No algorithm will ever be able to defend democracy.

This text is my contribution to the Information and Democracy Commission of Reporters Without Borders

Also available in French.


No algorithm ever will be able to defend democracy.

(But any of them operating within large-scale toxic technical architectures is capable of corrupting it.)

On September 4, 2018, Mark Zuckerberg published an article in the Washington Post entitled "Protecting democracy is an arms race. Here’s how Facebook can help". Ten days later, it is on his platform that he announces "preparing for elections": he lists the risks, biases, diversions, logic of influence and manipulation that have already endangered the democratic balance and the organization of elections. And he announces, once again, that he will try to correct all this. He will not succeed.

How did we find ourselves at the beginning of the 21st century in a situation where Mark Zuckerberg – and a few other digital platform owners – set themselves the recurring objective of "protecting democracy", mainly through "algorithms" and "artificial intelligence" and claim that this will be one of their main "missions" within their company and on a global scale?

In 2011, two theorists of artificial intelligence dealing with the ethical problems that algorithms would raise wrote that:

"Increasingly complex decision-making algorithms are both desirable and inevitable, as long as they remain transparent to inspection, predictable to those they govern, and robust against manipulation." (Bostrom and Yudowski, "The Ethics of Artificial Intelligence")

Today, "the algorithms" we are talking about are deployed within toxic technical architectures involving millions or billions of users.

Today, "the algorithms" we are talking about are based on proprietary and therefore totally opaque data sets.

Today, "the algorithms" we are talking about are explicitly developed to provide themselves with a level of autonomy (or "learning") that makes their "behaviour" and decisions often unpredictable for their creators themselves.

Today, "the algorithms" we are talking about are constantly interacting with other toxic algorithms, data sets and technical architectures ; and they are doing so on ever larger scales and in ever more constrained environments that further increase the level of risk and uncertainty.

That is why today, for all these reasons, it is absolutely impossible to guarantee that they are transparent to inspection, predictable for those they govern and, above all, robust against any manipulation.

For the past 10 years, the main struggle of activists, journalists and defenders of digital rights has been to limit the impact of the algorithmic fingerprint on our private and intimate lives. This battle is over, obsolete and, for the most part, lost. This is another battle we must fight today, on a completely different front, with a completely different urgency and on a completely different scale.

We have to limit the impact of the decision-making algorithmic footprint on our public life, on our common social infrastructures and on our collective destiny.

It is urgent and imperative that any form, ambition or project of algorithmic governance, as soon as it touches on sovereign sectors (transport, education, health, justice, security), be necessarily and by legislative constraint, developed on the model of the GNU GPL licenses of free software to guarantee a complete and sustainable auditability of the processes at work.

It is urgent and imperative that the development of a universal model for the portability of all our data** be a priority for governments, and that it be imposed on all actors in the economic world in relation to the conservation or storage of digital data, whatever its nature, volume and use.

(** the SOLID project led by Tim Berners Lee could be a first approach)

It is urgent and imperative that companies that currently receive the bulk of data and digital flows (roughly GAFAM, NATU and other BATX) be taxed at the real level of their business volume and that this tax directly finances the above-mentioned actions, as this redistribution process must remain beyond the control of these companies. Because the opposite process has already begun, the one in which a few omnipotent companies claim the right to challenge public power and the general interest in tax collection, as Amazon and Starbucks recently demonstrated in Seattle.

It is urgent and imperative that a common information system be positively defined in the law and that it can include the algorithms and code that can be mobilized in the context of any public action.

Finally, it is urgent, imperative and vital that everything directly related to the democratic process (such as voting, election, counting) be placed beyond the reach of any form of assistance, guidance or algorithmic substitution (starting with "voting machines"). Electronic voting' must be considered for what it is : a threat that is strictly and definitively incompatible with respect for the confidentiality of the vote and therefore democracy.

The question is whether we will be able in the very short term to build an alternative which, after the digital time of "disintermediation" of the last 20 years, will be that of forms of algorithmic remediation that respect the social body and start from its most fragile, poorest and most exposed part. Then perhaps, and only then, can questions of algorithmic governance begin to be considered calmly.

Aside from all these conditions, we are going to offer future generations a world in which the main problem will not be that Mark Zuckerberg and a few other industry bosses are prone to be the protective guardians of our democracies, but that they are indeed the only ones still in a position to be so, while having themselves only a very vague and approximate idea of how to proceed and the chances of achieving it.

This is not just about Fake News and free will. It is not simply a question of freedom of information or freedom of the press. It is not just algorithms, platforms, states and nations. It is not just a question of human intelligences and other "artificial" ones. It is about the freedom of peoples. It is about freedom at all.

(Translated with


Translated with

Laisser un commentaire

Votre adresse e-mail ne sera pas publiée. Les champs obligatoires sont indiqués avec *

Ce site utilise Akismet pour réduire les indésirables. En savoir plus sur comment les données de vos commentaires sont utilisées.

Articles similaires

Commencez à saisir votre recherche ci-dessus et pressez Entrée pour rechercher. ESC pour annuler.

Retour en haut