top of page

The AI Ethics Principle of Autonomy in Health Recommender System


Journal: Argumenta

Author: Dr. Simona Tiribelli, ITGH Director for AI Ethics


Abstract:


The application of health recommender systems (HRSs) in the mobile-health (m-health) industry, especially for healthy active aging, has grown exponentially over the past decade. However, no research has been conducted on the ethical implications of HRSs and the ethical principles for their design. This paper aims to fill this gap and claims that an ethically informed re-definition of the AI ethics principle of autonomy is needed to design HRSs that adequately operationalize (that is, respect and promote) individuals’ autonomy over ageing. To achieve this goal, after having clarified the state-of-the-art on HRSs, I present the reasons underlying the need to focus on autonomy as a prominent ethical issue and principle for the design of HRSs. Then, I pursue an inquiry on autonomy in HRSs and show that HRSs can both promote individuals’ autonomy and undermine it, also leading to phenomena of passive ageing. In particular, I claim that this is also due to the concept of autonomy underlying the debate on HRSs-based m-health, which is sometimes misleading, as it mainly coincides with informational self-determination. Using ethical reasoning, I shed light on a more complex account of autonomy and I redefine the AI ethics principle of autonomy accordingly. I show that autonomy and informational self-determination do not overlap. I also show that autonomy encompasses also a socio-relational dimension and that it requires both authenticity conditions and social recognition conditions. Finally, I analyze the implications of my ethical redefinition of autonomy for the design of autonomy-enabling HRSs for healthy active ageing.


The application of artificial intelligence (AI), and specifically machine-learning (ML) algorithms, in the field of healthcare has expanded significantly in the last decade (Jian et al. 2017, Esteva et al. 2019, Tran et al. 2019). ML algorithms—probabilistic models capable of mining and learning from vast amounts of raw, unstructured, and heterogeneous datasets to discover valuable patterns and correlations, often invisible to the human eye, and make predictions on the basis of them—have shown great potential for preventive and personalized medicine and healthcare (Miotto et al. 2016, Harerimana et al. 2018, Cowie et al. 2018, Barton et al. 2019, Dudley et al. 2015), for home-care health monitoring (Zoppo et al. 2020, Zheng et al. 2020), and e-health services via mobile-health (m-health) such as telehealth and telemedicine (Rubeis et al. 2018, Karako et al. 2020, Santoro 2020); domains recognized as crucial for the design of agetech, i.e., digital systems (from apps and wearables devices to domotics for ambient assisted living) that aim to improve autonomy for


Comments


bottom of page