Pith. sign in

REVIEW 4 major objections 4 minor 53 references

Towards Industrial Convergence : Understanding the evolution of scientific norms and practices in the field of AI

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read AI research converges through mixed academic-industrial teams, with industrial habits winning in papers and code.

desk verdict A descriptively rich but causally overreaching study of academic, industrial, and mixed AI teams; the mixed-team convergence result is worth testing properly, but the current analysis does not support the causal claim. read the letter →

arxiv 2505.17945 v1 pith:7IUMQ2U3 submitted 2025-05-23 cs.DL physics.soc-ph

Antoine Houssard This is my paper · ORCID
classification cs.DLphysics.soc-ph
keywords AIresearchacademia-industrycollaborationMode2knowledgeproductionasymmetricalconvergencescientificnormsbibliometricsopen-sourcecoderepositoriesmixedteams
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to test whether AI research is moving away from a purely academic mode of knowledge production toward a hybrid, industry-shaped one. It compares how purely academic, purely industrial, and mixed teams produce two linked artifacts: the scientific paper and the public code repository. It finds that pure academic and pure industrial teams still follow clearly different norms in topic breadth, publication strategy, code complexity, and division of labor. The central result is that teams containing at least one industrial author behave like industrial teams in those choices and succeed more in both arenas, with papers earning more citations and code repositories earning more popularity. The paper concludes that whatever convergence is happening in AI research passes through mixed teams, and interprets that as an asymmetrical convergence that mainly benefits industrial actors.

What carries the argument

The analytical engine is the comparison of two concurrent artifacts, the scientific paper and its associated code repository, across three team types: purely academic, purely industrial, and mixed. The data come from a platform that pairs papers with their official code implementations, and the paper follows each artifact from preprint or release onward to publication and popularity. It combines topical-diversity measures, including Shannon entropy over topics and a rewiring-based z-score for topic-combination typicality, with publication venue and timing analysis, repository documentation and language-stack metrics, a degree-of-authorship score for how concentrated code authorship is, and time-series clustering of citation and popularity accumulation. The mixed-team contrast is the load-bearing comparison: whenever an industrial author is present, the team's choices shift toward the industrial profile.

What would settle it

Re-run the publication-status analysis counting peer-reviewed conference papers, such as NeurIPS, ICML, CVPR, and ACL, as publications; if the industrial publication rate moves close to the academic rate, the 'industry publishes less' conclusion is an artifact of the journal-only definition. A matched comparison of mixed versus purely academic teams on similar topics and resources would likewise test whether the mixed-team success advantage survives selection effects.

Watch

Extended reading notes

Core claim

The central discovery is that the mixed team, rather than the pure academic or pure industrial group, is the site where industrial norms enter academic research and where both artifacts perform best. In the sample, mixed teams publish less and more slowly, choose narrower and more conventional topic combinations, and maintain more complex, better-documented code repositories, patterns that lean industrial, while their papers draw more citations and their repositories draw more popularity than either pure group. The author's own reading is that this supports asymmetrical convergence rather than full Mode 2 unification: institutional-level norms remain distinct, but mixed teams adopt industrial logic, and in exchange they get resources and success.

Load-bearing premise

The load-bearing assumption is that a study counts as 'fully published' only when it appears in a journal, so the reported gap between industrial and academic publication, 4.7% versus 23.2%, depends on excluding the conference venues where most AI research is actually peer-reviewed.

