Dr Paulina Schenk, University College London

Behavioural science–based artificial intelligence (AI) tools have the potential to accelerate developing evidence-based interventions that target behaviours to improve health and other important policy outcomes. One such tool, evorAI, analyses publicly available text (e.g. blogs and social media posts) using established behavioural science frameworks to generate structured summaries of factors influencing behaviour. These insights could be used to support the design of public health communication campaigns. 

This project will explore how behavioural science AI tools such as evorAI can support developing inclusive behaviour change communications and be integrated into routine UK public health workflows. The work will involve in three stages. First, a diverse Patient and Public Involvement and Engagement (PPIE) panel will be established, including six people with lived experience of non-communicable diseases. Through a workshop, the panel will reflect on existing health behaviour campaigns, helping identify what makes them feel inclusive and where certain groups might be overlooked or misrepresented. Second, ten public health officials will interact with evorAI as a prototype tool during “think-aloud” interviews, providing real-time feedback on usability, relevance, fit with their day-to-day work and challenges to developing inclusive campaigns. Finally, in co-design session informed by insights from Stages 1 and 2, the ten public health officials will highlight tool features that support integration into the everyday public health workflows, and identify potential processes for bias detection and mitigation. 

The project will produce open resources and academic outputs, providing guidance for more inclusive AI tools in public health. The protocol for this research is available on the Open Science Framework. 

--

Note: This application was submitted and peer reviewed under the title: "Embedding Equity in Behavioural Science AI Tools: Co-Design with UK Public Health Teams"

Deliverables/Outputs

(1) Feedback to guide the co-creation workshop

(2) infographics summarising discussions for use in future funding applications. 

Insights will focus on inclusivity and bias detection, shaping practitioner workshop priorities. 

The co-creation workshop will result in a peer-reviewed publication outlining design specifications and inclusive principles, alongside conference presentations (e.g., UCL Centre for Behaviour Change Conference) to share findings with academic and practitioner audiences.  Both phases will contribute to establishing a practitioner–researcher–public network for future tool development and testing.  All publicly shareable materials will be stored in an open repository (e.g., OSF), extending the project’s impact beyond immediate partners. 

Outcomes 

Embedding impact from project inception, we will ensure findings directly inform the next generation of AI behavioural science tools for public health. Outputs will guide both the refinement of evorAI and broader sector learning on inclusive AI design. The practitioner–researcher–public network established will sustain collaboration, feedback, and knowledge exchange beyond the life of the project. Ultimately, this work will support the creation of public health communication campaigns that are evidence-based, culturally sensitive, and equitable - contributing to improved health outcomes in communities across the UK.