Human-Centered Al for Health, Autonomy and Wellbeing

The proposed profile focuses on human-centered AI for health, autonomy and wellbeing with a particular focus on the individual’s perspective. The novely of this profile is the grounding in a holistic view on the individual’s everyday situation where AI-based systems are embedded, to study how AI-based systems influence health, autonomy and wellbeing and how the individual may influence and collaborate with the AI-based systems as part of everyday living to improve health, autonomy and wellbeing.

The development and integration of artificial intelligence (AI) in society affects living conditions for individuals as well as the development of societies. This has lead to increased calls for human-centered AI, since the development will likely affect individuals’ privacy, integrity, autonomy and control, health and well-being, as well as how individuals think about and practice these values.

Health-related AI-based systems are becoming increasingly important as societies try to i) prevent illness, increase wellbeing and reduce morbidity and mortality, and ii) manage the care of different age groups with diverse needs, preferences and life situations. However, it is the individual who, to increasing extent, is expected to be in charge of their illnesses and treatments on a daily basis, and who interacts with these systems, sometimes inside but mostly outside of the health care setting. Moreover, health-promoting AI systems are embedded in and competing for user attention with other tasks (shopping, household chores, entertainments, work, school, social life, etc), often guided by other AI-based recommender systems.

In the present urge for new innovations that can boost the AI economy, and in the drive for developing AI based on large amount of data, the perspectives of the individual user and his/her various needs and values are often forgotten. Within the areas of health promotion, much effort is put on developing recommender systems, persuasive technologies, digital social companions and other health-promoting applications based on patterns of behavior without involving the target group in the development process. Some appear as human-like characters to promote a social relationship similar to relationships between humans, which is expected by designers to affect the human. This lack of involvement and transparency leads to decreased possibilities for the individual to evaluate the trustworthiness in the different options when searching for tools that can solve health-related concerns. As improving health requires that the individual changes behaviors in order to develop and maintain a healthy lifestyle, it is vital that the individual’s perspective is taken into consideration to furthering understanding and discussion from ethical and societal points of view. Rather than unilaterally working to increase trust in AI systems, we need to make it possible for both individuals and society at large to evaluate when and how to trust AI systems. For this, it is crucial to understand the needs and goals of the individuals using AI-based systems, how the technologies empower (or disempower) their users, how they are integrated (or not) into everyday activities, and what they thereby make possible (or impossible).

A way to address the above risks is to develop AI-based systems that are collaborative, working together with the individual and groups of individuals during daily activities over time. Sweden and other Scandinavian countries have a long tradition of applying participatory design methodologies in health care settings, which is a good starting point for developing such systems. However, current methodologies are not developed for systems that may change autonomously over time such as self-learning AI-based systems, for systems that allow coordination of initiative between the human and system, systems that need to autonomously adapt behaviors to not violate integrity, or systems that encompass adaptive, person-tailored behaviors promoting autonomy, self-efficacy and wellbeing.

Relevant areas of concern at the forefront of research include:

I) Autonomy: What aspects of self-determination are at risk when individuals interact with health-promoting AI-based systems that may require change of behaviour? How should these AI systems function in order that such risks be mitigated? How can autonomy be strengthened through the support from AI-based systems?

II) Behaviour change systems for improving health: Sense of control and positive experiences are key preconditions for promoting behavior change. How can a broad perspective on experiences, beyond usability engineering, be adopted in order to promote trust, understanding, transparency, motivation and control?

III) AI-based systems as social companions: Individuals create close relations with the health-promoting AI systems in their everyday lives. How are such relations culturally, socially and emotionally invested, and what does this mean for the empowerment or disempowerment of the individual?

IV) AI-based systems as collaborators: how does collaborating with AI-based systems affect the individual’s decision making, autonomy, self-determination and fundamental values?

V) Design methodology for adaptive, person-centered AI-based systems/companions: How can users take active part in shaping the AI systems’ behaviour? How do users make AI-based systems individually relevant and acceptable?