Let us not kid ourselves, sooner or later, the power of artificial intelligence (AI) will penetrate deeply into the functioning of public institutions. If only because they are penniless, they struggle to take on increasingly numerous and demanding missions and AI will often be seen as a pragmatic solution.
Everything also indicates that AI will not appear for long as a simple tool “improving” the functioning of these services , as the typewriter was able to do during the time of Courteline’s administration. We can even envisage that within a few years, the consequences of AI will lead us to profoundly evolve our institutions, or even to overhaul our constitutional law.
As of today, we know that the structural lack of explainability of algorithms is a challenge for public action: why am I not entitled to Benefits when my neighbour has obtained them? Why didn’t my son get into the university he was aiming for? This is just a taste of the questions that will arise with AI.
A first response to avoid a brutalising technocracy will consist of establishing strong transparency of software algorithms and services, institutionalized through citizen participation. This would apply not only to the State, but also to large digital platforms. This force will have the capacity to ensure that the biases of the models are identified, that the “dark patterns” called to steal our attention are neutralised.
We could even envisage that this citizen force could co-develop certain public services. Nothing new in reality. It is a question of transposing the citizen reserve that the Swiss have established with regard to the defense of their territory to the field of algorithms. I know how theoretical, even utopian, this statement may seem, but I do not see in the medium term any other mechanism capable of replacing it satisfactorily. To do this, it will obviously be necessary to rethink our Constitution, to the extent that a “soft” law can be enacted by citizens, in a flexible and quantitative manner.
Concretely, what does this mean? For example, there is little doubt that legal decisions will ultimately be made by AI . First minor disputes, then little by little judgments with more serious consequences. Underlying this claim is the fact that there is a very large body of legal documents, providing a great opportunity for teaching strong AIs in law. The same goes for police investigation processes, rights allocation commissions, management of building permits, etc.
In health, various research studies tirelessly remind us that at least 30% of practitioners’ time is occupied by administrative processing. Not only is this a net destruction of productivity, but it is also a significant source of medical errors. In the long term, in education, this can be a way to provide individualised mentoring.
Of course, some could easily denounce these remarks as being excessively techno-solutionist, not taking into account the subtlety of the judge, the administrative agent, etc. We can easily answer them that it is probably not them who are the victims of the denial of rights, of the non-access to aid to which some are legitimately entitled, of the Kafkaesque complexity of this bureaucracy that Bourdieu denounced as a primary factor of social inequality.
Others will mock science fiction scenarios and assert that their reality is not for tomorrow. Everything suggests that they are largely wrong: a modern nation should, on the contrary, start now to think about the nature of its institutions in the age of AI, especially when this nation is burdened by complexity and the silos of its decision-making.