Finding the research that could improve public decisions

During his PhD at UCL, supported by Elsevier, Basil Mahfouz studied why some research reaches policy while closely related work receives little attention. His analysis found that governments often relied on narrow sections of the available literature. Author networks and visibility sometimes had more influence on policy uptake than conventional indicators of academic quality.

Mahfouz used semantic matching and machine learning to narrow the search. Starting with papers cited in policy, the models identified uncited papers covering closely similar ground. Comparing the two helped reveal relevant evidence sitting outside the networks and sources policymakers already knew.

Access to a broad, structured research corpus allows models to compare papers through their content, authorship and connections. Language models can then narrow a vast evidence base to a set that specialists can review.

Expert insight remains essential at the point of decision. Researchers and policymakers must assess the quality of the evidence, its relevance to the population involved and the uncertainty around its findings.

It's also noted that predictive systems learn from past patterns, including previous gaps and biases in evidence use. Their results require scrutiny rather than automatic acceptance.