//

When scientific ideas fail to diffuse

Research shows that limited diffusion of knowledge predicts abrupt collapses in the use of scientific ideas, like bursting market bubbles.

Start Reading

When scientific ideas fail to diffuse across diverse audiences, their influence often diminishes rapidly. An analysis of extensive biomedical publication data, employing representational learning methods, reveals that limited dissemination of such ideas can result in sudden and drastic declines in their attention —comparable to the collapse of market bubbles. This phenomenon underscores the critical role of diverse perspectives in fostering sustained scientific advancement and innovation.

Why it matters

The growth and influence of scientific ideas are not always guaranteed, as even those with early promise can experience abrupt declines in relevance and usage. Such declines, similar to “epistemic bubbles”, occur when research gains initial prominence but remains confined to narrow scientific communities and limited topics. While some of these collapses may signal a natural correction in scientific attention, others can represent wasted resources, divert focus, and delay meaningful progress.

The concept of knowledge diffusion, or the spread of ideas across scientific and social networks, offers a counterpoint to these bubbles. Ideas that fail to diffuse across diverse audiences are more prone to collapse, whereas those with wider dissemination undergo broader scrutiny, fostering greater validation and longevity. Restricted diffusion often inflates perceived value while concealing challenges such as replicability, clinical relevance, or general applicability. Understanding these dynamics offers valuable insights into strengthening the sustainability and impact of scientific research.

The framework explained

Biomedical research papers were analysed using two high-dimensional “spaces”:

  • Scientific space: Represents topic areas based on combinations of Medical Subject Heading (MeSH) keywords.
  • Social space: Captures collaboration networks among researchers, extending beyond direct co-authorship.

Knowledge diffusion was assessed by examining how ideas propagated across these spaces. This approach was illustrated by comparing the diffusion patterns of two prominent, highly cited research papers (Figure 1, credit: Authors):

  1. Cardiac stem cell research: Initially hailed for its potential in heart regeneration, this work faced significant replication challenges, leading to its decline. Limited diffusion revealed the presence of a social-epistemic bubble. Despite early enthusiasm, the lack of broader engagement constrained the field’s robustness, resulting in a dramatic collapse.
  2. Cancer immunotherapy: Research that led to groundbreaking advancements in cancer treatment, culminating in the 2018 Nobel Prize for a key contributor, demonstrated sustained diffusion. This broad dissemination ensured its enduring impact and highlighted the benefits of widespread scrutiny and validation.
Figure 1. The left side (A) shows a hypothetical case where research gets massive attention early on in terms of citation but quickly loses it by failing to diffuse broadly, which illustrates a bubble that bursts. The right panel (B) is a contrasting case that performs like a late bloomer.
Credit. Author

Signature of bubbles and collapses

An analysis spanning nearly 30,000 biomedical subfields revealed key characteristics linked to the formation and collapse of scientific “epistemic bubbles.”

  • Limited diffusion predicts collapse: Subfields constrained within narrow scientific or collaboration networks were more likely to experience abrupt declines in citations. Ideas that remain within insular circles tend to be overvalued and insufficiently tested outside their immediate community.
  • Elite concentration increases risk: Subfields dominated by influential scientists and their collaborators were particularly vulnerable. While such networks can sometimes foster breakthroughs, they often lead to inflated popularity and blind spots, increasing fragility.
  • Clinical translatability mitigates risk: Research with clearer pathways to practical clinical applications was less susceptible to collapses, benefiting from wider applicability and engagement.
  • Impact on researchers’ productivity: Scientists who entered a subfield shortly before its collapse faced long-term setbacks in productivity compared to those who joined earlier. This underscores the hidden costs of scientific bubbles, where early-career researchers risk investing time and resources in unstable areas, ultimately hindering their career trajectory.

The lag in funding dynamics

An analysis of funding patterns revealed a significant misalignment with scientific trends. Peaks in funding were observed to precede collapses in subfields by approximately seven years. Even after these fields lost broader relevance, substantial financial resources continued to flow into them.

This lag between funding and the evolving scientific landscape highlights inefficiencies in allocation strategies, emphasising the need for more adaptive approaches to ensure resources are directed towards areas with sustained relevance and potential for impact.

Implications

This research offers a practical framework to refine how biomedical science is assessed and supported. Funding agencies and policymakers can use these insights to evaluate knowledge diffusion and diversify investments. By embracing varied perspectives within the scientific community, decision-makers can minimise the risks associated with overhyping trends that fail to deliver long-term progress.

The framework extends beyond biomedical research, applying to fields such as technology, education, and public policy. It can also address issues like misinformation on social media, where artificially amplified ideas often mimic the dynamics of epistemic bubbles. Analysing diffusion patterns helps identify robust ideas, improving the reliability of both digital and scientific information systems.

Epistemic bubbles, however, are not inherently detrimental. Focused networks can nurture high-risk, high-reward concepts, but this work highlights that their ultimate success depends on broader validation and impact through diffusion.

Expanding this approach to additional domains can deepen understanding of how ideas evolve, spread, and sustain progress. Identifying patterns of resilient knowledge diffusion will enhance innovation, optimise resource distribution, and strengthen the foundations of discovery across disciplines.

🔬🧫🧪🔍🤓👩‍🔬🦠🔭📚

Journal reference

Kang, D., Danziger, R. S., Rehman, J., & Evans, J. A. (2024). Limited diffusion of scientific knowledge forecasts collapse. Nature Human Behaviour, 1-9. https://doi.org/10.1038/s41562-024-02041-0

Tags:

Donghyun Kang is a Postdoctoral Scholar at the Knowledge Lab at the University of Chicago. His research focuses on the emergence, diffusion, and evolution of ideas and practices in science, technology, and civic domains. By combining computational methods such as natural language processing, statistical modelling, and network embedding techniques with theoretical insights from the sociology of knowledge and social studies of science, Kang aims to advance our understanding of the social conditions and processes that shape the development of science and human knowledge.

View all

James Evans is the Max Palevsky Professor of Sociology and Data Science, Director of the Knowledge Lab, and Founding Faculty Director of Computational Social Science at the University of Chicago. His research uses large-scale data, machine learning, and generative models to study how human and machine collectives think and what they know, focusing on the emergence of ideas, reasoning patterns, and processes of attention, communication, and agreement. Supported by the NSF, NIH, and DOD, Evans’ work is published in Nature, Science, and PNAS, and covered by The New Yorker, The New York Times, The Economist, and more.

View all

You might be interested in