Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Oil & Water? Diffusion of AI Within and Across Scientific Fields

Eamon Duede, William Dolan, André Bauer, Ian Foster, Karim Lakhani

TL;DR

This paper tackles whether AI engagement is truly ubiquitous across science or confined to a subset of fields. It leverages a large-scale analysis of the Semantic Scholar Academic Graph (1985–2022), employing a keyword-based classifier to label AI-engaged papers, a Gini-based metric to quantify diffusion across publication venues, and SPECTER2 embeddings to study semantic trajectories relative to Computer Science AI papers. The results show AI-engaged publications rose ~13x across 20 fields and reached ~9% in 2022, with diffusion across venues accelerating within disciplines. However, embedding-based analyses reveal semantic tension: AI-engaged work becomes more CS-like but also diverges from non-AI-engaged work within fields, indicating oil-and-water dynamics where ubiquity grows but cross-field mixing remains imperfect. The study provides a foundational, quantitative view of AI's broad yet heterogeneous integration into science, with implications for policy, funding, and future research on AI's impact on disciplinary cohesion and scholarly practice.

Abstract

This study empirically investigates claims of the increasing ubiquity of artificial intelligence (AI) within roughly 80 million research publications across 20 diverse scientific fields, by examining the change in scholarly engagement with AI from 1985 through 2022. We observe exponential growth, with AI-engaged publications increasing approximately thirteenfold (13x) across all fields, suggesting a dramatic shift from niche to mainstream. Moreover, we provide the first empirical examination of the distribution of AI-engaged publications across publication venues within individual fields, with results that reveal a broadening of AI engagement within disciplines. While this broadening engagement suggests a move toward greater disciplinary integration in every field, increased ubiquity is associated with a semantic tension between AI-engaged research and more traditional disciplinary research. Through an analysis of tens of millions of document embeddings, we observe a complex interplay between AI-engaged and non-AI-engaged research within and across fields, suggesting that increasing ubiquity is something of an oil-and-water phenomenon -- AI-engaged work is spreading out over fields, but not mixing well with non-AI-engaged work.

Oil & Water? Diffusion of AI Within and Across Scientific Fields

TL;DR

This paper tackles whether AI engagement is truly ubiquitous across science or confined to a subset of fields. It leverages a large-scale analysis of the Semantic Scholar Academic Graph (1985–2022), employing a keyword-based classifier to label AI-engaged papers, a Gini-based metric to quantify diffusion across publication venues, and SPECTER2 embeddings to study semantic trajectories relative to Computer Science AI papers. The results show AI-engaged publications rose ~13x across 20 fields and reached ~9% in 2022, with diffusion across venues accelerating within disciplines. However, embedding-based analyses reveal semantic tension: AI-engaged work becomes more CS-like but also diverges from non-AI-engaged work within fields, indicating oil-and-water dynamics where ubiquity grows but cross-field mixing remains imperfect. The study provides a foundational, quantitative view of AI's broad yet heterogeneous integration into science, with implications for policy, funding, and future research on AI's impact on disciplinary cohesion and scholarly practice.

Abstract

This study empirically investigates claims of the increasing ubiquity of artificial intelligence (AI) within roughly 80 million research publications across 20 diverse scientific fields, by examining the change in scholarly engagement with AI from 1985 through 2022. We observe exponential growth, with AI-engaged publications increasing approximately thirteenfold (13x) across all fields, suggesting a dramatic shift from niche to mainstream. Moreover, we provide the first empirical examination of the distribution of AI-engaged publications across publication venues within individual fields, with results that reveal a broadening of AI engagement within disciplines. While this broadening engagement suggests a move toward greater disciplinary integration in every field, increased ubiquity is associated with a semantic tension between AI-engaged research and more traditional disciplinary research. Through an analysis of tens of millions of document embeddings, we observe a complex interplay between AI-engaged and non-AI-engaged research within and across fields, suggesting that increasing ubiquity is something of an oil-and-water phenomenon -- AI-engaged work is spreading out over fields, but not mixing well with non-AI-engaged work.
Paper Structure (13 sections, 4 equations, 6 figures, 4 tables)

This paper contains 13 sections, 4 equations, 6 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Data
  3. Methods and Findings
  4. RQ1: How has the percentage of scientific and scholastic research that engages with AI changed over time?
  5. Methods
  6. Findings
  7. RQ2: To what extent has AI engagement diffused within individual fields?
  8. Methods
  9. Findings
  10. RQ3: How is changing ubiquity of AI engagement associated with changes in the semantic character of fields?
  11. Methods
  12. Findings
  13. Discussion and Concluding Remarks

Figures (6)

  • Figure 1: Change in AI engagement across all fields from 1985 - 2022
  • Figure 2: Change in AI engagement percentage from 1985 - 2023 by field. Inserts tally the total change in percentage of AI-engaged publications for each field.
  • Figure 3: Boxplots representing the overall change in Ubiquity from 1985 to 2023.
  • Figure 4: Ubiquity over time from 1985 - 2022. Inserts tally the total percentage change in Ubiquity for each field.
  • Figure 5: Density of the distribution of semantic similarity between the centroids of all Non-AI-engaged and all AI-engaged papers and the centroids of AI-engaged papers in Computer Science across all fields for the period 1985 - 2023. Values closer to 1 on the x-axis represent higher semantic similarity.
  • ...and 1 more figures