Gender Summit 17: Open science, diversity and AI

One of the key themes for the Gender Summit in Amsterdam in October 2019 covered fostering diversity in open science and AI to better connect science to society.

Greta Byrum from the Digital Equity Laboratory in the US described the state of play as she sees it. “This data almost feels radioactive,” she warned. The rush to market for automated digital systems based on AI is leading to built-in bias in areas such as facial recognition, policing and health care. “In the US, there is little protection for highly sensitive data,” she said. “You should be very happy that GDPR exists in Europe!”

She cited the example of period tracker apps that share data with third parties without explicitly asking for user permission, potentially sharing highly sensitive information about pregnancy status in a way that could intersect very badly with changes to legislation around abortion in the US.

“As datasets grow, biases can be reinforced rather than ironed out, depending on who decides on data quality,” she explained. For her, society could – and should – insist on and drive for industry standards, a code of ethics and better public awareness of what might be happening to their data.

Cecile Greboval from the Council of Europe had similar stories to tell of countries using flawed algorithms to decide on disability benefit rather than face-to-face interviews, relying on AI to filter admissions to university and image databases that return sexist images for simple search terms, such as female police officer. Women are severely under-represented in ICT, at only 13% of ICT-related graduates working in digital jobs in 2016 and numbers have actually been falling in recent years.

“We think that technology is neutral, like the law. But 80% is designed by white men who can project their own world view on to the data,” explained Greboval. “Gender stereotypes in childhood and education, together with sexism and harassment in the workplace have led to a dearth of women in IT. This means women are excluded from well paid, powerful jobs.”

Greboval called for safe conditions conducive to women at work and pointed towards the Council of Europe action against sexism: See it. Name it. Stop it.

Built-in bias

Algorithms can have bias built in from the ground up if particular groups are under or over represented in datasets. Sexism and stereotypes become transmitted to machines and are sometimes amplified, with historical inequalities hardwired into the data. Ozgur Simsek from the University of Bath described how Buolamwini & Gebru (PMLR, 2018) demonstrated that facial recognition systems perform better at detecting men than women and at detecting light-skinned than dark-skinned people.“The worst rates are for darker skinned women, by up to 35%. When guesswork will give you a 50% error rate, there’s not much learning going on there,” said Simsek wryly.

Caliskan, Bryson and Naraynan (Science, 2017) analysed 840 billion words – from tweets, the US Declaration of Independence, Reddit threads and many other sources. They found the same biases in text as shown by people taking the Harvard Implicit Association Tests. Unsurprisingly, machines learning from these datasets will tend to incorporate the same biases as humans.

To counteract these biases, Bath University puts learning ethics and transparency on the same footing as algorithms on their AI courses, combining computer science, AI, engineering, social science and policy making. “Students must cover all these areas to graduate,” said Simsek.

For Ghislaine Prins of Randstad, the route to improvements is to review datasets, establish diverse teams for creating algorithms, train leaders – and check, check, check again. “Then check the check!” said Prins.

Gender Summit 2019, Amsterdam: Identifying concrete measures for change

In Amsterdam last week, gender and inclusion professionals gathered for the 17^th Gender Summit. Ingrid van Engelshoven, Minister for Education Culture and Sport introduced the three themes for the event, which chimed with her ambitions for Dutch science policy: national frameworks to advance gender balance, diversity and inclusion; fostering diversity in open science and AI to connect science to society; actions towards a team-driven, innovative academic culture.

For van Engelshoven, there are two points to bear in mind to achieve gender equality. “We should judge research institutes by their gender equality and assume gender equality is the norm.” She cited the example of Emily Warren Roebling’s largely unsung contribution to the construction of the Brooklyn Bridge in New York after her engineer husband developed caisson disease. “Let’s not take 14 years to build bridges to equality,” she urged. “I am looking for very concrete measures we can take now, not in 10 years time.”

Mind the gap

Belle Derks of Utrecht University was the first speaker to rise to the challenge by identifying a number of ‘gender gaps’. “In Europe, the gender pay gap in academia is about 800 Euros. Age is a large factor in this as 50% of women drop out earlier in their careers. However, we actually see the widest gap at the highest pay grades.” Derks reported that there was no evidence that women negotiate less for their pay. “In fact, their more precarious employment often means that they negotiate more.”

Women at higher grades also report they spend more time teaching and have fewer resources in terms of staff, budget and equipment, leading to a ‘resources gap’. When men look ‘up’, they see people at the top like them. For others, the lack of role models can correspond to a ‘belongingness gap’.

Unconsciously, we expect women to be communal and men to be agentic and tend to dislike those who do not conform to these stereotypes. Simply trying to be more agentic is not as effective for women and taking a communal approach is not as valued. Women feel a lack of fit, which can lead to less engagement, work exhaustion, lack of agency and higher turnover. “The concrete solution here is to control for masculine definitions of excellence, focus on team science and value a diverse set of qualities in our reward systems,” explained Derk.

Bias in a meritocracy

For Prof Simone Buitendijk of Imperial College London, universities need to recognise the pernicious effects of bias and accept that it exists. “It’s not about being nice to women and ethnic minorities, it’s about including all talent to tackle global challenges,” she insisted. “If you tell us bias is not true, how dare you!”

