published Monday 11 February 2019
A new Artificial Intelligence system which examined the role of men and women in the British workplace by reading every website originating from the UK has found that there is still huge gender bias.
The analysis was created by a team of AI engineers and data scientists at the UK start-up Glass A.I, whose technology read and interpreted every website of the internet from the UK and compiled information on employment practices.
The study, published this month by the Royal Statistical Society, reveals stark differences between gender work-roles, and huge segregation in some sectors of the economy.
It found that around 95% of receptionists, legal secretaries and care assistants are female in the UK while 85% of investment bankers are male, and the creative industries remain overwhelmingly male dominated.
Despite being almost as equally likely to participate in the work force – men remain far more likely than women to be in leadership roles, across all sectors.
82% of all CEOs, 92% of Chairpersons and 73% of directors are male, which confirms statistics already highlighted by the ONS (Office for National Statistics) research.
Of 108 economic sectors examined, 87% are biased towards men, with investment banking being 85% male, across all roles.
Civil Engineering, Oil and Gas remain 80% male.
Creative industries such as Media, Music, Internet and Photography also remain heavily male biased.
The study does reveal some female dominated sectors too – Veterinary science is 78% female, and Primary and Secondary Education is 71% female.
Data Scientist Ana-Maria Huluba who ran this study says:
“It is well known that there is a male bias in the board room, what has not been appreciated is the sheer scale of bias across ordinary jobs, and the level of gender segregation between sectors”
“What makes the study even more interesting is that men and women actually appear in almost equal numbers on the web in total, with 51% male, and 49% female – which matches the ONS numbers for gender in the workplace. And yet beneath this we get this massive segregation of roles and appearance in different economic sectors. This is a complex pattern that is supportive and yet goes beyond traditional stereotypes of activity”,
The Glass AI crawler was set to read the entire .UK top-level domain, for both public and private organisations. Sites were analysed if they were written in English, had a UK physical address, had some description of the organisation that the AI could recognise, and had people (either through team pages, biographies, or roles or descriptions).
Glass have built technology to monitor the entire internet for economic and social science analysis. The AI categorises content and builds summaries of activities, networks of companies and people, acting like a vast automated LinkedIn, so as to reveal economic and social trends.
This technology has already been used by UK government analysts to map the UK AI ecosystem, as well as to produce economic studies for drones, virtual reality and other emerging sectors of the economy. In this latest study, the AI assessed gender roles within UK businesses as represented online.
The study has been covered this month by the Royal Statistical Society’s Significance magazine and the full paper was recently published in Heliyon, an academic journal by Elsevier.
This unique study opens the door for more AI-based research using the internet as a live mirror of society and points to new ways for monitoring complex issues, as well as tracking the policy initiatives intended to tackle them.
Glass.AI is a machine-learning start-up co-founded by AI entrepreneur Dr Jason Kingdon, who previously led the UCL Intelligent System Lab spin-out Searchspace and was an early backer of FTSE listed AI Automation company Blue Prism. Glass AI is backed by Kingdon and the IP Group, and is led by co-founders Sergi Martorell and Dr Iain Mclaren with a team of AI engineers. “Glass.AI is a unique experiment in how AI can accelerate core research and bring a new perspective to complex social and economic analysis, in effect using AI to study ourselves. It opens an exciting as well as challenging future,” says Martorell.
Glass has developed a new system for deriving large scale social and economic insights from the web and other sources, by teaching an AI to understand written language. Their latest work, the UK Online Gender Report, has been published in the Heliyon journal by Elsevier, and it is the first ever study of gender representation in UK business space from online data using AI. The report covers some 2.3m people in 150,000 companies across a hundred different business sectors. Glass will make this interpreted data from over 200m .UK web pages available to both researchers and the public.
Full academic paper in the Heliyon journal (Volume 4, Issue 12): here