AI photo check exposes scale of diversity problem at top firms – New Scientist

Men are on board

H. Armstrong Roberts/ClassicStock/Getty

By Timothy Revell

Bias in boardrooms is tricky to assess. Many companies dont publish diversity reports, making useful information difficult to come by and hampering efforts to tackle institutional biases. Now artificially intelligent algorithms have been used to dig down into the data, confirming that there is a lack of diversity at the top of the worlds corporate ladder.

To evaluate the situation, researchers from biotech firm Insilico Medicine compiled pictures of the top executives taken from the websites of nearly 500 of the largest companies in the world. The final dataset comprised over 7200 photographs from companies spanning 38 countries.

They trained image recognition algorithms to automatically detect the age, race and sex of the board members, and compared the results to the age, race and gender profile of each firms country to see if they reflected the general population. AI is far from perfect at interpreting images and Insilico Medicine doesnt specialise in this particular area, so the results should be taken with a pinch of salt. But, nonetheless, they do give an impression of the current state of play.

Evidence from other studies suggests that boardroom diversity is increasing year on year, but it is clear there is still a long way to go. Overall, the team found that only 21.2 per cent of the corporate executives in the study were female. And in every single company, the percentage of female board members was lower than the percentage of women capable of work in that country. Twenty-two companies had no women on their boards, with the majority of those firms being in Asia.

Nearly 80 per cent of the corporate executives in the study were white, with 3.6 per cent black and 16.7 per cent Asian. South Africa had the highest proportion of black executives, representative of the fact that 80 per cent of its population is black. However, the two South African companies included in the list still only reached 54 and 35 per cent in terms of the proportion of black board members.

In the US, many companies reflected the 12 per cent of the population that is black in their boardrooms, although there were also 30 companies without any black board members at all. The median age across all corporate executives in the study was 52.

These huge companies lead industries and influence our everyday lives. Using machine learning makes it possible to examine their diversity in a way that couldnt be done before, says Polina Mamoshina at Insilico Medicine. The data for the study was collected on 20 March.

This paper confirms that we live in a biased world, says Sandra Wachter at the Oxford Internet Institute, UK. However, acknowledging the problems this causes is only a crucial first step. Having a public discourse about these issues is vital. It is important to find out where the biases stem from and tackle the roots, she says.

Anti-discrimination laws should be used to achieve parity at the top of companies, and a shift in mentality is required to start viewing diversity as an advantage, says Wachter. The studys methods could be used in any situation where management profiles are available, but diversity data isnt, and could help examine the diversity within governments, universities or media outlets.