The worker developed two distinct types of explanations for the worker's varying levels of expertise. Top: for experts, the AI explanation functioned as an extra eye. Bottom: for trainees, it served as an educational tool, providing more detailed information. Details can be found in Section \ref{['sec:exptypes']}.
• Figure 4: Complete Set of AI-DEC --- The AI-DEC consists of four types of cards, content, modality, frequency, and direction. In each facet, there are cards with design elements derived from existing design research in ML, HCI, human-robot interaction (HRI), and XAI. When utilizing the AI-DEC, end-users will select cards from the four communication facets to assemble their preferred AI explanation. The resulting explanation design will consist of a combination of cards from each facet, forming a tailored design solution to address user information or interaction requirements. This design solution will entail the content of the explanation, the mode of delivery, the timing or frequency of presentation, and the required level of user interactivity.
• Figure 5: Full Set of Participant Design Solution --- The table presents the complete set of participants' explanation designs utilizing the AI-DEC during the co-design session.