A lack of diversity is stifling STEM sectors. Evidence suggests that women, certain ethnic minorities, people with disabilities and those from lower socioeconomic backgrounds are underrepresented in STEM-related education, training and employment.
In January 2022, the Turing responded to the Diversity in STEM inquiry that’s being conducted by the House of Commons Science and Technology Committee to investigate the extent of underrepresentation in the STEM workforce, and how policy makers, funding bodies, industry and academia can work to address it.
We believe that high-quality research and innovation in STEM requires diversity. This is doubly true in the data science and AI community, where diverse research teams are needed to help us better understand and challenge biases in data and algorithms, so we can ensure that the outputs don’t perpetuate existing societal inequalities. Improving representation will help researchers to develop data-driven technologies that benefit all.
The Turing’s response to the inquiry summarised the current state of diversity in data science and AI. Perhaps ironically for these data-centric fields, there is a notable lack of diversity data for data science and AI across academia and industry. The Turing’s view is similar to that of the Royal Society: this data gap is a barrier to addressing underrepresentation. We need more universities, research institutes and organisations to publish their data on researchers in these fields.
At the Turing, we’ve just published our own diversity data in our first EDI Annual Report, and will continue to do so on an annual basis. We’re also gathering key evidence through our ‘Women in data science and AI’ project, which last year published its report ‘Where are the women? Mapping the gender job gap in AI’. This presented a new, curated dataset to map women’s participation in the data science and AI workforce, and identified persistent structural gender inequality, with women more likely to occupy jobs associated with lower status and pay.
The solutions to underrepresentation in STEM are as complex as the causes, and will require coordinated action across government, funding bodies, industry, academia and educational bodies. But every institute also has its own responsibilities. Our work at the Turing is just beginning: in September 2021, we launched our first EDI strategy and accompanying action plan, which set out what we want to achieve in this area, and how we’ll go about it. A key priority is to rectify underrepresentation in our community. We can’t yet say for certain what works, but we are committed to trialling, reviewing, and trialling again.
In our response to the inquiry, we also offered several broader suggestions for how diversity in STEM might be improved:
- Translate undergraduate diversity to postgraduate and early career research communities. While there is evidence that the diversity of the STEM undergraduate community has improved significantly over the last decade, some of these changes have not been translated further along the pipeline of talent into academia and industry. The following may all encourage a diversification of PhD candidates: targeted support through exposure to information about career options; mentoring programmes during the later stages of undergraduate and master’s degrees; and financial packages.
- Consider the positive action provisions contained within the Equality Act 2010. These provisions are potentially underutilised by funding bodies, industry and academia. While they are increasingly being used for targeted schemes, there remains a hesitancy to fully use them to address inequality due to fear of misinterpreting the guidelines and inadvertently creating discriminatory practices. Further guidance or review of how these provisions can be successfully employed would be welcome.
- Develop consistent research sector reporting on protected characteristics. While the funding councils and some other bodies do report this data, a sector-wide approach, similar to that for higher education (as managed by the Higher Education Statistics Agency), would allow far greater benchmarking, analysis and accountability. As mentioned above, we also need improved publishing of diversity data across the data science and AI community.
Top image: Robert Katzki / Unsplash