With the advancing sophistication of AI and automation-based tools, we explore their use for evidence synthesis. This topic has great importance in a world where insurmountable volumes of information present a growing challenge to the considered (and timely) inclusion of robust evidence at all stages of health and care decision-making.
The last decade has seen an increase in the availability and sophistication of artificial intelligence-based tools. And, as the range of complex tasks for which they may be deployed increases, the technology underpinning these tools has been regarded as ‘revolutionary’ (Gates, 2023).
Amongst other developments, rapid growth in tools and methods which can process and analyse large volumes of textual data has led to considerable interest in whether AI and automation may be used to enhance the speed and efficiency of evidence synthesis. As researchers seek to apply AI and automation-based tools to expedite the laborious stages of systematic review, while maintaining rigour.
Given the insurmountable volume of health research articles published every day, tools which can reliably streamline the process of identifying and reviewing evidence for inclusion have immediate appeal. Where time savings and reductions in manual workload spent searching and screening are reported by systematic reviewers and methodologists who have adopted AI-based tools (Khalil et al., 2022). However, we know considerably less about the adoption of AI-based tools for evidence synthesis outside of the systematic review context. This creates uncertainty as to how efficiently these tools might be deployed within rapid or scoping review contexts. Understanding how well the benefits (and drawbacks) reported by researchers generalise to more rapid or pragmatic review methods is crucial to resolving this uncertainty.
To explore this uncertainty, this short blog introduces and frames our briefing note on automated tools for evidence synthesis. The briefing note outlines the context in which our evidence and knowledge team works in greater detail. While sharing examples of seven automation and AI-based tools, we also reflect on the benefits and drawbacks reported for different tools. Seeking to understand how well these tools might apply to the rapid and pragmatic context in which we work.
