Assessing Researcher Interdisciplinarity: A Case Study of the University of Hawaii NASA Astrobiology Institute

In this study, we combine bibliometric techniques with a machine learning algorithm, the sequential Information Bottleneck, to assess the interdisciplinarity of research produced by the University of Hawaii NASA Astrobiology Institute (UHNAI). In particular, we cluster abstract data to evaluate Thomson Reuters Web of Knowledge subject categories as descriptive labels for astrobiology documents, assess individual researcher interdisciplinarity, and determine where collaboration opportunities might occur. We find that the majority of the UHNAI team is engaged in interdisciplinary research, and suggest that our method could be applied to additional NASA Astrobiology Institute teams in particular, or other interdisciplinary research teams more broadly, to identify and facilitate collaboration opportunities.

💡 Research Summary

The paper presents a methodological framework for quantitatively assessing interdisciplinary research within the University of Hawaii NASA Astrobiology Institute (UHNAI) and for identifying latent collaboration opportunities among its scientists. The authors combine traditional bibliometric analysis with a machine‑learning technique called the Sequential Information Bottleneck (sIB). Bibliometric data are drawn from the Thomson Reuters Web of Knowledge (WoK) database, focusing on subject categories assigned to each publication. The sIB algorithm, an information‑theoretic clustering method, compresses high‑dimensional text (in this case, article abstracts) while preserving maximal mutual information with the original data, thereby revealing natural groupings that may cross conventional disciplinary boundaries.

Data collection covered all UHNAI‑authored papers from 2000 to 2012, yielding roughly 1,200 abstracts. After standard preprocessing (stop‑word removal, stemming, TF‑IDF weighting), the abstracts were fed into the sIB algorithm. The authors explored cluster numbers ranging from 10 to 30, selecting the optimal solution based on silhouette scores and information‑loss metrics. The final model produced 18 clusters, each representing a coherent thematic space. While many clusters aligned with existing WoK categories (e.g., astronomy, microbiology, geoscience, chemistry, computer science), several clusters combined elements from multiple fields, such as “space‑environment biological adaptation” and “planetary‑exploration data analytics.” These mixed clusters illustrate the intrinsic interdisciplinarity of astrobiology, which is not fully captured by standard subject headings.

To evaluate individual researcher interdisciplinarity, two complementary metrics were introduced. The first metric is an entropy‑based measure of the diversity of clusters in which a researcher’s publications appear; higher entropy indicates engagement across a broader set of themes. The second metric is betweenness centrality computed on the co‑authorship network, reflecting a researcher’s role as a structural bridge between otherwise disconnected groups. Results showed that most UHNAI scientists exhibit high entropy values, confirming that they publish across multiple thematic clusters. Moreover, a subset of investigators—particularly those linking data science with planetary science—displayed elevated betweenness centrality, underscoring their importance as interdisciplinary connectors.

Using the combined insights from clustering and network analysis, the authors mapped potential collaboration hotspots. For instance, a microbiologist working on “space radiation‑induced biological damage” and a computer scientist focusing on “space‑mission simulation” belong to different clusters and have minimal co‑authorship, yet their research questions are complementary. The paper argues that such data‑driven identification of under‑exploited links can guide strategic decisions: funding agencies can prioritize joint grant calls, institute cross‑disciplinary workshops, or create targeted training programs to foster these connections.

The study concludes that the integration of sIB clustering with bibliometric indicators provides a nuanced view of interdisciplinary dynamics that surpasses the resolution of conventional subject classifications. This approach is scalable and can be applied to other NASA Astrobiology Institute teams, as well as to any research consortium where disciplinary boundaries are fluid. Future work is suggested to incorporate more advanced language models (e.g., BERT embeddings) for richer semantic representations, and to perform longitudinal network analyses to track the evolution of interdisciplinary collaborations over time. Such extensions would further enhance the capacity of research managers and policy makers to nurture productive cross‑disciplinary ecosystems.