Facial recognition can be responsible, ethical, and trustworthy by Amine Kacete
Facial recognition: A concept that deserves clarification
Most of these debates lean heavily on the definition of facial recognition. They commonly reduce it to identity verification software rather than its broader applications. From a purely scientific standpoint, a fairer definition encompasses all intelligent algorithms that can automatically analyze a face taken from images or videos. These applications include locating faces, predicting macro/micro-expressions and emotional states, reconstructing 3-D geometric information about the face, and automatically inferring facial features such as hair color, eye color, or age. By adopting a broader definition of facial recognition, a more objective assessment of its technological and social impact becomes possible.
Impacts of a multifaceted technology
As with any emerging technology, there are two sides to facial recognition, one negative and one positive.
The first is based on a commonly held view of the general background in which this technology has developed. The argument goes as follows: CCTV (Closed-Circuit Television) surveillance systems are massively expanding (about 800 million worldwide and about 400,000 in France[1]); historically reserved for the most vulnerable locations, they have become widespread and are now part of the urban landscape. Criminal activity is strongly correlated with crowd movements, making it very difficult to predict and avoid. Governments are increasingly falling back on this technology to remedy such security flaws, feeding worries of a big brother scenario[2]: Constant large-scale surveillance with in-depth profiling and control that threaten individual freedom.
The more positive side is that facial recognition has never been so accurate, fast, and convenient. It can help overcome major technological obstacles and greatly improve human quality of life. For instance, by incorporating its full potential to accurately identify facial features, medical care be delivered in an increasingly personalized way.
To give another example, thanks to ultra-precise identification, missing persons can be located in record time, and recently, in some countries, this capacity for identification via facial recognition has proved critical in managing lockdown scenarios to reduce the spread of viral illnesses. Elderly people can also get automated assistance based on identification of their expressions and emotional state.
Additionally, the spread of deepFake technology has put personal digital identity at risk; facial recognition is a strong way to protect content by identifying fakes.
An opportunity for Europe
History has taught us that banning a technology entirely to avoid its misuses while ignoring all of its advantages has never been a wise solution. Furthermore, in the context of facial recognition, the tools and algorithms that make it effective are accessible to everyone through open-source sharing.
Today, eliminating this technology altogether would be nearly impossible. A regulatory approach is therefore the smarter choice. Without regulations, or in the presence of nearly partial regulations, the negative effects will potentially outweigh all the benefits of facial recognition. Some Asian models show potential misuses of a widescale facial recognition deployment, both with respect to the number of people identified and the number of public places targeted. Additionally, these deployments often involve automated interpretations of the facial recognition data. It is reasonable to ask how these automated decisions are affecting the everyday lives of Chinese citizens. The partial regulation of the North American model raises other questions. Currently there is no strict control regarding access to or collection of the data that is used to train and improve facial recognition. Where is this data stored? What purposes can/must it be used for? By including restrictions based on ethics, law, and collective responsibility, Europe has already undertaken serious regulatory initiatives such as GDPR.
The white paper on AI that it just published, already supported by major players such as Microsoft, confirm that a European “third way” is possible. A new paradigm centered on ethical, responsible, trustworthy facial recognition is possible. First, the impressive performance of artificial intelligence, which continues to grow, will further accelerate the resolution of “ethical” issues and clear the way for other benefits to this technology. Second, rapidly designing solid regulatory processes that govern both algorithm training and the deployment of decisions made by the technology will increase society’s acceptance and trust. Several concrete initiatives offer a model. For instance, in Germany, some areas deemed to be vulnerable visually indicate that an automatic facial recognition system is in use, just as with signs indicating CCTV surveillance.
Prospects not yet considered
With the artificial intelligence revolution, multiple digital technologies are emerging and attempting to incorporate in their core algorithms a pure-AI logic.
The best example are videos using 3D Computer Graphic Imagery. This technique allows past and present movie stars to be brought to life from photographs with a very advanced degree of realism. It is conceivable that actors could film their scenes based solely on their 3D digital data without needing to be physically present. This would open up new movie industry business models that depart from established ethical and legal practices.
Finally, solving the problem and addressing the issues raised by facial recognition potentially means developing a data protection model that can serve as a standard for all other technologies that pose similar obstacles. In summary, when supported by regulation and substantial efforts at standardization, facial recognition offers hope for any technology based on digital data.