AI is Failing Women. For Everyone’s Sake, that Needs to Change.

Table of Contents

[ad_1]

Add to favorites

Technology has always had a diversity problem but as we enter a world driven by AI innovation, this will have major implications on how the global population engages in the digital future.

From robots and voice recognition to smart fridges and driverless cars – artificial intelligence is becoming widespread – but will these advancements meet the requirements of both men and women? Without diversity, the answer is no, writes Shawn Tan, CEO of AI ecosystem builder Skymind Global Ventures.

Research compiled by the Datatech Analytics for the Women in Data campaign found that only 25{f08ff3a0ad7db12f5b424ba38f473ff67b97b420df338baa81683bbacd458fca} of UK jobs in artificial intelligence and other specialist technology roles were filled by women in 2019 – the lowest proportion in two decades. Numbers elsewhere reflect a similar figure. The World Economic Forum (WEF) estimates that 78 percent of global professionals with AI skills are male — a gender gap three times larger than that in other industries.

Women and men don’t hold the same types of AI jobs, either. Males are more likely to be in senior positions, such as software engineer or head of engineering. Women in AI usually do less influential jobs, such as data analyst or researcher.

The consequences of a homogeneous ‘male’ workforce is the creation of machines and systems that are designed with inherent gender and racial biases.

In her book Invisible Women: Exposing Data Bias in a World Designed for Men , Caroline Criado Perez reveals how women are being shortchanged by the limitations of gender-blind technologies, leading to user outcomes that can be amusing and annoying at the best of times, as well as harmful.

She gives examples such as map apps that fail to show the ‘safest’ routes to a destination in addition to the ‘fastest’ routes; and seat belts and airbags that are tested on dummies with male torso and height dimensions – leading to greater female casualties on the road.

Voice-command technology also fails to meet the needs of women. In her book, Perez tells the story of how her mother tried to call her sister using the voice recognition system in her Volvo. She kept on failing in her attempt until her daughter suggested she lower her voice like a man. It worked.

Given the car was produced by a company founded in Sweden – a nation with a reputation for gender equality – you’d expect the car designers would get this technology right – but the code for the system was almost certainly created by men miles away in Silicon Valley.

The flaws inherent in voice AI are worrying, given its growing popularity. Google estimates that 20{f08ff3a0ad7db12f5b424ba38f473ff67b97b420df338baa81683bbacd458fca} of their searches are now done through voice query – and that number is expected to rise to 50{f08ff3a0ad7db12f5b424ba38f473ff67b97b420df338baa81683bbacd458fca} by 2020.

But research on Google’s own speech recognition software reveals that their system is 70{f08ff3a0ad7db12f5b424ba38f473ff67b97b420df338baa81683bbacd458fca} more likely to recognise male voices over female. Moreover, speech recognition struggles to understand different accents , which will severely impact the efficacy of the innovation and industries like IOT, which has built vast service offerings with voice activation at its core. Everything from turning on the lights to setting the temperature in your house and locking the gate. Imagine if you’re immobile, at home alone, and rely on this technology to give you autonomy – and it doesn’t work because you sound different to the test cases used to produce it? The consequences could be devastating.

The same biases found in voice activation also exist with facial recognition technology. Tech titans like Amazon have been called out for delivering AI systems that fail to perform accurate recognition on female and non-white faces. When it comes to recognising the gender of a face, most systems identify male faces better than female faces and have error rates of 1{f08ff3a0ad7db12f5b424ba38f473ff67b97b420df338baa81683bbacd458fca} for lighter-skinned men. White women are misclassified as men 19{f08ff3a0ad7db12f5b424ba38f473ff67b97b420df338baa81683bbacd458fca} of the time, and, according to research conducted by Algorithmic Justice League, the errors increase to 35{f08ff3a0ad7db12f5b424ba38f473ff67b97b420df338baa81683bbacd458fca} for non-white women.

The impact of this gender and racial bias is profound. Facial technologies are being developed for commercial purposes and as businesses start to market services using facial recognition for security, policy and vetting job seekers, women and people of colour will continue to be marginalised, this time by machines – instead of humans

Diversifying the Workforce

If artificial intelligence is to reach its full potential, we need to diversify the people building these systems and attract more women to the industry.

But how?

First, STEM skills must be prioritised in primary and secondary school curriculums, and available to all students with an emphasis on coding and software skills.

Second, we need more mentoring programmes to encourage women and people from different backgrounds to enter technology and AI professions. This should start at secondary school to inspire the next generation of digital workers.

Mentoring should also continue throughout an employee’s career and there should also be initiatives and programmes in place that help to build communities and networks that allow people to support one another – especially in the coding world – which forms the foundation for AI.

A good example of community building in the UK is the work being done with the software bootcamp Makers. 35{f08ff3a0ad7db12f5b424ba38f473ff67b97b420df338baa81683bbacd458fca} of its cohort is female – twice the national average and they attract students from different social and racial backgrounds. Makers’ diverse talent is in high demand – and they also help to create positive community engagement programmes that celebrate role models for those underrepresented in tech such as the women in software powerlist.

As an AI ecosystem builder, Skymind Global Ventures is also investing in community and education – supporting programmes around the world that put training and diversity at the heart of their coursework. We are planning to open one of the world’s biggest AI universities by the end of the year – and, with help from industry leaders, will devise educational training that reflect the types of skills that are required by the sector today – and sponsor people from all walks of life to become vital AI talent for the companies building our future.

Finally we must legislate diversity in AI. Nothing can happen without support from the government.

Despite the sobering statistics around diversity, we’re seeing some positive changes. Many businesses and institutions are making concerted efforts to recruit more women and people from different backgrounds. Coding schools are attracting more female students and returnship programmes, aimed at luring women back into the workforce after years off raising families, are gaining popularity. Apprenticeships are also becoming more effective in training up and getting people through the door of AI companies.

Artificial Intelligence has a long way to go before it truly embraces diversity and inclusion, but the growing debate is creating a movement that can help to shape an AI future that is reflective of society as a whole – and therefore good for everyone. Let’s continue this progress… Let’s continue to talk – and take action!