Large Language Models (LLMs) are increasingly used to generate clinical documentation, yet most systems focus on easing individual users’ clerical burdens rather than supporting multi-stakeholder work. We explore how LLMs expand the design space of clinical multi-stakeholder systems. Through a research-through-design study with 10 dermatologists using a teledermatology prototype, we explore how designing for secondary stakeholders opens new design opportunities and reshapes primary users’ work. Applying the distributed cognition lens to our findings, we conceptualised representational design, i.e., the process of encoding secondary stakeholder perspectives (representations) in LLMs through, among others, prompts, templates, and constraints, while attending to LLMs’ latent knowledge. We envision representational design as supplemental to data- and interaction-level approaches to designing multi-stakeholder artificial intelligence (AI) systems. We demonstrate how LLM-mediated representations cause misalignments between generated outputs and primary users’ notions of acceptability, and how attending to these misalignments sets requirements for human-AI interaction techniques.

Generative AI technologies are reshaping everyday environments by enabling multimodal interaction. As their ubiquity and agentic capacities grow, there is a pressing need to understand how these systems reshape human–computer interaction in relational, social, and systemic terms. We introduce a scenario-based design pack for investigating Human–GenAI relations. Grounded in assemblage theory and structured around a three-stage process—Prepare, Make, Reflect—the pack supports the prototyping, analysis, and critical reflection of emergent sociotechnical configurations. We evaluated the pack across three deployments: an ACM workshop (n=22), a multidisciplinary design session (n=20), and a university HCI class (n=260). Participants generated scenarios that surfaced relational issues of power, agency, visibility, and care. We contribute the design pack alongside an exploratory framework to advance relational enquiry into multimodal Human–GenAI relations, support more inclusive and socially responsive GenAI practices, and complement FATE approaches by grounding fairness, accountability, and transparency in lived, multimodal configurations.

Despite recent advancements, real-world use of Artificial Intelligence (AI) in radiology remains low, often due to the mismatch between AI offerings and the situated challenges faced by healthcare professionals. To bridge this gap, we conducted a field study at nine medical sites in Denmark and Kenya with two goals: (1) to understand the challenges faced by radiologists during chest X-ray practice and (2) to envision alternative AI futures that align with collaborative clinical work. This study uniquely grounds the AI design insights in the comprehensive characterisation of diagnostic work across multiple geographical and institutional contexts. Building on ideas articulated by interviewed radiologists (N = 18), we conceptualised five visions that transcend the traditional notions of AI support. These visions emphasise that the clinical usefulness of AI-based systems depends on their configurability and flexibility across three dimensions: type of clinical site, expertise of medical professionals, and situational and patient contexts. Addressing these dependencies requires expanding the clinical AI design space by envisioning futures rooted in the realities of practice rather than solely following the trajectory of AI development.

Artificial Intelligence (AI) repeatedly match or outperform radiologists in lab experiments. However, real-world implementations of radiological AI-based systems are found to provide little to no clinical value. This paper explores how to design AI for clinical usefulness in different contexts. We conducted 19 design sessions and design interventions with 13 radiologists from 7 clinical sites in Denmark and Kenya, based on three iterations of a functional AI-based prototype. Ten sociotechnical dependencies were identified as crucial for the design of AI in radiology. We conceptualised four technical dimensions that must be configured to the intended clinical context of use: AI functionality, AI medical focus, AI decision threshold, and AI Explainability. We present four design recommendations on how to address dependencies pertaining to the medical knowledge, clinic type, user expertise level, patient context, and user situation that condition the configuration of these technical dimensions.

Medical Imaging (MI) datasets are fundamental to artificial intelligence in healthcare. The accuracy, robustness, and fairness of diagnostic algorithms depend on the data (and its quality) used to train and evaluate the models. MI datasets used to be proprietary, but have become increasingly available to the public, including on community-contributed platforms (CCPs) like Kaggle or HuggingFace. While open data is important to enhance the redistribution of data's public value, we find that the current CCP governance model fails to uphold the quality needed and recommended practices for sharing, documenting, and evaluating datasets. In this paper, we conduct an analysis of publicly available machine learning datasets on CCPs, discussing datasets' context, and identifying limitations and gaps in the current CCP landscape. We highlight differences between MI and computer vision datasets, particularly in the potentially harmful downstream effects from poor adoption of recommended dataset management practices. We compare the analyzed datasets across several dimensions, including data sharing, data documentation, and maintenance. We find vague licenses, lack of persistent identifiers and storage, duplicates, and missing metadata, with differences between the platforms. Our research contributes to efforts in responsible data curation and AI algorithms for healthcare.

One of the core goals of responsible AI development is ensuring high-quality training datasets. Many researchers have pointed to the importance of the annotation step in the creation of high-quality data, but less attention has been paid to the work that enables data annotation. We define this work as the design of ground truth schema and explore the challenges involved in the creation of datasets in the medical domain even before any annotations are made. Based on extensive work in three health-tech organisations, we describe five external and internal factors that condition medical dataset creation processes. Three external factors include regulatory constraints, the context of creation and use, and commercial and operational pressures. These factors condition medical data collection and shape the ground truth schema design. Two internal factors include epistemic differences and limits of labelling. These directly shape the design of the ground truth schema. Discussions of what constitutes high-quality data need to pay attention to the factors that shape and constrain what is possible to be created, to ensure responsible AI design.

Artificial Intelligence (AI) in medical applications holds great promise. However, the use of Machine Learning-based (ML) systems in clinical practice is still minimal. It is uniquely difficult to introduce clinician-facing ML-based systems in practice, which has been recognised in HCI and related fields. Recent publications have begun to address the sociotechnical challenges of designing, developing, and successfully deploying clinician-facing ML-based systems. We conducted a qualitative systematic review and provided answers to the question: “How can HCI researchers and practitioners contribute to the successful realisation of ML in medical practice?” We reviewed 25 eligible papers that investigated the real-world clinical implications of concrete clinician-facing ML-based systems. The main contributions of this systematic review are: (1) an overview of the technical aspects of ML innovation and their consequences for HCI researchers and practitioners; (2) a description of the different roles that ML-based systems can take in clinical settings; (3) a conceptualisation of the main activities of medical ML innovation processes; (4) identification of five sociotechnical interdependencies that emerge from medical ML innovation; and (5) implications for HCI researchers and practitioners on how to mitigate the sociotechnical challenges of medical ML innovation.

Virtual Reality allows users to embody avatars that do not match their real bodies. Earlier work has selected changes to the avatar arbitrarily and it therefore remains unclear how to change avatars to improve users' performance. We propose a systematic approach for iteratively adapting the avatar to perform better for a given task based on users' performance. The approach is evaluated in a target selection task, where the forearms of the avatar are scaled to improve performance. A comparison between the optimised and real arm lengths shows a significant reduction in average tapping time by 18.7%, for forearms multiplied in length by 5.6. Additionally, with the adapted avatar, participants moved their real body and arms significantly less, and subjective measures show reduced physical demand and frustration. In a second study, we modify finger lengths for a linear tapping task to achieve a better performing avatar, which demonstrates the generalisability of the approach.