Policy ideas and evidence are buried in a vast and fragmented landscape of sources, from government websites and think-tank reports to academic publications.

In the fast-paced world of policymaking, effectively navigating this complexity and fully assessing available information is challenging.

Currently, policymakers lack robust tools to synthesise existing knowledge about what has been tried, what works, and where gaps remain. This limits opportunities for learning and innovation.

Recognising these challenges, this project aims to develop advanced tools and methodologies leveraging AI to efficiently navigate and synthesise extensive policy and academic literature.

Our objective is to enable better-informed policy design, as well as foster innovation and diversity of ideas, by broadening the range of evidence and perspectives available to policymakers.

Through innovative AI-powered tools, we aim to significantly reduce the time and effort needed to explore, synthesise, and apply relevant evidence in policy development – transforming how policymakers engage with evidence.

The vast and fragmented nature of the policy evidence landscape, spread across various sources such as academic publications and third sector reports, makes a comprehensive assessment difficult. This can result in incomplete analyses and missed opportunities for innovative policy design, particularly when considering potential insights from international examples.

There is a need for streamlined, open-source approaches and tools capable of synthesising existing knowledge and enabling policymakers to engage more effectively with evidence.

Currently available AI-powered synthesis tools, despite their growing capabilities, often fall short in terms of openness, transparency, completeness and specificity to policymakers’ needs.

We are developing a digital tool designed for policymakers and researchers to explore, synthesise and engage with policy evidence data. 

Alongside tool development, we will conduct user research and testing to better understand policymaking scenarios where AI-driven solutions can add the greatest value.

Our project builds upon recent foundational work by Nesta’s data science resident Luke McNally, and we will initially focus on three core features: search, synthesis, and simulation.

• Search: This feature will facilitate rapid scanning of research and policy documents, enabling users to efficiently screen for relevance and employ large language models (LLMs) to extract structured information, accelerating systematic reviews.
• Synthesis: Leveraging both traditional data science techniques and generative AI, this feature will produce visualisations and summaries of selected policy evidence.
• Simulation: Users will be able to construct systems diagrams of policy areas and conduct simple simulations of policy interventions by dynamically interacting with system elements

View a demo of an early prototype version of this tool (called LitSynth) below.

With recent advances in generative AI significantly enhancing evidence synthesis capabilities, this field is experiencing rapid progress.

Initiatives such as the UK Research and Innovation’s programme for transforming global evidence and ARIA’s Collective Intelligence Engine underscore the growing interest and investment in AI-driven policy innovation. Research communities like ai@cam are actively exploring optimal ways to leverage AI technologies. 

We are keen to work in the open, learn from best practice and share our learnings. If this project aligns with your interests or expertise, please get in touch with the Mission Discovery team!