It’s a disappointment for Patrick Breye. The German MEP was denied his request for access to all documents relating to the artificial intelligence (AI)-based lie detector project: “iBorderCtr”. A decision issued on September 7, 2023 by the Court of Justice of the European Union that continues to maintain opacity around this controversial project, reports Usine Digitale.
The tool, intended for travelers wishing to enter the European Union (EU), should make it possible to detect whether or not they are lying during border controls at the airport. To do this, they must respond to an artificial “control agent” via a webcam, explains the European Commission.
This virtual border guard is based on a machine learning system capable of detecting suspicious facial “micro-expressions” when travelers give their answers. For example, an eye that goes to the left or right due to hesitation. If the machine considers the answers given to be true, the traveler crosses the border. Otherwise, a human agent must perform a more in-depth check.
Fear of discriminatory bias
However, the precise functioning of this technology remains a mystery. For this reason, MEP Patrick Breye wanted to consult the documents relating to the project. His first application was rejected by the European Executive Research Agency (REA), a funding body that manages research grants from the European Union (EU).
For this reason, Patrick Breye turned to the Court of Justice and justified his request in the name of “public interest” for a project, moreover, financed entirely with public funds, worth 4.5 million euros. The MEP is also concerned about possible discriminatory biases related to the use of this technology against certain populations. The significant use of biometric data in the project. iBorderCtr It is also a cause for concern for Patrick Brey.
The work on this AI-based lie detector is also denounced by Article 19. This British association for the protection of human rights reacted, on September 7, to the decision of the European Court of Justice.
Furthermore, the detection of “microexpressions” is not based on any scientific fact and this technology is not infallible, the association considers.
Criticized operation
“The link between microexpressions and lie detection continues to be highly criticized,” says Laurence Devillers, researcher at the Laboratory of Informatics of Mechanics and Engineering Sciences at the CNRS. Releasein 2019. “The detection of facial movements has limits, which may be due to morphology, skin color or the fact that the person is wrinkled or not, for example,” he later added.
Already in 2019, the specialized media The Intercept tested the system, then in an experimental phase, on the border between Serbia and Hungary. The technical limitations of this project were then pointed out, with a mediocre score despite the honesty of the journalist’s answers, as well as a total opacity about the result obtained, communicated only after a legal request under the GDPR.
This September 9, the Italian version of the Wired medium published an article in which it analyzed the information that, however, Patrick Breye was able to obtain through his request. Limited elements, with missing data, such as the name of the ethicist who was to supervise the project, at the request of the European Commission.
Source: BFM TV
