MIT CSAIL researcher quantifying how AI reshapes science: adoption economics and AI's cost–performance frontier
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-15) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.
Research profile
Output—
Not yet scored: Evolve lacks sufficient publication-volume data for this axis. It would measure total and per-year output — papers, preprints and venue spread — but the current snapshot covers only the two reviewed 2026 arXiv papers.
Impact—
Not yet scored: no citation counts, h-index or uptake metrics are present in this snapshot. This axis would measure citation volume and velocity, including uptake of the cost–performance frontier claims by later meta-science work.
Focus0/100
The focus axis scores 0 on the 0–100 scale, and the raw field distribution explains why: work is split exactly evenly — 0.5 meta-science / AI for science and 0.5 economics of science / science of science. Both catalogued 2026 papers occupy this same intersection, so the pattern reads as a deliberate two-field programme rather than drift, but no single dominant subfield has yet accumulated enough record mass to register concentration.
Collaboration61/100
Collaboration scores 61. The record shows 20 distinct coauthors and 12 recurring partners, among them Gabriel Manso, Emma Fu, Mihai Codreanu, Alex Imas, Juan Mateos-Garcia, Joseph Emmens, Evalyne Muiruri, Arthur Turrell, Julian Jacobs, Atoosa Kasirzadeh, Ana Trisovic and Yiyuan Chen. Both 2026 papers are multi-authored team efforts, with Thompson in a middle position on the Google-led telemetry study and in the last position on the frontier-assessment paper — a pattern of joining large empirical teams while holding the closing, synthesis-oriented slot.
Activity—
Not yet scored: the snapshot contains only two papers, both dated 2026, so publishing cadence cannot be established. This axis would track papers per year and recency of contributions over time.
Reproducibility38/100
Reproducibility scores 38, the mean readiness across 2 assessed papers. 'AI in Science: Early Insights' rates low: its evidence rests on proprietary Gemini usage logs that third parties cannot readily re-obtain. 'The AI-Enabled Scientific Frontier' rates medium: its head-to-head comparisons are partially checkable, though Evolve notes curation and metric caveats. Releasing benchmarks, prompts and comparison protocols would be the clearest lever for lifting this axis.
Research summary
Neil Thompson is a researcher at MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) whose work sits at the junction of two fields, split evenly between meta-science / AI for science and economics of science / science of science. His 2026 catalogued output addresses one question: how, and how unevenly, artificial intelligence is reshaping scientific practice. 'The AI-Enabled Scientific Frontier,' where he is the last author, claims AI reshapes science unevenly, reporting head-to-head comparisons in which AI beats traditional statistics at roughly 10× the cost and rivals scientific simulation about 614× cheaper, with curation and metric caveats tempering the evidence. 'AI in Science: Early Insights,' a Google-led study with Thompson as a middle author, analyzes Gemini usage logs covering 2,690 specialist models and 637 scientists, finding broad AI adoption and complementarity with human work — rich but descriptive and vendor-tilted, with low reproducibility readiness. The record lists 20 distinct coauthors and 12 recurring partners, with a collaboration score of 61 and a mean reproducibility score of 38 across the two assessed papers. Output, impact and activity axes lack sufficient data and are unscored. This profile is name-resolved from public bibliographic data tied to ORCID 0000-0001-9888-573X.
Collaboration pattern
Collaboration scores 61, with 20 distinct coauthors in the record and 12 recurring partners — including Gabriel Manso, Emma Fu, Mihai Codreanu, Alex Imas, Juan Mateos-Garcia, Joseph Emmens, Evalyne Muiruri, Arthur Turrell, Julian Jacobs, Atoosa Kasirzadeh, Ana Trisovic and Yiyuan Chen. Both 2026 papers are team productions: Thompson takes a middle-author slot on the Google-led telemetry study and the final slot on the frontier-assessment paper, suggesting he joins large empirical collaborations while anchoring the closing, synthesis-oriented position.
Research keywords
Highlights
Telemetry-scale evidence that AI adoption in science is broad and complementary
Placing AI on science's cost–performance frontier
A dual-field programme linking adoption economics to frontier measurement
Thompson's two 2026 papers split evenly between economics of science / science of science and meta-science / AI for science (field distribution 0.5/0.5), pairing large-scale usage telemetry with head-to-head capability-versus-cost comparisons — one agenda asking who uses AI in science and another asking when AI actually pays off.
Timeline
- 2026Co-authors 'AI in Science: Early Insights' (middle position), a Google-led analysis of Gemini usage logs spanning 2,690 specialist models and 637 scientists.
- 2026Senior-authors 'The AI-Enabled Scientific Frontier' (last position), reporting head-to-head results in which AI beats statistics at ~10× cost and rivals simulation ~614× cheaper.
- 2026Accumulates a collaboration network of 20 distinct coauthors, including 12 recurring partners, across the two catalogued arXiv papers.
Papers in the Evolve catalog
Frequent co-authors
About this profile
This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.
- Bibliometric dimension scores are computed from public bibliometric data only; coverage of the wider publication record may be incomplete.
- 'Neil Thompson' is a relatively common name; this profile is name-resolved from public bibliographic records tied to ORCID 0000-0001-9888-573X and the MIT CSAIL affiliation.
- Output, impact and activity dimensions are unscored because the underlying data is insufficient; no citation counts or h-index are available in this snapshot.
- Evolve catalog coverage is limited to two reviewed 2026 arXiv papers, so the timeline, highlights and reproducibility mean reflect only that reviewed subset.