12,445 papers · continuously updated · last export: 10 Aug 2026livingmeta.ai
← Browse all papers
AI evidence extraction

Large Language Models in Academia: Boosting Productivity but Reinforcing Inequality

Journal of the Association for Information Systems · 2025

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10
Relevance
1/4
Quality (LMQS)
E
Evidence
0
Citations
0.00
FWCI

Methodology & findings

Study design

Observational study using publication data analysis.

Sample

N = 218723, 3 groups

Primary method

Difference-in-differences analysis, generalized synthetic control analysis. Software and specific statistical packages not specified in abstract.

Main result

The study found that "publication rates rose by about 8%, with growth accelerating to 3.2% in 2023 and 12.8% in 2024" following the introduction of LLMs. Additionally, "Junior scholars benefited more than seniors, with the productivity gain declining roughly 1% per year of experience," while "native English-speaking (NES) researchers published more than their non-native English-speaking (NNES) peers, widening linguistic disparities."

Reports effect sizes.

Research paradigm

Positivist/empiricist

Author conclusions

The authors conclude that "LLMs boost scholarly productivity and lower barriers for early-career researchers, while also reinforcing inequities rooted in language proficiency."

Risk of bias

Selection bias: Sample limited to computer science scholars from top 194 U.S. universities only; Confounding: Unable to establish causal mechanisms distinguishing LLM adoption from other contemporary productivity changes; Measurement bias: Publication rates may not fully capture research quality or actual productivity; Geographic/institutional bias: Limited to U.S. institutions; Selection bias: Study limited to computer science scholars from top 194 U.S. universities; Temporal confounding: Publication rate changes may be attributable to factors other than LLM introduction; Missing confounders: No control for changes in research funding, institutional policies, or other technological advances; Language proficiency proxy: Native English-speaking status may not perfectly capture LLM benefits related to language; Selection bias: Analysis limited to computer science scholars from 194 top U.S. universities, which may not represent all academic disciplines or institutions; Confounding variables: Multiple factors other than LLM introduction could affect publication rates (funding changes, career stage effects, pandemic effects on 2019-2024 period); Language classification bias: Categorization of researchers as native vs. non-native English speakers may not capture variation in language proficiency or discipline-specific language barriers; Temporal confounding: The 2019-2024 period includes COVID-19 pandemic, which may have differentially affected publication patterns

Open questions raised

  • Not explicitly stated in abstract
Data: not_statedCode: not_statedExtracted from: pdfAgreement 62%

Explore related topics

Productivity EffectsResearch ProductivityEmpirical Study

Related papers

← Browse all papers