The pervasive nature of bias is at odds with scientific research as a meritocracy. For example, BAME students experience more mental health issues while studying, which can become a vicious circle, impacting their eventual results.

“Just like a canary in the coal mine, minorities suffer most from competitive, individualistic and vicious atmospheres,” she reminded us. “Generally, we are poor at measuring excellence. No one can achieve perfection, even the superstars.”

“We need to tackle the system, but not shame individuals – unless they are in denial!” she said. Buitendijk called for leaders to be strategic in their approach and not let equality and diversity become the topic they aim get to once everything else is fixed. “Don’t leave it to the lone diversity officer in their cubicle,” she urged to a rueful laugh from the audience. “If we blindly insist that research is a meritocracy, then people blame themselves for bias in the system.”

The success factors linked to sustainable change are outlined in LERU’s recent report: “Equality and diversity at universities: The power of a systemic approach.”

• Discover and include a wide range of students
• Realise the potential of all staff and students
• Enhance performance and well-being
• Create an attractive community for all
• Increase the quality of knowledge production
• Connect with societal challenges

“Leaders should understand the statistics, listen to individual stories and take a strategic response,” urged Buitendijk. Hopefully, everyone in the audience is here to do just that.

Going for Athena SWAN Gold at University College London with Prof Sara Mole

“The gender challenge for STEM is that we are losing talent.”

Professor Sara Mole’s opener for the latest Equality in Science talk at the Wellcome Genome Campus on 16 September neatly summed up the ethos behind University College London’s approach to Athena SWAN. As a Gold award winner, the MRC Laboratory for Molecular Biology (LMCB) has a success story to tell. For those of us currently grappling with Athena SWAN applications, the key questions is – how did you do it?

In Physics in the UK, women students are mainly lost between GCSE and A level, then numbers remain relatively stable. Chemistry sees a fairly steady decline in numbers from A level onwards. In Biology, there is actually an issue with missing male talent up to undergraduate level, where 65% of students are female. Female talent is lost sharply in the transition from postdoc to lecturer, with numbers declining still further beyond that. Athena SWAN targets this loss of talent at institutional and departmental level.

UCL’s LMCB first won a Silver Athena SWAN award in 2009, graduating to Gold in 2016 when it secured only the eighth Gold award in the UK. As a postgraduate research department, with 18 Principal Investigators, postdocs do not stay on as lecturers – effectively LMCB trains them up for success elsewhere. Lecturers in their turn come from postdocs trained by other institutions. This has led to two key areas of focus for LMCB: postdoc career development and encouraging applications from women to lecturer positions. Appraisals then focus strongly on promotion, to foster a career path to associate and full professorships.

For a successful Athena SWAN application, you need to demonstrate that your actions have impact. “Changing the culture of a department benefits everyone,” Prof Mole reminded us. “We homed in on rationale, action and impact. Why did we do what we did, how and when did this happen and what were the effects?” The key for LMCB lies in solid data, otherwise it is difficult to know what changes have actually made that crucial difference.

The team behind the changes targeted a 50:50 approach, where everyone contributes and all protected characteristics were represented within the team, including a range of organisational roles and career stages. The impact of this approach has been significant – 36% of professors are now female compared to 18% in UK biosciences generally and they have achieved gender parity in applications, shortlisting and offers for PhD students. Staff surveys show that there has been a big increase in the numbers willing to encourage both girls and boys to consider a career in STEM. LMCB has taken positive action by inviting applications from under-represented groups and improvements in awareness of personal development training led to much better uptake.

LMCB works hard to bring speaker ratios closer to 50% by encouraging staff to nominate a speaker of the opposite gender. Cultural changes have led to a supportive, collegiate atmosphere underpinned by active mentorship. Postdocs have career timelines, support with writing and interview practice, meaning that female and male postdocs are now equally successful in winning promotion. Role models also have their part to play, including young group leaders who are one step ahead of the postdocs. “What is inspirational is to see someone like you or me who has taken that next step,” said Prof Mole. Careers talks also cover pathways outside academia for both genders. “We don’t want men to feel ‘trapped’ in academia or for women to feel ‘excluded’. Everyone should have equality of opportunity and the information they need to make career choices according to personal preference.”

For Prof Mole, it is important to reward and celebrate equality and diversity activities, including those that provide career development opportunities that benefit everyone, as well targeting under-represented groups. “We found our vision about half way through our Athena SWAN journey,” admitted Prof Mole. “Our hashtags #simplygoodpractice and #benefitsall indicate that everyone benefits but under represented groups will benefit most. We’re aiming for 50:50 and this should be the norm for the next generation.”

To make your changes really stick, Prof Mole has the following top tips:

• Gather your enthusiasts
• Know your data
• Find your vision
• Identify your priorities
• Fix what is in your immediate power to change
• Plan actions to address priorities
• Implement and keep up the momentum
• Keep communicating the vision
• Monitor for impact
• Don’t reinvent the wheel (“We plagiarised constantly!”)
• Keep deepening and start to beacon

Asked for two things that we could all take away with us, Prof Mole replied, “Be a really good colleague and treat others the way you would like to be treated. Get to know people to break down unconscious biases and assumptions.”

