AI and data science systems impact every member of the public, yet there isn’t enough diversity in the teams designing the technology itself.
Diversity is essential to the development of a robust data science skills base.
Progression into postgraduate training in data science is as low as 11% for black and 8.4% for disabled students, while the inequitable impact of Covid-19 on women and BAME communities is likely to have slowed progress on diversity and inclusion in all sectors. When we consider the Office for National Statistics’ estimate that over 70% of 1.5 million roles at risk of automation are held by women*, it is clear that urgent intervention is required.
Newcastle University is one of 13 UK universities to have received funding from the Department for Digital, Culture, Media, and Sport (DCMS) and the Office for AI for a project to widen participation in those studying data science and AI degrees. This will support the Government’s ambition to see 2,500 people from these under-represented groups leave university with qualifications in data science and AI.
In response, the University has created 45 MSc scholarships for historically under-represented groups in the field of data and AI.
The first were awarded this academic year and the scheme will run over the next three years. These scholarships provide £10,000 towards course fees or living expenses for individuals who are:
- Black, Asian and minority ethnic (BAME)
- registered disabled
- from POLAR Q1 and Q2 (young participation by area)
- care leavers
- estranged from their families
- from the traveller community
- children from military families
So why the need for this diversity? We need diversity in the teams who are developing and delivering AI systems to ensure that bias is not inadvertently built into algorithms. To prevent this, people with different life experiences need to be designing, building and governing these technologies.
Combatting algorithmic bias
At Newcastle University, throughout our teaching, we develop strong awareness of algorithmic bias and fairness which risks creating exclusionary experiences and discriminatory practices.
These biases may arise, for example, from lack of diversity in training sets for machine learning models. Open source libraries and pre-trained models then perpetuate these biases, with very low friction for them to be incorporated into projects which have a real impact on society.
We teach our students to explicitly check the impact of bias during development and testing.
Nurturing a diverse cohort
We advocate strongly for diverse data science teams with different life experiences governing the development and use of data science and AI. We work closely with community groups promoting diversity in data science roles. We also offer a pre-sessional summer school to help applicants build their confidence to return to Higher Education and enhance their skills prior to study.
Students can gain access to professional mentors comprising senior leaders in data science, many of whom are from underrepresented groups, to share experiences and act as positive role models.
We are committed to providing our students with valuable industry experience throughout the course. We collaborate with the National Innovation Centre for Data to deliver a two-week incubator course on data-driven innovation and encourage students to engage with industry partners in projects throughout the course.
We team up with industry partners to offer paid internships to our students while they are working on their Masters thesis projects. One example of this is with Hexis Lab, which is a research company looking into DNA damage in skin ageing. HexisDNA is a secure app created to enable consumers to provide data on their skincare habits. It also allows them to request a PCR based test to assess the level of skin DNA damage and assess if the reported age of the consumer matches the AI-derived predicted age.
This project to widen participation will go a long way to help us address diversity and ethics in data science education.
At Newcastle University we continue to develop our data science and AI curriculum to meet the needs of the current and future workforce. We are committed to supporting our graduates to develop the confidence and skills required to address future data challenges across many sectors.