Taking forward the Scottish Government’s Equally Safe strategy, Equally Safe Online (ESO) is an interdisciplinary, collaborative project where social science meets Machine Learning and NLP. Experts in safeguarding, violence against women and girls (VAWG), and digital education from the University of Edinburgh School of Informatics, School of Social and Political Science (SPS), and the School of Education are working with computer scientists at Heriot-Watt University (HW) to build a user bot for prevention, intervention, and support in online gender-based violence (OGBV), with the project scope being discourse-based OGBV on social media platforms.

Our research methods entail intersectional participatory ‘grassroots’ design ‘bottom-up’ with the third sector partners and victim/survivor stakeholders. Co-creation focus groups, both online and in-person, were conducted with local and national feminist charities, including Rape Crisis, Women’s Aid, SafeLives, EmilyTest, Amina (the Muslim Women’s Resource Centre) and Glitch (the UK’s only OGBV charity) as well as Children's Rights charities such as YoungScot. Sessions with young people (YP) included general population mixed gender groups and YP not in education, as well as Rape Crisis specialist YP workers, teachers, and youth workers.

To create a programming framework, ESO’s first work package (WP1) co-created a ‘taxonomy’ – or classification system – for OGBV by asking the question ‘what does OGBV look like?’ and considering existing law and policy, including the Online Safety Act. This is a first for the computational field, as an OGBV taxonomy does not exist. The taxonomical approach allows for AI solutions to detect GBV to empower internet users to filter instances of hate and abuse as well as permit the production of ‘counter-speech’, generating responses such as fact-checking, signposting, and bystander help.

Considering high levels of youth victimisation, our second work package (WP2) invited young people (YP) to participate in the development of the taxonomy and advise on counterspeech and OGBV typologies. Using creative materials, this data-collection prioritised young GBV survivors including transgender, rural/island residents, and minority ethnic participants.

Focus groups took place over 1.5 years, facilitating challenging discourse across locations and audiences in Scotland (and the wider UK), from children to experienced practitioners coming from a variety of occupational roles, mainly frontline service workers and policy. Fieldwork discussions included gathering testimonies on what works to effectively respond to OGBV, and providing insights from those particularly at risk and with direct lived experience. To make our technical project accessible for all, we produced graphics with Dog and Fox Ltd, who specialise in equalities, to use in our participatory sessions.

The project is now at the dissemination stage, building our user bot, writing publications on areas like methods and ethics, and applying for art community outreach funding. We are undertaking a range of exercises to access datasets, train our models with real-world examples, and user-test with annotators. ESO will soon be presenting to Ministers at the Scot Gov Policy Forum, and we are in talks with partners to produce educational materials. The project has received letters of support from Ofcom, Members of Scottish Parliament (MSPs), and the Equality, Inclusion and Human Rights Directorate.