“Don’t make choices now based on what may or may not happen, just do what you want to do – that’s the best advice I’ve ever received,” she said finally. “Don’t waste time on those who are very negative at the start – let the impact and results convince them later.”

I declare myself convinced!

Zebras in the workplace – how to support employees with long term conditions and invisible disabilities

“cropped zebra”by Alice Chaos is licensed under CC BY-NC-SA 2.0

“When you hear the sound of hooves, think horses, not zebras.” This is the advice given to medical students in training, to encourage them not to ignore common conditions when in search of a patient diagnosis. This week I attended a training session on invisible disabilities by Dr Sara Booth, Honorary Associate Lecturer at the University of Cambridge. Invisible disabilities include what are sometimes called ‘zebra conditions’ due to their rarity, such Ehlers-Danlos syndrome, lupus, autism spectrum disorder and many others.

Under the Equality Act 2010, disability is a protected characteristic, which means it is illegal to discriminate against someone on the basis of their disability. Unfortunately, the Equality Act does not mention invisible disabilities and only some conditions, such as cancer, are specifically named so that they are protected from day one of diagnosis. For others, the qualifying line between a long term condition (LTC), an invisible disability and a medical condition can become very blurred.

When many of us think of disability, we picture an impairment that is permanent and fixed, such as sight loss or paraplegia. Employers might consider that once they have put reasonable adaptations in place, their legal duty has been met. However, a frequent characteristic of an invisible disability or a long term condition is that symptoms will vary significantly over time, whether to gradually worsen, develop co-occurring conditions or flare up unpredictably. The impact of taking medication for a life-long condition, over decades, can be unforeseen and may cause someone’s health to deteriorate in the long term. For sufferers, this can pose significant challenges at work as the adjustments they need day to day can also change dramatically.

For most of us, our jobs provide not just earnings but also social networks and status. Being excluded from work by a disability, even temporarily, affects all of these areas. By law, this exclusion should not happen, but invisible disabilities present particular challenges.

According to the Shakespeare 2010 Social Model of Disability, there is an important difference between someone’s medical condition – their impairment – and society’s response to it – their disability. So while someone’s medical status might qualify as a disability under the Equality Act, whether this impairs their capacity to work will depend on how well their workplace adapts to their needs.

Unfortunately, long term conditions may not be well known or understood by others and can often affect young people. Stigma and scepticism can dog young people with these invisible conditions, coupled with the severe fatigue that often accompanies them. Young people in this situation may be seen as manipulative, lazy or ‘letting colleagues down’. Once in a flare up, employees often work harder and longer to compensate when they feel their effectiveness is slipping, leading to further flare ups and a deteriorating cycle of sick leave. Ultimately this can lead to staff leaving the workplace or working permanently in roles below their capabilities, as they never feel they can ‘catch up’.

Dr Booth asked a network of lupus sufferers in the UK about the impact on their working lives. Many reported reductions in hours, income and promotion prospects. Ironically, employees tend to be in better health higher up the career ladder, as they are more in control of their hours – if they can achieve the higher status in the first place.

Part of the issue can be due to the binary view many employers take towards illness – you are either well, or you are sick. The Bradford factor, sometimes used by employers to assess the impact of sick leave, can be more punitive towards staff who have repeated short absences rather than fewer, longer leaves. Multiple, short-term absences are a classic pattern for those with long term conditions. The associated stress of the constant threat of disciplinary action or job loss also takes its toll.

Dr Booth outlined some of the huge time and energy wasters that come with the territory when you have a long term condition. Just getting hold of medication is a challenge when GPs will only prescribe a month or two’s supply, leading to endless filling of repeat prescriptions for life long conditions – sometimes for multiple drugs. Most of us have struggled with getting a GP appointment at one time or another. Having a long term condition means spending hours arranging (re-arranging), attending and waiting for appointments as well as having countless forms and bureaucracy to deal with. If you cannot use public transport, you spend extra time and money getting around – all the time being given unwanted advice, put downs and judgement from others.

So what should change to make holding down a job less of a challenge for those with so much else on their plate? Dr Booth recommended having specialists on staff in occupational health and HR roles to provide advice on long term conditions to both managers and staff affected. Addressing the attitudes and knowledge of colleagues can help, as well as being aware that someone’s state of health may vary in complexity beyond the sick / better binary state. Discouraging ‘presenteeism’ by colleagues with low grade infections that they may pass on to their more vulnerable team members can also make life less unpredictable.

Another idea gaining currency with some employers is the ‘employee passport’. This is a document where the employee describes what support they need to stay as well as possible, which can be shared at work as needed. It provides continuity when managers change and can reduce resentment from colleagues, who may otherwise only see you when you look ‘well’ and appear unimpaired. These passports could not only benefit those with invisible disabilities or long term conditions, but also staff with caring responsibilities or mental health issues.

For me, openness, honest dialogue and an ability to move beyond making assumptions about the work patterns of colleagues would probably benefit most of us.