Can facial recognition really improve crowd management?

Three days after failing to maintain order around the Stade de France, the mayor of Nice Christian Astrosi has called for the use of facial recognition to facilitate the conduct of the main event. But what contribution can algorithms make to seemingly unorganized crowd management? One year from the Rugby World Cup, two years from the Paris Olympic Games, and the full work of European regulations for artificial intelligence, it is not trivial to question.

TechnoSolution or Technodiverition?

At first, there were questions about the organization of the Champions League final. Home Secretary Gerald Dermanin quickly noted that with 30,000 to 40,000 British supporters holding fake tickets, an estimate is so unlikely that the president of the Liverpool club is demanding an official apology. At Real Madrid, the same thing: we want a quick explanation for the May 28 chaos. Arthur Mesaud of the Digital Rights Defenders Association believes that talking about face recognition in such a context is not a way to take the debate to the details but rather to focus on the questions. Don’t give up.

Guillaume Gormand, author of a study on the effectiveness of the video surveillance system used by Gendarmerie, agrees: “The problem is not having a tool like the way it is used. Properly covered, it will withstand a great deal of adverse conditions. But if the company does not follow, no camera or algorithm will compensate for management errors.

Behind the algorithm, video surveillance

Similarly, Christian Astrosi’s proposal fits a particular context. The proliferation of surveillance devices in public spaces, firstly, is described by Guillaume Gormand as being consistent with a lobby of private actors. “After the 1995 legislation, we first saw the big municipalities equipping themselves, then the market stagnated. It all started again when Nicolas Sarkozy launched an ideological campaign of video surveillance during his campaign and then during his mandate. “In 2019, the Gadget des Commune had 11,470 cameras in France’s 47 largest cities, 2.4 times more than in 2013. These are just a few of the foundations,” the researcher continued: “Today, we add a lot of equipment. , Calculation software, algorithms, to run tests. Better a poor horse than no horse at all. A

However, the effectiveness of the two technologies – pure video surveillance, firstly, of the algorithms applied to it, secondly – is a matter of debate. In 2018, sociologist Laurent Muchielli reported that only 1 to 3% of crimes committed on public streets were solved for camera images. In 2020, court auditors noted that the amount spent on these devices continues to rise, while very few studies have proven their effectiveness. Guillaume Gormand, for his part, noted that video surveillance has no effect on crime or displacement. Furthermore, only 1.13% of the collected images were used in the survey. Similarly, he noted, video surveillance “can be effective in conducting large events because it allows authorities to project themselves on the ground”, to see first-hand what is happening there.

Control to avoid abuse?

From this perspective, video devices can be useful. However, the researcher further points out, “In sensitive events, field actors go back to what they can master best: radio, live, information gathered on the scene … new technology comes second. And then that algorithm doesn’t solve the question.” The use of stadium bans “identifying people and using them to” wake them up “as Christian Astrosi advises, fits the context of more or less regulated technical testing at sporting events. In January 2019, FC Metz supporters, for example, Surprised to learn through the press that they had been tested by the two-eye company’s algorithm.

To avoid this kind of problem, three senators submitted a report on biometric recognition in public space on May 10. In the thirty proposals, they recommend “passing an experimental law for more than three years to draw a red line to avoid surveillance society and to experiment with different uses,” summarizes co-correspondent Mark-Philippe Dowbres (LR). Within the proposed range: prohibit real-time facial recognition without the possibility of obtaining human consent, “except for access to dangerous sites such as nuclear sites or major events with the risk of overflow or terrorist attacks such as the Olympic Games”. Indeed, such a proposal would rekindle the controversy over the use of facial recognition in 2024, although the possibility was ruled out in October 2021 by an inter-ministerial representative for the Olympic Games.

Little music of surveillance technology

Above all, senators have called for clearing the way for controversy over the use of such algorithms. In it, they join researchers on the legal implications of the artificial intelligence chair at Grenobles-Alpes University, who is the author of a detailed mapping of the use of facial recognition in Europe. In short, these actors say that authenticating a user with their consent (for example, comparing a photo with their face at the entrance to a stadium that they allegedly took a few days ago from their smartphone) has far less impact. Crowds like the London police.

But for Arthur Mesaud of La Quadrechar du Net, “at a time when there is little ideological debate about queuing, in the interest of good organization, the government or the police are the most adept at pushing privacy-threatening technologies.” Adopting recognition tools may seem harmful, relatively neutral, but it makes the population accustomed to aggressive technology. “It’s almost unreasonable that Christian Astrosi is giving face recognition at the Stade de France. But every six months he offers this kind of thing.

In the background, the control of artificial intelligence

If Mark-Philippe Dowbras and his co-reporters Arnaud de Belenet (LREM) and Jerome Durein (socialist) called for a framework for biometric recognition to allow certain uses and prohibit others from La Quadrechar du Net, then we Firmly vs. In early June, the association even launched a campaign to file a joint complaint against “Technopolice” with CNIL, a term behind which it incorporates many technologies promoted by Christian Astrosi – video surveillance, facial recognition algorithms or even automated behavior detection. .

In it, the two perfectly illustrate a debate that could shake not only the preparations for the Olympic Games, but also the European Parliament, when it comes to controlling algorithmic surveillance technology: is it better to ban them altogether? Germany, in particular? Or rather accept specific applications subject to testing?

Leave a Comment