The tech industry is unequal – we know that. Gender differences in AI research are rife all over the world. According to a report by the World Economic Forum only 16% of people working in this area are women. This is a significant problem. Half the world’s population barely even participates in researching and developing a technology that affects every one of us and will be crucial in shaping our future.
It should be obvious that we need both diversity and gender equality to reflect the whole of humanity in these decision processes. The issues are clear: if we don’t include a range of perspectives, we risk incorporating bias and discriminatory behaviours into algorithms. There are lots of examples of this, including an AI recruitment tool that Amazon invested in a couple of years ago. The company used CVs they had received over the past ten years to train the tool. Most of the CVs were from men, which meant the software learnt to prefer men for the jobs. Amazon scrapped the project in 2018.
Another example of built-in discrimination concerns facial recognition, a rapidly evolving field of AI. Statistics show that it is up to 99.97 percent accurate under certain circumstances. But there are big differences. If you are a woman of colour the accuracy drops significantly. A New York Times article reported that the algorithms of one particular program recognised the faces of white men in 99 percent of cases, but women of colour just 65 percent of the time.
The software is only as intelligent as the data used to train it. These examples clearly demonstrate that if we don’t make changes, we’ll just get more of what we already have – bias and assumptions based on gender inequality. In fact, it gets worse – due to the lack of the diversity that is present in human decision making, algorithms amplify the human biases present in the training data, since that is all they know. This leads to algorithms not just propagating but strengthening the inequalities.
Unfortunately, we’re seeing the same stereotypes in the virtual world that we have in the real world. The best-known humanoid robots are ‘female’- Sophia from Hanson Robotics and Erica designed by Hiroshi Ishiguro. Sophia appeared on the TV programme 60 Minutes and on the Tonight Show with Jimmy Fallon, where she went on a date with Will Smith. Physically stronger robots such as Atlas and MIT’s rescue robot Hermes also have human features, but with a more masculine shape.
Then there’s the bias in chatbots. Not that long ago almost every chatbot on the market had a woman’s name and avatar. Customers pointed out these gender stereotypes to companies and it’s now more common to see chatbots with gender-neutral names and avatars.
We need to remember that AI itself is neither biased nor unbiased. Humans, however, are very biased, both consciously and unconsciously. And it’s humans who choose what we use AI for, as well as how we use it and what data we use to train it. So we need to include lots of different perspectives and approaches when we develop AI solutions, if only to be able to identify potential sources of bias.
Our social concepts about gender are at risk of being reflected in AI and robotics as well. We are building in basic gender differences in an area where gender doesn’t even exist and, therefore, upholding our prejudices and unequal structures.
Lack of women in the sector
According to an article in Wired, Google had 641 people working on machine learning and AI, of which only ten percent were women. Facebook had 115 people working in AI, 15 percent of whom were women. There’s no shortage of these kind of figures, including an Allbright Foundation report from May of this year called Tech Dudes Caught in Their Own Myth. The report shows not only that men are overrepresented in the tech sector, but also that men give jobs to men and, in particular, men invest in men. October saw the release of another report called Pandemin backar bandet [The Pandemic Is Pressing Rewind], which shows that the coronavirus crisis has slowed progress on gender equality made by listed companies. This is a worrying development.
So what can we do about this dual gender inequality in the tech sector? How do we ensure women are not excluded? And how do we make sure that all of society is represented? We believe there are three parts: training, recruitment and corporate culture. A lot is already being done to encourage girls to enrol in technical courses and, of course, that’s where everything starts. But from that point onwards, it’s managers at tech companies who have a major responsibility, both in terms of recruitment and corporate culture. We have to force the change until it becomes the norm.
More speed, less progress
Rapid development is a core element of the tech and start-up world. New technology evolves fast and speed to market is vital for many companies. A lot of tech companies grow quickly, sometimes going from five to five hundred employees in just a few months. But this pace poses a problem for both gender equality and diversity. It leads to rushed recruitments and women being side-lined in favour of men.
You often hear, ‘We don’t hire based on gender, we hire the best person for the job.’ But since women have accounted for the majority of university graduates for some time now and have excellent qualifications, it’s not the ‘best’ that are being hired, but rather the ‘most suitable’. And ‘most suitable’ often means people like the other employees, i.e. men, perpetuating the gender imbalance. Because it’s easy. Hiring someone like you is the safe option, and it’s even more likely to happen when hiring is rushed.
Corporate culture that creates opportunities
Gender inequality is a societal problem, but businesses in general, and tech companies in particular, have a big responsibility. The majority of Solita employees are still men, so we’re not top of the class. But we’re working hard to change this. We’re taking a proactive approach to all aspects of gender equality and inclusivity – from how we write our job ads to our long-term, strategic business decisions. We believe strongly that corporate culture forms the basis of everything, and our values aren’t just empty words – they steer everything we do.
Our culture has a firm focus on gender equality, diversity and inclusivity. We see each individual and aim to build teams with a range of views, ages, genders and backgrounds, because we think diverse groups innovate better. We want our solutions to reflect the modern world and believe this is achieved through different perspectives and opinions. All of society must be represented.
The fourth industrial revolution has arrived and together we have a unique opportunity to build a new, gender-equal world. It’s the tech sector’s responsibility to hire and promote more female talent in AI and AI research. We need to establish corporate cultures that take gender equality and diversity issues seriously. Investors need to support more women founders and we must create more gender-neutral images of AI and robots in the media.
Tech companies have the power to shape our society right now. Excluding half of humanity is unacceptable and hinders new thinking and innovation.
Want to find out more about Solita’s approach to AI? Here is our report: The Impact of AI about AI and ethics.