Biases entrenched in AI hiring software mean many well-qualified women may be falling at the first hurdle, according to tech diversity non-profit FrauenLoop
* German non-profit offers tech training courses, CV tips
* Automated hiring seen entrenching long-standing biases
* Migrants, ethnic minority women face additional barriers
By Umberto Bacchi
June 14 (Thomson Reuters Foundation) - If you're a woman from an ethnic minority group hoping to land a job in Germany's tech sector, you might want to rethink your CV.
As firms use Artificial Intelligence (AI) software to filter job applications, the mention of a foreign school or maternity leave could trigger automatic rejection, according to FrauenLoop, a non-profit aiming to boost diversity in tech.
While discrimination in the labour market is nothing new, the increasing use of algorithms in hiring means many well-qualified women may be falling at the first hurdle, said FrauenLoop founder Nakeema Stefflbauer.
"If you don't know that you need to have a certain concentration of keywords in your CV and to keep off potential trigger words like 'married' then you might simply be applying and applying and not understanding why you're not getting interviews," Stefflbauer said.
"You may not even be among the candidates that's even being seen by humans. If an algorithm decides that none of the successful candidates the company has hired for a technical role ever took maternity leave, then that's a disqualifying factor."
Standard CV formats in Germany usually include personal details such as age, nationality and marital status that are seldom listed in other countries.
Stefflbauer, who is American and has a long career in software development, founded FrauenLoop in 2016 with the aim of making tech companies more diverse and their products less discriminatory.
From AI recruiting engines favouring men over women to facial recognition software struggling to identify people of colour, campaigners have warned that new technologies often reflect societal bias.
This tends to happen because women and people from ethnic minorities are rarely involved in the development of such products, said Stefflbauer, whose parents are from Jamaica and Panama.
"If we're not represented in a productive and proactive role in the testing, creation and roll-out of software, we're either going to be ignored, or we may even be harmed by it," she said by phone.
DISPARITIES
In Germany, less than 17% of tech specialists are women according to the German internet industry association Eco.
While data on ethnic diversity is scarce, a 2019 survey of 1,200 European tech founders by venture capital firm Atomico found 84% identified as White or Caucasian and less than 1% as Black, African or Caribbean.
To address such wide disparities, FrauenLoop trains women from different ethnic backgrounds who have no prior experience in tech to become web developers, data analysts and AI specialists.
The 12-month evening course is subsidised for refugees and asylum seekers, while others are asked to make a donation, said Stefflbauer.
Sara Alkilani, 30, a Syrian who made a perilous trip to Germany at the height of Europe's migrant crisis in 2015, said she joined FrauenLoop in 2020 after reading about how AI automation would make many jobs redundant.
Having been forced to start a new life after leaving her home behind, the idea that she could one day lose her job to a machine made her feel "very insecure - and (made me) want to get into the tech industry".
The course changed her life, she said. Besides equipping her with skills that led her to land a job as a data analyst, she found a "safe and supportive community" to rely on.
"I feel part of a group that cares about me, who are happy if I get a job ... If I'm confused about things, they offer help and don't expect anything in return and know that we're stronger when we support each other," she said.
'UNBELIEVABLE'
Tech jobs are not hard to come by in Berlin, which is home to a thriving start-up scene employing about 80,000 people.
But getting a foot in the door in an industry that is overwhelmingly male and white can be challenging, hence the focus on CVs and presentation in the training, Stefflbauer said.
Gelavizh Ahmadi, a Kurdish Iranian with dual German citizenship who holds a PhD in physics from the Free University of Berlin, said she received a barrage of rejections when she started looking for work in computer science.
"It was unbelievable because I was even overqualified for some job descriptions while my profile explicitly fit others but I wasn't even getting past the first screening," the 40-year-old said.
The experience left her wondering whether racial bias might be involved - algorithmic or not. "I didn't see any other explanation," she said.
Removing reference to her Iranian nationality from her CV led to more responses, said Ahmadi, who took part in FrauenLoop's training in 2018 and recently got a job as an AI developer for the auto industry.
About 40% of FrauenLoop's graduates have so far found employment, Stefflbauer said.
The share is growing as more tech companies realise the value of diversity, which in turn is allowing women and people from ethnic minorities to choose whether a firm is worth working for, she said.
"Are you going to choose to go work in a company where all of the tech team are men or that has never bothered to hire anybody who is not white?" said Stefflbauer.
"Think of all the unpaid, stressful work to educate others, to manage expectations, to battle micro-aggressions, and to advocate for yourself that this involves. Why would you do that?"
Related stories
Content moderation in conflict zones: What role for big tech?
U.S. small towns take on energy-guzzling bitcoin miners
Right to disconnect gains ground as pandemic brings work home
(Reporting by Umberto Bacchi @UmbertoBacchi; Additional reporting by Valeria Cardi; Editing by Helen Popper. Please credit the Thomson Reuters Foundation, the charitable arm of Thomson Reuters, that covers the lives of people around the world who struggle to live freely or fairly. Visit http://news.trust.org)
Our Standards: The Thomson Reuters Trust Principles.
