Ethnic diversity boosts scientific impact
The impact of scientific work is boosted when it features co-authors who have a diverse range of ethnicities. That is according to an analysis of nine million scientific publications by six million authors carried out by researchers based at Khalifa University in Abu Dhabi, United Arab Emirates. The study looked at five types of diversity – ethnicity, gender, discipline, affiliation and academic age – and found that ethnic diversity is the strongest predictor of scientific impact.
The ethnicity of authors in the study was determined using a machine-learning technique that analysed author names. The researchers – Bedoor AlShebli, Talal Rahwan and Wei Lee Woon – looked at two types of ethnic diversity. One type, dubbed “group level”, is the variety among the author list of a paper. The other – “individual level” – is the variety in a researcher’s own set of collaborators. The study found that group level has an even greater effect on scientific impact than individual level. “This matters as it implies that an author’s open-mindedness and inclination to collaborate across ethnic lines is not as important as the presence of co-authors of different ethnicities on a paper,” the study says.
AlShebli, Rahwan and Woon say in their study that they were surprised by the findings because other forms of diversity, such as affiliation, are thought to be more related to technical competence. But it turns out that bringing together people from different cultures and social perspectives could have more of a pay-off than just an ethical one.
One limitation of the new study, however, is the method used to determine ethnicity. Although the researchers used a machine-learning technique with a large database of names to classify authors’ ethnicity, this could still result in mistakes creeping in. “There are a number of problems in the study and essential information is missing,” says Ludo Waltman, deputy director of the Centre for Science and Technology Studies at Leiden University in the Netherlands. For instance, he points out that the authors focus on the study’s statistical significance but do not consider the size of the effects. Although there are many benefits of using a large sample of papers, one downside, he notes, is that it is likely to almost always yield statistically significant results.
Dalmeet Singh Chawla