Artificial intelligence and healthcare

On 12 October 2023, the Garante per la Protezione dei Dati published a document entitled ‘Decalogo per la realizzazione di servizi sanitari nazionali attraverso sistemi di Intelligenza Artificiale’. This is an interesting document, which deserves reflection and further study.

The Garante’s document (number 9938038), can be downloaded by clicking here, and contains a number of aspects related to the implementation of artificial intelligence algorithms in healthcare. The ten rules that the Garante puts on paper are:

1. The legal bases of processing.
2. The principles of accountability and privacy by design and by default.
3. Roles.
4. The principles of knowability, non-exclusivity and algorithmic non-discrimination.
5. Data Protection Impact Assessment (DPA).
6. Data quality.
7. Integrity and confidentiality.
8. Fairness and transparency.
9. Human supervision.
10. Further profiles with respect to data protection regulations related to dignity and personal identity.

Each of these points has been described, detailed and examined by the Authority. This article will comment on some of them that are considered more complex than others although they all have particular relevance.

On algorithmic discrimination

The Garante addresses this aspect by hinging it on the three key principles set out by the Council of State:

1. the principle of knowability, according to which the data subject has the right to know about the existence of decision-making processes based on automated processing and, if so, to receive meaningful information about the logic used, so that he/she can understand it;
2. the principle of non-exclusivity of the algorithmic decision, according to which there must still be in the decision-making process a human intervention capable of checking, validating or refuting the automatic decision (so-called human in the loop);
3. the principle of algorithmic non-discrimination, according to which the data controller should use reliable AI systems that reduce opacity, errors due to technological and/or human causes, periodically verifying their effectiveness also in the light of the rapid evolution of the technologies used, of the mathematical or statistical procedures appropriate for profiling, putting in place appropriate technical and organisational measures. This is also in order to ensure that factors leading to inaccurate data are rectified and the risk of errors is minimised, given the potential discriminatory effects that inaccurate processing of data on health status may have on individuals (see recital 71 of the Regulation)

Algorithmic discrimination is a very real risk, mostly caused by potentially culpable reasons in the artificial intelligence training process. The risk can be mitigated by adopting known and shared training protocols that pursue equally known and shared goals. If a common policy in the management of training processes were not adopted, there would be a risk of unexpected, uncontrolled and hardly manageable results.

On human control

Human control is not debatable: the need to keep the process in an anthropocentric dimension is essential. The physician, at any point in the algorithmic process, must have the faculty to detach himself from it and supervise it in accordance with the principle of knowability but above all intelligibility. If the process that the physician has to verify were incomprehensible, the power to supervise it would be in vain. This is especially so when the doctor disagrees with the algorithm and, for the benefit of the patient, decides to question himself trying to understand how the artificial intelligence arrived at the different result. Human control must not, therefore, be alienated at any point in the use of technology.

Indeed, one has to get away from the concept of trust. Trust requires reliance on something or someone because one does not know the context completely. One must enter into a realm of knowledge, of certain knowledge, which therefore does not require reliance on anything.

A young society

Italy is a country that is facing Europe’s largest ‘brain drain’: there are highly qualified professionals fleeing to seek their fortune in other countries. The causes are well known and are many: salaries are too low, career expectations too slow, a suffocating regulatory framework.

Of course, acting as the immovable guardian of absolute continuity, there is bureaucracy, which is not only a lot of bureaucracy and boring (bureaucracy is a lot of bureaucracy and boring all over the world, but it is important, it is needed), but it is above all irrational, incomprehensible, inefficient, grotesque, always ready to surprise us with renewed vigour

Source: The Post “The Stasi and the End of Imagination”, 15 May 2023, Link

In the last few days, Italy has appointed a new commission: the Algorithms Commission, which will obviously have to deal with artificial intelligence. This is an important step, even though insiders will recall the large number of task forces, working groups, and technical tables that have overlapped over the years and which raise the question of the need for yet another commission. However, this government initiative is at the centre of a different controversy: the chairman of the commission will be Giuliano amato, 85 years old, born in 1938.

With all due respect to Mr Amato and his professionalism, the question arises as to whether there was not a younger and more professionally adherent candidate for the leadership role for a commission that will have to radically innovate the country. This is not only a matter of internal representation but also external representation, notwithstanding the valuable experience of the Honourable Member.

In the 2019 Three-Year Plan, AgID assumed a central role in the field of artificial intelligence; insiders will recall that a special task force was appointed. Also in the light of these technical tables, these cabine di regia, which have characterised the actions of governments in recent years, one wonders whether the candidature of Hon. Amato was the only useful alternative. Time will assess, hopefully positively, the work in terms of effectiveness and relevant results achieved.

Il Piano Triennale dell’ICT

The role of Europe and Italy

Europe is the future. Europe can coordinate the knowledge of countries, it can establish common operational protocols, carry out controls on companies and technological solutions in compliance with the principles of knowability while maintaining industrial secrecy. But to achieve this, a change of mentality is needed: priority must be given to the cohesion side of resources, to the strategic level before the technological one. In a landscape where major corporate interests clash, where there is a race for the best performing ‘technological platform’, it becomes difficult but extremely necessary to understand that cohesion and integration are two strategically essential requirements for successful technological innovation.

In a country that often prefers to do, undo and redo what was done before, this paradigm shift appears as mammoth, if not alien. Instead, this is precisely what is needed: a common strategy, first of all European and then national. Can Italy take up the challenge in a strategically relevant way? Will it succeed in putting the strategic level before individual market interests? The current challenge will condition the future, like the thread held by the Moires in Strudwick’s painting. If we lose this challenge, Italy will lose the chance to bring back to the centre the values of quality production that have always made it a pole of excellence in the world.