SRAD community updates

Systematic Identification of Tortured Phrases

In our July meeting, Alexandre Clausse (University of Toulouse) shared his doctoral research focused on the systematic identification of tortured phrases. Using the Dimensions API full-text search, and a self-built application, Alex is working to uncover additional tortured phrases and detect their prevalence in scientific literature over time.

This work builds on and contributes to the Problematic Paper Screener, developed by Alex’s supervisor, Guillaume Cabanac, which identifies papers containing tortured phrases (more than five, or obviously tortured, phrases).

Simulating citation behavior and cartels with agent-based models

Our June meeting featured George Chacko (Research Associate Professor, Siebel School of Computing and Data Science & Executive Director of Research Analytics, Grainger College of Engineering, at the University of Illinois Urbana-Champaign), who presented his group’s work utilizing Agent-Based Models (ABM) to simulate human citation behaviors and citation cartels in artificial networks. The simulations revealed a fascinating “quenching effect,” showing that the introduction of “superstar” high-fitness nodes actively siphons citation counts away from average papers, operating as a zero-sum game within a scientific domain.

Furthermore, by modeling artificial citation cartels, the data showed that colluding members would achieve a distinct early and cumulative citation advantage.

The meeting concluded with an open invitation for the SRAD community to access and experiment with the shared simulation code on GitHub.

Applying economic complexity models to global science discussions

This month, Yajie Wang presented a fascinating project that applies the Economics Complexity Index (ECI) to measure the complexity of public attention to science on X/Twitter. Analyzing a massive dataset of 54.6 million posts, the research demonstrated that raw post/tweet volume does not inherently equal attention complexity.

The findings sparked a great discussion on how language barriers and platform preferences can create a divide between a country’s scientific knowledge production and its footprint in global digital discussions.

For data access, Kathryn Weber-Boer provided guidance on navigating API limits, advising users to utilize the query_iterative function for up to 50,000 records or to pull researcher information directly from Google BigQuery using Python.

Why static citation counts fail to capture real attention dynamics

April’s session highlighted Jessica Rodrigues Da Silva’s doctoral research on modeling attention dynamics. She argued that public and academic attention is a dynamic, structured phenomenon that is poorly represented by static counts like total citations or likes.

By representing attention as a continuous “trajectory of distributions over time,” her work uses representation learning to map how engagement flows across different attention sources.

During our technical Q&A, we highlighted the limitation of 1.5 million posts from X per month, using the Altmetric APIs. A question was also raised about the research classifications for grants in Dimensions, which Kathryn confirmed are derived from classification of the text of the grant abstract, using a function much like the DSL classify function.

Testing the causal impact of social media on research visibility

In March, Kim Yi Dionne shared her team’s ongoing multi-arm field experiment investigating the causal impact of scientific communication on research impact. To move beyond purely correlational data, the project randomly assigned political science articles to five distinct communication treatments, including Reddit posts, long social media threads, blogs, and short-form videos like TikToks and Reels.

The discussion addressed structural constraints in academia, the immense labor required to create digestible scientific content, and the potential safety risks for researchers online.

Measuring the impact penalty when researchers explore new fields

This month’s meeting featured an insightful presentation by Yian Yin on “The pivot penalty in research,” which explored the dynamics of researchers exploring new areas versus exploiting existing knowledge. The study revealed that larger topic pivots are systematically associated with lower citation impacts, demonstrating a clear “pivot penalty”.

On the technical side, the group tackled data linking challenges; Kathryn Weber-Boer shared a SQL query for Google BigQuery designed to help users effectively link and deduplicate book, monograph, and chapter records between the Altmetric and Dimensions datasets.