By Ella Walters, Head of Data Science, Constellation AI

It came as no surprise that when a recent report from McKinsey found that, although companies say they are committed to improving diversity, little actual progress has been made to improve the representation of women – as well as other under-represented minorities – in the workplace.

This is exceptionally prevalent in the technology sector, where the proportion of women to men is actually falling.

For many years, a significant proportion of software programmers were female. Indeed, in 1967, an article in Cosmopolitan magazine, entitled “The Computer Girls”, described computing as an ideal career path for young women. Today, however, only around one in five (22 percent) of professionals working in AI – comparably, a form of modern day computing – is a woman.

As it stands, the AI industry isn’t as appealing women as it is for men.

A study by PwC reveals that just over a quarter of women, 27%, would consider a career in technology, compared to 62% of men. Of this quarter, only three percent said it would be their first choice. And according to Girls Who Code, only 7,000 of the 40,000 Computer Science graduates in the US in 2017 were women. Perhaps worst of all, it’s unlikely that these figures shocked anyone.

There are a number of different plausible explanations for this gender gap. A neglect of focus on STEM subjects in schools, for example, may be limiting the number of possible career opportunities at an early stage, while a largely male culture and a lack of female role models and mentors is likely a limitation for young women considering a job in the industry.

Unconscious masculine bias in the recruitment process could also be preventing potentially stellar female candidates from progressing beyond the starting post – and this bias may be making its way beyond just human decision-making.

With AI becoming an increasingly mainstream technology within corporations, and used to make real-world business decisions (including those around recruitment), there is an ever-present, and increasingly sinister, threat of bias from the white-male-dominated tech scene.

Biased algorithms

Last week, it was revealed in a report by UNESCO that artificial intelligence voice assistants, such as Amazon’s Alexa and Apple’s Siri, are perpetuating and spreading gender stereotypes, fuelling the stereotypes that women are “obliging, docile and eager-to-please helpers.”

Some cases even found assistants “thanking users for sexual harassment” and that sexual advances from male users were tolerated more than from female users.

While shocking, in some ways, this is unsurprising, given the dominance of a singular perspective of white males who are developing the technology. AI relies on a series of algorithms to function.

These algorithms tell Facebook which ads to serve, tell Netflix what films to suggest, and tell Google how to prioritise the news it shows.

While in these instances a little bias might not have catastrophic consequences, AI is being used in increasingly sophisticated contexts, such as making instant decisions on loan applications, sentencing criminals in a matter of minutes, or automatically sorting through CVs as part of the recruitment process.

Ideally, such decisions should be fair and just, unencumbered by human emotion or prejudice. In reality, however, increasingly more AI algorithms will be subject to bias.

In some cases, this is down to environmental conditions. As AI moves to the mainstream, now is the time to act on levelling out this unconscious white male bias, before its implications start to have hugely detrimental societal implications.

Amazon, for example, recently brought an AI-powered recruitment experiment to an end when it was discovered that the machine learning technology behind it was displaying bias against female applicants.

Upon investigation, it transpired that after observing that the majority of CVs submitted to the company were from male candidates, the solution had taught itself that men were preferable to women.

We’re in an age where AI pervades many aspects of our lives. But if this AI is only being created by individuals of one perspective, then it stands to reason that there will always be bias. Indeed, if men continue to outnumber women by a ratio of four to one, it’s possible that such bias could eventually become the norm.

If we hope to eliminate this bias, then, it’s vital that the tech industry strives for greater representation of every race, culture, sexual orientation, and gender, for the sake of our

Begin at the beginning

Of course, eliminating unconscious bias isn’t the only answer. We need to do more to attract young women into considering a career in science, AI and technology in the first place, and this needs to start by re-evaluating how STEM subjects are taught in schools – especially to girls.

Only 35% of STEM students in higher education around the world are women. But by improving primary and secondary education, we would hope to see this rise. Primary schools have recently started teaching basic coding programmes, which is a step in the right direction, but there’s still much more to do.

Encouragingly, the volume and variety of post-university roles for STEM graduate is evolving – a world of possibilities is opening up for women and under-represented groups. But these groups need to be able to see that these jobs are for them, and not just jobs for people like those who came before them.

If we hope to further diversity – and that encompasses diversity of all genres – into the tech industry, we need to address the issue of perception. If we hope to keep them there, we need to address the issue of bias. The time has come to level the AI playing field.