Conference report: Bias at core of data sets a bad example
Caroline Criado Perez, author of Invisible Women: Exposing Data Bias in a World Designed for Men, outlined her work showing how data could be used much more effectively to ensure equality.
This was not merely an issue of equality, but life and death, she said; for example, women are 47 per cent more likely to be seriously injured and 17 per cent more likely to die in a car crash as a result of vehicle design based on male bodies – which does not take account of breast tissue or the need to have different designs for pregnant women.
“We need to take the emotion out of it and turn it into a numbers game [by using hard data] as that seems the best way to explain it,” said Criado Perez, arguing that the world was designed around “reference man”, a white Caucasian male designed to represent everyone, including women.
“We all default to him and it makes him harder to kill off,” she said. “He is the product of thinking that goes back centuries and starts with Aristotle – that sees women’s travel patterns, working patterns and bodies as far too complicated.
“We continue building and designing our world around the lives and bodies of men, because it’s simpler and cheaper.”
Criado Perez said 71 per cent of women who wear personal protective equipment (PPE) in their work have to use PPE not specifically designed for them: “Most occupational research is done on men or in male-dominated professions, and women are often excluded from research as ‘confounding factors’.”
She also highlighted how gender bias in artificial intelligence was having an impact, with a chatbot asked to diagnose a health problem in a male and female patient with exactly the same symptoms diagnosing a heart attack in the man and a panic attack in the woman. “The stronger the original bias, the greater it is in the application,” Criado Perez said.
She pointed to research by the London School of Economics, which showed that quotas in recruitment tended to weed out incompetent men rather than promote competent women.
Criado Perez concluded with three key points:
n Equality does not mean treating everyone the same, as women are not men.
n Sex-disaggregated data is crucial in doing data right – and if everyone did it, the world would change very rapidly.
n Diversity is not a tick box; it will affect how we collect data and what we look for.
In a panel discussion chaired by Criado Perez, Talat Yaqoob, director of Equate Scotland, which works to get more women participating in STEM (Science, Technology, Engineering and Maths), said there was a very strong body of evidence showing that diversity in the workplace increased productivity, profitability and innovation.
Asked why there appeared to be so little change – with women entering subjects like computing science at university actually falling in recent years – Yaqoob said: “Attitudes, behaviour and culture change don’t happen overnight and change is happening on the shoulders of women who are experiencing inequality in the first place.”
Yaqoob said many women took on additional roles in the workplace – such as co-ordinating equality and diversity committees – and that this work might take place in a quiet corner of an organisation and “was not often given the opportunity to drive culture change – and while it’s viewed like this, it’s always going to be slow-burning change and that change will be glacial.”
She added: “You might think that the biases are harming women but at Equate Scotland, we show people that the biases are harming their business, their sector. We work with an evidence base and can give examples of how gender equality benefits everybody and that means we get positive responses most of the time.
“You need to provide really obvious practical steps – like offering training and development around gender competency, because people need to understand inequality to do something about it.”
Professor Gillian Hogg, deputy principal at Heriot Watt University, a key partner in the Edinburgh Data-Driven Innovation programme, said: “We still do not get enough girls applying [for STEM subjects] and there are barriers that girls put up themselves and how they perceive themselves in relation to boys. The pipeline coming through to university is not strong enough.”
Hogg noted that when parcels arrived at her home addressed to Professor Hogg, delivery drivers would regularly assume they were for her husband.
“We all have a responsibility to highlight, question and challenge, and not allow assumptions to go unchallenged,” she added. “We have to acknowledge and challenge unconscious biases; we need to accept these unconscious biases are all around us; they start young and go right through.”
Former Labour MEP Catherine Stihler, now chief executive of the Open Knowledge Foundation, said we needed to work at community level to make a difference in how we use and perceive data.
“When something is closed, make it open,” said Stihler. “We need open artificial intelligence so we can actually look at what’s going on and so we have trustworthiness and accountability.”
Rachel Miller, a data scientist at Accenture, said it was important to show that careers in data were not “big and scary” and were open to women from a range of backgrounds, not just those who were great at maths. She thought it was also crucial to get buy-in for change from senior leaders.
Criado Perez said New Zealand Prime Minister Jacinda Ardern was a role model for a new form of female leadership. “The way she handled the Christchurch shooting was inspirational,” she said. “She showed real leadership and real empathy, which we do not associate with leadership. It is key to see empathy as not a weakness, but a strength.”
Making real change and ensuring the full data picture was taken into account in creating products was also about getting diversity into design teams, said Criado Perez.
“That affects everything,” she said. “If you have a homogenous group, they will just not ask the right questions.”
Stories from the cutting edge
The conference was the setting for the launch of the Women in Data project – which is tasked with providing “a snapshot of women who are inventing, innovating and changing the world in their offices, labs and homes”.
Poppy Gerrard-Abbott, a sociology lecturer at the University of Edinburgh, led the project to capture the “professional and personal observations, concerns and hopes of a wide range of women across the data innovation landscape”.
Gerrard-Abbott added that it had delivered a “volume, breadth and depth our imagination did not expect in the beginning”.
Gerrard-Abbott interviewed almost 60 women in data science and found they had very mixed views on how to deliver systemic change.
While many women thought upskilling girls and women and getting them into STEM subjects was the key issue, many believed the toxic nature of workplaces – and making them more welcoming to women – was fundamental.
“If we want to see inclusive growth and innovation, we must give women the same authority we give to men in the world and the workplace,” she said. “If we exclude women’s voices, we are perpetuating women’s exclusion as we enter the fourth industrial revolution.”
Batches of interviews from the Women in Data project will be released weekly on the Data Driven Innovation programme website – found at www.ddi.ac.uk – and Gerrard-Abbott encouraged delegates to go online and “read, reflect, enjoy and share this eclectic mix of stories”.
She urged anyone who despaired about “the long and laborious move towards change” to value, legitimise, believe in and listen to women’s stories. If we did that, she said, we would be taking a big step towards doing data right.