Machine learning can give us a clearer picture of research

Artificial intelligence can already describe research. With the right data and expertise, it might evaluate it, say Jonathan Adams and his colleagues.

For 25 years, quantitative research analysis and, inevitably, performance evaluation has focused on research in science and technology. This was because journals in these fields dominated the global databases of publications and citations. The social sciences are only recent entrants into such databases, and the humanities, where the monograph retains priority, have never found them relevant.

The analysis of research, however, is becoming less dependent on citation indices. Machine learning has the potential to make “transformative advances in a range of areas, and create social and economic opportunities”, as the Royal Society noted in a report published in April. It could also provide a more balanced analysis across the research spectrum, reuniting a divided landscape.