ARTIFICIAL INTELLIGENCE, NOT WITHOUT HER !
"Nothing is lost, nothing is created, everything is transformed" and with AI, everything is amplified, and women disappear."
AI has become part of our daily lives without us sometimes being aware of it: from content suggestions on social networks, to applications using biometric authentication, automatic medical diagnoses or hiring procedures. It goes without saying that it is now a major strategic tool. However, the infringements of fundamental rights are very real without us being able to really discern the contours, measure the risks or even control them.
AI is like teenage sex, everyone talks about it, everyone is convinced that others are doing it but nobody knows what it is?
In simple terms, AI is the set of methods used to create machines capable of simulating human intelligence. More technically, according to computer scientist Shortliffe, AI "consists of designing algorithms that allow computers to perform tasks that require symbolic reasoning, going beyond simple mathematical calculations". In other words, these intelligent algorithms are characterized by the autonomy to make their own decisions, to predict results, to adapt and evolve according to the data collected through the mechanism of self-learning. Beyond a simple calculating machine, AI would therefore be a thinking machine.
However, it may seem like a misnomer to speak of "intelligence" because it has nothing to do with human intelligence. Indeed, although AI's calculation capacities are impressive, the machine can only learn and reproduce what it has already seen and what humans teach it. Designed to automate certain human activities, it has no capacity for discernment, reflection, morality or common sense. AI is therefore nothing more than a reflection of human thought.
AI, a reflection of Man, for better or for worse ...
Free from the prejudices and biases of human affect, algorithms are often mistakenly seen as an ideal of neutrality and reliability, conducive to equal treatment. However, like any tool, there are intrinsic risks and, in this case, algorithmic neutrality does not exist. Indeed, AI is increasingly pointed out as many studies show that it contributes to reinforcing the sexist biases present in our society. It is as much a subject of fantasy as of uncertainty, of optimism as of concern.
"Artificial intelligence is the icon of modernity. On the contrary, when it comes to gender equality, it is rather archaic.”
- L’intelligence artificielle, pas sans elles ![1]
Our latest digital inventions therefore continue to convey sexism and gender stereotypes, undermining the place of women in society with each passing day. AI not only reproduces a world riddled with gender inequalities, but also amplifies them. It creates the exponential risk of reinforcing discrimination automatically while giving it an objective appearance, as demonstrated in the report by the CNIL and the Defender of Rights[2].
AI's fuel is data
AI owes its success to the presence of massive data combined with gargantuan computing capacities. It is based on three elements in particular: algorithms, data and computers. While computers are in principle neutral and cannot be accused of bias, algorithms and data can be biased to arbitrary degrees.
Indeed, when a category of population is not proportionally represented in the training databases compared to the real world, this creates a distorted representation of the world for the algorithm from which it will draw biased conclusions. Similarly, erroneous or incomplete data present for a category of individuals will lead to biased decisions, increasing the risk of error and discrimination against them. These conclusions can have serious consequences on the lives of individuals: access to employment, career development, access to information, credit, education, etc. As they are described as black boxes, these intelligent algorithms therefore call for obligations of transparency and interpretability.
Under-representation of women in AI professions
The low numbers, or even absence, of women in the artificial intelligence professions is also one of the reasons for the risks of biased results delivered by AI. While forecasting organizations claim that these professions will account for the majority of new jobs over the next decade, women represent only 15% of data scientists and the situation continues to deteriorate. Indeed, the share of women in the world of AI continues to decline both in terms of choice of post-baccalaureate studies and at the stage of professional integration and access to positions of responsibility.
Thus, at a time when the gradual elimination of women in the sector is taking shape, the feminization of digital professions is more urgent than ever and essential to respond to these problems. On the one hand, it will make it possible to rebalance the presence of women in these professions in order to create an inclusive environment. On the other hand, it will reduce data bias in algorithms thanks to the diversification of profiles.
An ethical question: what future do we want for mankind?
"Science without conscience is the ruin of the soul”- Rabelais
Let us remember that ethics is a set of moral principles, a conception of the "good", which guides and motivates the behavior of Man in society. Thus, since AI is amoral, it is not so much that it should be ethical or not, but rather its uses, imagined and developed by humans, because only the latter have freedom of choice in conscience. However, the designers of algorithms, like the organizations using this type of system, do not show the necessary vigilance to avoid the automation of sexist discriminatory effects. Hence the need to think of a trusted AI, respectful of fundamental freedoms, in order to meet these challenges.
Alerted to these problems, the scientific community, on the one hand, has embarked on research projects in Ethical AI (Fair Machine Learning), aiming in particular to correct the discrimination bias of models but also to make the results of the algorithms explainable. Governmental organisations[3], on the other hand, are trying to set up a normative framework to regulate the development, use and control of AI.
Specifically, the Draft Recommendation on the Ethics of AI[4], adopted by Unesco's 193 members on 24 November last, sets out values and principles that constitute a common basis for the formulation of laws and regulations by member states. For example, one specific recommendation concerns the issue of discrimination :
"Member States must ensure that gender stereotypes and discriminatory biases are not carried over into AI systems, but rather proactively identified and corrected.”
Furthermore, the European Commission presented in April 2021 the "Artificial Intelligence Act"[5], a proposal for an AI regulation that advocates trustworthy AI, respectful of European human values and fundamental rights.
It should be noted that responsible technological innovation is not a hindrance to the competitiveness of companies, on the contrary. On the contrary, companies that are able to explain how their algorithms work in a transparent manner can have a real advantage over their competitors. Trust through transparency is a guarantee of retaining or even expanding a customer base. We could then think of creating a Digital Corporate Responsibility, following the example of Corporate Social Responsibility (CSR).
Ultimately, it is up to humans to define, think, introduce and maintain ethics at the very heart of AI systems.
[1] L’intelligence artificielle, pas sans elles !, 2019, Aude Bernheim et Flora Vincent.
[2] Rapport mai 2020, Cnil et Défenseur des droits : « Algorithmes : prévenir l’automatisation des discriminations »
[3] CNCDH, Avis relatif à l’impact de l’IA sur les droits fondamentaux, avril 2022 : https://www.cncdh.fr/sites/default/files/a_-_2022_-_6_ _intelligence_artificielle_et_droits_fondamentaux_avril_2022.pdf
[4] Projet de recommandation pour une éthique de l’IA, UNESCO, 2021 : https://unesdoc.unesco.org/ark:/48223/pf0000379920_fre.page=15
[5] Commission européenne, “AI Act”, avril 2021 : https://digital-strategy.ec.europa.eu/en/policies/european-approach-artificial-intelligence