American companies, from Hollywood studios to tech giants, are increasingly betting on artificial intelligence tools (AI) and expect from their employees they do the same. The online media Business Insider reported in late July that Microsoft began to include in its assessment grids the capacity of employees to use AI. But even if the use of this technology is in the process of digging its place in our work standards, some employees refuse to use it. And the real obstacle is not necessarily the one we think of.
According to a study by researchers from the University of Beijing and the Hong Kong Polytechnic University, this reluctance is explained by a phenomenon called “Penalty Compence”. This is a bias where the use of artificial intelligence tools leads colleagues to perceive AI users as less competent, regardless of their real results. A prejudice which may seem insignificant at the start, but which could prove to be particularly detrimental for women occupying posts with technical skills, indicates the media Fast Company.
The study was conducted within a leading technological company, the name of which has not been revealed. In an article published in the Harvard Business Review, the authors explain that this company had already deployed to its developers a cutting -edge IA coding assistant, supposed “Significantly improve productivity”. However, twelve months later, only 41% of nearly 30,000 engineers interviewed had tested this assistant. Among them, there were 39% of male engineers aged 40 and over, against 31% on the side of women. The company having put as many possible devices as possible to promote the use of this tool, the researchers went deeper.
They selected a sample of 1,026 engineers from the same company and asked them to assess Python code. This code was strictly identical, but presented sometimes as hand -produced, sometimes with the help of an AI, by men or women engineers. Result: the engineers thought that those who had used AI to produce their code were 9% less competent on average. The penalty was worse for women (13% less) than for men (6% less). IA non-user engineers were also the most critical and non-user men particularly penalized women using this technology (26% more severely).
A double persistent standard
Knowing this, many employees – notably women and those over 40 – avoid AI for fear of this jurisdiction penalty, when they could however benefit from these productivity tools. Contrary to the idea that AI would equalize the chances, this study shows that its effects can worsen existing biases, especially in very male environments. The adoption of AI tools is on average 25% lower in women.
The authors of the study specify: “Those who dreaded the jurisdiction penalty in the tech industry, in particular women and older engineers, were precisely those who adopted AI to the least”.
Kamales Lardi, Author of the Book Artificial Intelligence for Business,, recently declared in an article by Fast Company: “According to my experience, women are often the subject of increased vigilance concerning their technical skills and capacities. There may be a deeply anchored fear that the use of AI is perceived as a way of cheating or as a sign of incompetence. ” This means that if women use AI less, it is not because they need it less, but because they have more to lose.
