Introduction

The need for reflection on the rationale and purposes of law in the field of artificial intelligence and health

Definitions and experimentation

Artificial intelligence (AI) can be defined as a “theoretical and practical interdisciplinary field aiming to understand cognition and reflection mechanisms, as well as their imitation by a hardware and software device, for the purpose of assisting or substituting human activities”1. For ease of reference, the expression “algorithmic medicine” will designate the use of tools and systems in the medicine field, based on algorithms associated with computer processing (Lambert 2019). The notion of health has been described by the World Health Organization as “a state of complete physical, mental and social well-being, and not merely the absence of disease or infirmity”. The present contribution strictly focuses on medical science, considering the notion of “well-being” to be rather ambivalent and inaccurate2.

Each in their own way, medical acts such as prevention, diagnosis and care, as well as prognosis, invite contributions from “medical AI”. A few examples illustrate the diversity of situations involved (for an understanding of the use of AI in the field of health, through experiments and scientific collaborations, refer to Chassagne et al. (2021)). First of all, in terms of medical imaging, experiential feedback is “promising”, not only due to the increase in data resulting from the devices’ technical progress, but also due to the application of a global data management standard (Chassagne et al. 2021, No. 50). Research on the pre-interpretation of mammograms is being explored in relation to AI, its purpose not being to replace the radiologist, but to facilitate diagnosis by identifying the points of attention to be prioritized and assessed by the imaging specialist (Chassagne et al. 2021, No. 62 sqq.). Furthermore, AI is being used in certain innovative medical devices for the diagnosis of sleep pathologies (e.g. the detection of sleep apnea at home. See Chassagne et al. (2021, No. 57 sqq.)). More generally, the refinement of medical and surgical techniques, as well as the improvement of patient care and the strengthening of remote care, are among the objectives of algorithmic medicine3.

The use of the “goal law”

All decision support systems and medical devices incorporating AI require the most accurate and reliable algorithms (for a technical and legal analysis of the challenges raised by decision support systems, see Desmoulins-Canselier and Le Métayer (2020)). In terms of data collection, the purpose is to acquire the most diversified data. Data also need to be cleaned, structured, formatted and sometimes enriched. In this context of innovation and research requiring significant investment, GAFAM and BATX4 appear as the health data “big players” for the first time. The public authorities have not remained indifferent to this factor. They had recourse to the law and, in a privileged way, to the law understood as an instrument at the service of certain ends. This was referred to as the “goal law” (expression coined by Paul Cuche (1919) and quoted by Oppetit (1999)):

To speak of the goal and final cause law, whether we like it or not, is to announce a metaphysical problem. […] The legislator who develops a legal rule – a goal law, as we have said – must have taken sides on the value of the goal pursued […] the legislator must have developed a certain conception of social progress; otherwise, how could he/she enforce a legal constraint, encouraging some directions on human activity while prohibiting others? (Cuche 1919)

Reflection on “the goal law”, its purposes and rationale

The equilibrium of a “goal law” can be fragile, either because the goals pursued by the legislator are likely to elicit conflict, or because they have been insufficiently considered upstream. Where appropriate, it is necessary to investigate the ethical, anthropological, sociological or practical grounds underlying the law, beyond legal reasoning. Thanks to this approach, it is possible to gain further insight into the values defended by the law, their content, hierarchy and their nature:

Issuing a goal law obviously implies that one envisions a certain hierarchy in the goals to be reached! Issuing a goal law involves formulating a value judgment (at least implicitly) on the chosen and imposed goals, as well as on disdained and forbidden goals (Cuche 1919).

This approach is particularly important when “algorithmic medicine” is involved. In fact, the use of algorithms and their regulation primarily concerns a human being, that is, the patient. However, it also challenges the very definition of medicine and has implications for another human being: the doctor. Under these circumstances, it is easy to understand the need for an “anthropological coherence imperative”. “Act in such a way that your action does not destroy your human nature!” (Lambert 2019).

Differentiated approaches to “algorithmic medicine” in “goal laws”

So far, “algorithmic medicine” has been approached by the legislator through the prism of health data. While the main texts relating to the protection of personal data apply to the medical field (French Law No. 78-17 of January 6, 1978 (LIL); regulation (EU) 2016/679 known as RGPD), the French Law No. 2019-774 of July 24, 2019 aimed to enhance the value of health data and, ultimately, the deployment of an AI medical5 sector. It is precisely this “goal law” that is relevant to scrutinize the so-called first objective approach to algorithmic medicine. The data’s prism leads to an objectified representation of the patient’s health. The human being, in its integral dimension (physical, psychological, spiritual), disappears behind objective knowledge (patient’s health data). This way of apprehending medical AI through data must be differentiated from another subjective interpretation.

Algorithmic medicine can be inscribed within the doctor–patient interpersonal relationship. This analysis is based on a certain conception of the person: “The human being is not just a lump of clay; it is a living being, in relation with others” (Aynès et al. 2022, footnote No. 1). In line with this doctrine, civil law strives to characterize the trust relationship between doctor and patient, in order to determine their reciprocal rights and obligations. Recently, health professionals have been held responsible for informing the patient when a medical device relying on algorithmic data processing is used (Article L. 4001-3 CSP). It is on the grounds of this innovation – introduced by Article 17 of the French Law No. 2021-1017 of August 2, 2021 – that the second approach to the so-called subjective algorithmic medicine should be studied.

When questioning the purposes and rationale of the law in the field of algorithmic medicine, two approaches need to be differentiated: an objective approach, based on health data, and a subjective one, at the heart of the doctor–patient relationship.

An approach to algorithmic medicine based on health data

The purposes of the goal law

French Law No. 2019-774 of July 24, 2019, on the organization and transformation of the health system (OTSS), was set within a global strategy for AI6. Two specific goals were of particular interest in this sense.

First and foremost, health data were considered as data to be valorized. Following the submission of the Villani report, the President of the Republic announced the creation of a Health Data Hub as one of the strengths of the French Artificial Intelligence (AI) strategy (Villani 2018). In June 2018, the French Minister of Solidarity and Health launched a prefiguration mission on this platform for the exploitation of health data. A public interest group, called the Health Data Platform (Plateforme des données de santé), was created via the French Law of July 24, 20197. During this time, it was argued that the implementation of the platform, as well as the enrichment of the National Health Data System (SNDS, Système national des données de santé), were intended to promote new uses of health data, in particular those related to the development of AI methods:

In order to increase the use of data from the National Health Data System both in clinical research and in terms of new uses, particularly those related to the development of artificial intelligence methods, the National Health Data System will be enriched with all the data collected during the procedures covered by the health insurance. This enrichment will position France among the leading countries in terms of structuring health data, while maintaining a high level of privacy protection8.

Structuring an Artificial Intelligence and Health sector was one of the ambitions of the report from the Health Data Hub prefiguration mission:

The analysts’ opinions have converged on a double-digit growth of the market for AI-based products and services until at least 2025 and a market size exceeding...