‘Science Is Universal—It Knows No Borders’

Fuad Aleskerov

© HSE University

Fuad Aleskerov, Tenured Professor and Director of the International Centre of Decision Choice and Analysis at HSE University, together with his colleagues, has developed methods of network analysis in bibliometrics that have made it possible to identify patterns in the appearance and citation of publications in academic journals, as well as their influence on each other. When one or a number of studies are frequently cited by a wide range of journals, this is an indicator that the research is of high quality. By contrast, extensive cross-citation within a limited group of journals increases the likelihood of identifying a network of predatory publications.

The findings, based on an analysis of almost 40,000 articles on Parkinson’s disease, have been presented in the book Bibliometric Analysis by Network Models: Identifying Trends in Scientific Literature, published by Springer. In an interview with the HSE News Service, Fuad Aleskerov discussed the origins of the book and the new directions of his research.

— How did the idea for the book come about?

— At first, we were not thinking about writing a book. It all began several years ago when one of the Deputy Governors of the Central Bank called me. He said they had analysed the interbank lending market as a network structure and examined it using measures of centrality, obtaining counterintuitive results.

We then proposed centrality indices that take into account both the characteristics of individual nodes and the collective influence of groups of nodes on other nodes. We developed sophisticated indices based on those previously used in game-theoretic models. Later, we managed to devise simpler ones and apply them to the analysis of academic publications. My colleagues and I first published an article in the Journal of the New Economic Association, showing the conditions under which particular node characteristics—where the nodes represented economics journals and the directed links represented citations—could reveal both positive and negative trends.

— Could you explain the basics of network analysis to our readers?

— A network is constructed according to the task at hand. The nodes may represent objects such as journals, articles, or authors. They are connected by links, known as ‘edges,’ which may be directed or undirected. For example, if Journal A cites Journal B 600 times, while Journal B cites Journal A only five times, this forms a network.

Centrality indices identify the nodes that are the most important within a network—the principal transmitters of information. There are more than 400 different centrality indices designed to determine the key nodes in networks.

We were the first to propose models that simultaneously take into account both the characteristics of individual nodes and the collective influence exerted by groups of nodes on a given node.

— Why is this important?

— Let me give you an example. If I borrowed one million roubles from Sber and failed to repay it, the worst that would happen is that German Gref and the bank’s other senior executives would probably stop saying hello to me. Sber itself would not collapse. A small bank, however, could fail under the same circumstances. Or consider another example. Suppose you and I each borrow 500,000 roubles from a small bank. If one of us repays the loan and the other does not, the bank may survive. But if neither of us repays, it is unlikely to remain solvent. We decided to apply a similar principle to academic journals.

© HSE University

— How was the study designed?

— The object of the study was journal citation patterns, while the node characteristics were defined as the proportion of citations relative to the total number of citations received. If the average number of citations is 300, then, for group influence to be considered significant, a journal’s share of citations must be at least 10%, while the node characteristic itself must be no lower than 30.

When we applied this approach to economics journals, the American Economic Review ranked first, which is exactly what one would expect.

However, as we increased the threshold values, the leading positions were taken by journals that frequently cited one another. In economics, the top-ranked journals still proved to be high-quality publications, but in other disciplines the method can also identify networks of predatory journals.

— Why did you decide to analyse medical journals?

— Olga Khutorskaya, our colleague from the Trapeznikov Institute of Control Sciences of the Russian Academy of Sciences, was studying cybernetic approaches to the treatment of Parkinson’s disease. I agreed that it was an important topic, and we decided to examine publications from 2014 to 2021.

— What criteria did you use to select the articles, and how did you analyse them?

— We used the Microsoft Academic database, which unfortunately was later discontinued. It contained around 70,000 articles. My colleagues filtered the dataset by the presence of a DOI, excluding anonymous publications, papers without an English-language version, and articles citing papers that had not yet been published. As a result, the final dataset comprised 39,811 articles.

Two of my undergraduate students, Ksenia Zinovyeva and Anna Stepochkina, carried out an enormous amount of work selecting the articles and analysing the citation data. I had promised them they would be listed as co-authors. The late Vyacheslav Yakuba, who sadly passed away not long ago, was responsible for developing the software.

We analysed several components separately, both by individual year and across the entire period: citations of articles, journals, authors, institutions, and scientific terms.

— What did the analysis reveal?

— It produced some very interesting findings regarding journals, co-authorship, and a number of other indicators. At times, we were strongly tempted to state that we had identified predatory journals citing one another. However, we decided it would be better to leave it to specialists in medicine and scientometrics to determine whether those journals were genuinely high-quality or predatory.

For example, the most prominent papers were those describing the symptoms used to diagnose Parkinson’s disease. These proved to be the most influential publications, yet they were often published in journals with relatively low citation metrics. A similar pattern could be observed for certain institutions and authors.

Another equally important finding concerned scientific terminology. Our analysis clearly showed how the interests of the research community changed even over a relatively short period. At the beginning of the study period, publications focused primarily on surgical approaches, whereas by the end the emphasis had shifted towards genetics.

It could be said that the quality of scientific research has traditionally been judged by the expert community rather than by journal rankings. However, the sheer volume of publications has now made it necessary to develop quantitative methods of evaluation. This represents a fundamentally different perspective and a new approach to bibliometrics and scientometrics.

© HSE University

— How difficult was it to have the manuscript accepted by a Western publisher?

— Once the manuscript was completed, we approached Springer. They were glad to accept it, but the publication process then slowed down. They explained that they were unable to pay the authors’ royalties. When we suggested donating the royalties to a charitable foundation instead, the completed book remained on hold for about a year before finally being published this year. That delay was related to the events of recent years.

— How do you feel about that?

— Quite calm. I always tell my students that while there are scientists of different nationalities, there is no such thing as Russian or American mathematics, physics, or chemistry. Science is universal—it knows no borders. Had Springer refused to publish Russian researchers simply because they were Russian, the global scientific community would have lost access to valuable new knowledge. Fortunately, that did not happen, and international scientific cooperation continues.

— How has the book been received by your international colleagues?

— Very positively. We received many warm congratulations and an abundance of encouraging feedback.

— What are your current research plans?

— We have recently completed a study based on data on seismic activity and movements of the Earth’s crust. These movements are recorded by monitoring stations, and when we incorporated this data into our model, we were surprised by how well it performed.

The resulting models are highly promising. They make it possible to predict, with up to 80% probability, whether an earthquake will occur several hours in advance. These findings deserve to be shared with the international community, particularly with countries such as China, India, Turkey, Brazil, and, of course, Japan, as well as earthquake-prone nations in Southern Europe.

I hope this work will lead to another book. We have already presented the research at a number of major international conferences, where it received very positive feedback from colleagues. If these methods can be put into practical use, they could help prevent thousands of deaths.