Get started
Work

Engineering-Led Breadth: How Far Iranian Artificial Intelligence Research Reaches Beyond Computer Science, Compared with 19 Research Systems, 2012–2019

Year: 2026
Type: preprint
Abstract: Purpose. Artificial intelligence (AI) is increasingly a general-purpose method applied across fields, and countries differ in how far their AI research connects to other disciplines. We measure how far Iranian AI articles reach beyond computer science, in which fields, and whether this differs from comparable research systems. Design and methods. We drew a random sample of 400 English-language journal articles with DOIs, published 2012–2019 and primarily classified in the OpenAlex Artificial Intelligence subfield, for each of 20 research systems (8,000 articles). Using OpenAlex's topic hierarchy (up to three topics per work), we computed the share of articles with at least one non-computer-science topic, the mean number of fields, the entropy of the field distribution and an applied-field share, tested Iran against an emerging-system pool (11 countries), a core pool (7 countries) and each country (Fisher tests with Holm correction, bootstrap intervals), and checked robustness by leaving each pool country out. Findings. 67.3% of Iranian AI articles (95% CI 62.5–72.0) carry a non-computer-science topic, third of 20 systems and 9.3 percentage points above the emerging pool (95% CI 4.5–14.0; 7.3–11.1 when any one pool country is left out). The excess is a cluster effect, not an Iranian peculiarity: Iran is not distinguishable from Türkiye (+2.0), Egypt (+3.0) or Russia (−1.5), and South Africa is higher. It is engineering-led: 28.2% of Iranian articles have an engineering topic against 16.7% in the pool (Holm p < 0.001) and 9.2% against 4.7% for environmental science (Holm p = 0.003), whereas links to health and life sciences are, if anything, weaker (health-science domain 1.8% against 3.7%, unadjusted p = 0.047). 39.0% of Iranian articles have an applied-field topic (pool 28.0%, core 25.0%). The evenness of the field distribution (entropy) does not differ from the pool (difference 0.02, 95% CI −0.08–0.11), and no time trend distinguishes Iran from the pool (difference in change +2.5 points, 95% CI −7.7 to 12.4). Originality. The paper separates breadth from bridging: Iranian AI research touches a second field more often than that of most comparable systems, but the second field is mostly engineering or environmental science rather than the health, life and social sciences, a distinction that a single interdisciplinarity index would hide.
DOI: 10.5281/zenodo.23448583
Language: English
Open Access status: green
OpenAlex ID: W7222903713
Cites:
Cited by:
Related to: