The question of the effectiveness and limitations of a diagnostic AI arises, from a technical, but also ethical and legal point of view. On February 27, the Aristotle Association is organizing a seminar on the relationship between AI and health. In what way is this sector deeply shaken by the arrival of these new technologies? And why is this new approach to health attracting so many players, from Gafam to SMEs? What are the consequences, both economic and civilizational? An interview conducted by the Aristotle Association prior to this conference, with the two seminar organisers Judith Nicogossian, Anthropologist, and Christophe Denis, researcher at the Lip6 of the Sorbonne, demonstrates the issues and challenges of AI in the world of health.
In April 2018, the FDA (Food and Drug Administration) kicks off the international competitiveness of the use of diagnostic AI in health care by authorizing an AI to make a diagnosis in the field of diabetic retinopathy, in the absence of a human expert. This decision raises many questions about the potential of the tool, and in particular, the prospect of having to rethink health training by integrating the autonomous technological third party into the doctor-patient relationship, and in the light of knowledge about the human factor.
Similarly, the question of the effectiveness and limitations of a diagnostic AI arises, from a technical, but also ethical and legal point of view, in particular the strategic question of the sovereignty of health data.
Finally, the question arises of a diagnostic health RN with civilizational scope. Is it a disruptive innovation or a (simple) incremental perspective? What is the posture to adopt, between fantasy and interest? For finally, should this diagnostic AI, at best, not integrate the values and real needs of patients into the programming, in respect of a therapeutic alliance necessary for the care relationship?
This isn't the first time Aristotle's taken an interest in artificial intelligence?
Christophe Denis : Yes, this is a seminar topic that we have been having regularly for the last two years. The first seminar consisted mainly of a global overview. We had chosen to have a general point of view on applications around AI. Is it a buzzword or not? Then we focused in a second seminar on the current challenges: how to make sense of the results produced by artificial intelligence. We then tackled in detail the notion of explicability. The goal was to take a look into the black boxes, whether through academic research or with startups like Craft IA. After these two seminars, and after many exchanges, we decided to make a more particular point on the approach of AI in health.
Why not enjoy unlimited reading of UP'? Subscribe from €1.90 per week.
Does digital technology change a lot in the health field?
Judith Nicogossian Yes, and we have addressed this topic in previous seminars. For example, we have talked about the contribution of digital technology to health in the context of the paradigm shift that is taking place in this field, a 4P (predictive, preventive, personalized and participatory) medicine.
At the moment, our relationship to digital technology is redefining the contours of health and illness, the care pathway is becoming 2.0, and the doctor-patient relationship is changing.
We also note the emergence of exponential medicine, interpreted as the application of transhumanist theories to health. All this had to be cleared up. In particular, we focused on the contribution of intelligent systems (AI) techniques, the questions surrounding their programming and their ethical and legal impacts.
What questions does this raise about quality of care, responsibility for diagnosis, treatment, trust, for example. We had to question the explicability of the algorithms. In health more than elsewhere, we need to know what we do with black boxes. To what extent can we accept them?
And in this seminar, what are you going to focus on?
JN : We will work around the issue of diagnostic expertise and trust, and thus the issue of medical training and the "human factor", the unbreakable part of human behaviour in human-machine interaction.
Since April 2018, the FDA (Food and Drug Administration) has authorized an AI to make a diagnosis in the field of diabetic retinopathy, in the absence of a human expert. So we will question what can be achieved today in technical and technological terms thanks to these diagnoses.
For medicine is now spread over very different fields. And the contribution of these tools is not the same between these medical fields. For example, AI has more input in radiology than in general medicine. Surgery is also a special case. So we have to think about all these technological potentialities by field of expertise, both medical and transversal to other disciplines.
On this subject we have invited speakers such as Guy Vallancien (urologist) Arthur André (neurosurgeon), Antoine Tesnières (I-Lumens and vice-president of the universities of Paris)
Pierre-Antoine Gourraud (hospital practitioner), Nicolas Pivert (Medtronic), Bénédicte Poumarède (Doseo), Nicolas Pons (INRA), Bruno Patin (Dassault) or Denis Caromel (ActiveEon) who ask themselves this question from the point of view of their reciprocal specialties.
CD In a second step, we will discuss the limitations of these intelligent tools. If the legal, ethical or philosophical discourse has already been dealt with, there will be different contributions here, such as the sovereignty of the data covered by the jurists Bénédicte Bévière-Boyer or Alain Legendre.
JN : It is still difficult to marginalize oneself from the Gafam monopoly on data, which is recovered today in a global way. We have launched the Health Data Hub, a platform for exploiting health data, managed by the Ministry of Solidarity and Health but which is ultimately a pale copy of Google's concerns at the moment. We will therefore seek to further question the sovereignty of data thanks to the specialist insights of Alexandre Templier (QUINTEN).
To fight against disinformation and to favour analyses that decipher the news, join the circle of UP' subscribers.
Finally, on the civilizational level, we will enter more concretely on the positions to be taken towards AI with experiential testimonies, such as that of Loïc Etienne, an emergency doctor (Medvir), who will practically talk about his experiences.
The goal is to get out of fuzzy logic to stick as closely as possible to the needs of reality and to be able to work on linguistic programming, the natural treatment of languages (we will have the testimony of the expert Frédéric Landragin on this subject). Without, however, refraining from projecting oneself and going ahead of the prospective.
We will be fortunate to have the contribution of the publishing house La Volte, with an editorial director Stuart Pluen Calvo and a science-fiction author, Raphaël Granier de Cassagnac, who has the double skill of also being a physicist.
A very interdisciplinary vision, so...
JN Yes. The field of health is a concrete transformation with an impact on the human body, in society, it is a question of making sense of these digital technologies applied to medicine. It is a niche for industrialists who are seeking to develop new tools and are in fact themselves becoming the relay for this new field of knowledge, the impact of automated systems on humans, our species.
Faced with these technological and economic changes, we are determined to change our approach and create interdisciplinary meetings.
What are the foundations of the new medical paradigm?
JN The first driver of these changes is an imperative: the reduction of healthcare costs.
In the face of the population explosion, health care costs must be reduced. So rely on the economics of these new technologies to continue to improve and guarantee human health while minimizing the money spent.
At the same time, we are moving away from the mechanistic, functionalist vision of the human body, which is ultimately very Cartesian. Western medicine is taking a new look at the body, with a more holistic approach, and leaving more room for traditional medicine.
Will technology be able to live up to this hope? What are the limits of the healthy human being?
So the health field is very special?
CD Health is a typical sector because we are witnessing a positioning of different players around the Gafams who are investing massively. This raises the question of cybersecurity, but also security in general. The automation of processes has a retroactive effect on people, and these effects must be questioned: loss of vigilance, confinement. And the questions we ask ourselves about health are applicable to other areas.
JN: Conversely, the arrival of RNs also calls for medical practice. Because this technical automation did not wait for the current AI to start up. We are not moving towards disruptive, but rather towards incremental. On the other hand, this transformation makes us reflect on the real needs of humans. On drug validation, for example, with more rigorous methods for measuring clinical effects.
The tool can then become very convincing. Whether in the health field or elsewhere, AI encourages us to change our behaviours and play with our emotions. Our human reactions evolve and above all co-evolve with those of machines. We are not talking about replacement but about complementarity.
(Source: Aristotle, February 2020)
Seminar "An AI check-up for digital health, it's serious doctor(s)? by the Association ARISTOTE on February 27th, from 8h30 to 18h - Ecole Polytechnique, Route de Saclay - 91128 Palaiseau