Human autonomy in the age of artificial intelligence

Abstract

Current AI policy recommendations differ on what the risks to human autonomy are. To systematically address risks to autonomy, we need to confront the complexity of the concept itself and adapt governance solutions accordingly.It is hard to overstate the important role autonomy plays for our moral and political institutions. A cornerstone of human dignity and a prerequisite of liberal democracy, autonomy is often considered a fundamental human value [1-4]. Progress in artificial intelligence (AI) development opens up new opportunities for supporting and fostering autonomy, but it simultaneously poses significant risks. Recent incidents of AI-facilitated deception, manipulation, or coercion suggest that AI technologies could seriously interfere with human autonomy on a large scale. Cambridge Analytica's attempt to manipulate voters is just one example [5]. Facebook's "emotional contagion" experiment, in which users were swayed towards adopting certain emotional states, yet another one [6]. Consequently, human autonomy has become a central theme across guidelines and principles on the responsible development of AI. The European Commission's High-Level Expert Group (HLEG) lists 'respect for autonomy' as the first of its four key ethical principles in their Guidelines on Trustworthy AI [7]. Several other policy documents, including the Association for Computing Machinery's Code of Ethics [8], The Montreal Declaration for Responsible Development of Artificial Intelligence [9], and the European Commission's White Paper on Artificial Intelligence [10], equally emphasise the need to protect and respect autonomy whereas the Organisation for Economic Co-operation and Development (OECD) lists autonomy as one of their human-centered values [11]. Despite this frequent call for the protection of autonomy, there remains substantial ambiguity within these documents as to (a) what exactly is meant by the term 'autonomy', as well

Rethinking Health Recommender Systems for Active Aging: An Autonomy-Based Ethical Analysis

“…While the relevance of autonomy in the fields intersected by our inquiry is evident, some outline how such a core AI ethics principle still tends to be vague or undefined. In other words, it is not clear what the respect and promotion of autonomy truly demand for the ethical design of AI (Prunkl, 2022 ) and RS (Prunkl, 2022 ).…”

“…An example can be the unfair reduction of meaningful opportunities due to biased correlations and socio-relational exposure, possibly leading to alienation/isolation (Pariser, 2011 ; Knees et al, 2024 ; Valentine et al, 2023 ). Such risks can only be understood through a more complex multidimensional concept of autonomy (Prunkl, 2022 ; Tiribelli et al, 2023b ). This paper (i) investigates HRS and proposes (ii) an understanding of the AI ethics principle of autonomy to elicit and circumvent the threats affecting AA, and (iii) possible mitigation strategies.…”

Artificial Intelligence and Groups: Effects of Attitudes and Discretion on Collaboration

“…AI may (or at least may be perceived to) remove discretion from operating employees or managers (Prunkl, 2022; Raisch & Krakowski, 2021), likely impacting the perceived autonomy associated with the job. Prior research suggests that the level of autonomy and its relationship with technology can have serious implications for team collaboration and other attributes of work (Parker & Grote, 2022).…”

Artificial intelligence and aging: potential and precautions

“…Moreover, the role of patient subgroups, and whether publication bias and other biases prevail cannot be determined or currently accounted for or validated to any meaningful degree. 33 The system may also fail to address the idea of natural healing or include data that are not considered medically mainstream, for example, herbal or alternative medicine approaches and the various interpretations of disease in these and other models.…”