Editorial extensions

If this is right

  • Teams with at least one industrial author align their research questions, publication strategy, and code practices with industrial norms, even when the team is mostly academic.
  • Mixed teams achieve higher average success on both artifacts, so industrial collaboration correlates with broader visibility for papers and code.
  • Pure academic AI research remains distinct enough that the results do not support a complete Mode 2-style unification; convergence is partial and channeled.
  • Industrial authors use publication strategically, favoring fewer venues, higher-impact targets, and longer time to publication, so collaborations change the publication rhythm of academic teammates.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper counts 'fully published' as journal publication only; because AI's main peer-reviewed venues are conferences, a reanalysis that counts conference papers could shrink or eliminate the reported 23.2% versus 4.7% publication gap.
  • The mixed-team success advantage may partly reflect selection, with companies attaching to promising projects, rather than a causal effect of mixing; a matched comparison on topic and resources would test this.
  • If convergence runs through mixed teams, policies meant to preserve academic norms in AI should shape collaboration structure, such as protecting academic partners' choice of artifacts, rather than only funding pure academic teams.
  • The paper notes frequent individual moves between academia and industry but does not track them; following individuals across moves would test whether convergence is carried by people or by team composition.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. This paper asks whether academic AI research is converging on industrial norms and practices. Using Papers with Code entries filtered into four AI fields, author affiliations from OpenAlex/GROBID, and GitHub repository metadata, the authors compare academic-only, industrial-only, and mixed teams on topic diversity, programming-language stacks, publication venues and timing, repository maintenance and labor distribution, and downstream success measured by citations and GitHub stars. They report that industrial and mixed teams focus on fewer topics, use more complex technical stacks, publish in higher-impact venues at lower rates, have larger and more evenly distributed development teams, and attract faster public attention, while mixed teams are successful on both papers and repositories. The paper interprets these patterns as evidence that convergence, if it exists, passes through mixed academic-industrial teams under an "asymmetrical convergence" framework.

Significance. The question is timely and the multi-artifact design, tracking papers and code together, is a real strength. The paper makes good use of externally collected data and several established metrics (rarefied Shannon entropy, Uzzi's atypical-combination z-score, the DOA authorship measure with a pre-existing 0.75 threshold), and the supplementary description of the collection pipeline is helpful. If the mixed-team result were robust, it would make a substantive contribution to the STS and Mode-2 literature on science-industry relations. However, the causal claim in the abstract and the interpretation in Sections 5 and 6 go beyond what the cross-sectional, uncontrolled comparisons can establish, and the journal-only definition of publication is particularly problematic for AI. The empirical regularities may survive a more careful analysis, but the current manuscript does not yet support the stated conclusions.

major comments (4)
  1. [Abstract; §5; §6] The central claim is causal: "the presence of industrials in academic studies leads to practices leaning toward the industrial side, but also to greater success" (Abstract). The evidence consists of cross-sectional group comparisons in Sections 4.1–4.3. These comparisons do not control for selection into mixed collaborations, team size, resources, field, or arXiv year; the paper itself reports that industrial repositories have approximately 29.2 contributors versus 16.9 for academic ones (§4.2). A within-author or matched design, or at minimum covariate-adjusted regression models, is required before "leads to" can be sustained. Without such controls, the results are consistent with selection into collaboration rather than an influence of industrial presence on practices.
  2. [§3; §4.2] The outcome "fully published" is defined solely as publication in a journal: §3 states that the arXiv life-cycle is tracked "to its subsequent publication in a journal". In AI, most peer-reviewed results appear at conferences such as NeurIPS, ICML, and CVPR, which are not journals. The reported gap of 23.2% (academic) versus 4.7% (industrial) in Fig. 4C therefore conflates "not published" with "not published in a journal". This is load-bearing for the conclusion that industrial authors "do not rely on scientific publications" and are "more selective" (§4.2, §5). The analysis should include conference proceedings or provide a justified reason for excluding them.
  3. [§4.2; Fig. 8] Several key comparisons lack adequate statistical support. The mixed-team time-to-publication result is introduced with "ANOVA test P≈0.09" and then described as "confirming the importance of industrial authors"—a p-value of 0.09 does not confirm a difference. Fig. 8 reports cluster compositions (e.g., 32% industrial in the fast-growth cluster vs. 23–24% elsewhere) without confidence intervals or significance tests, and Fig. 2's rarefied entropy estimates are shown without error bars. Please report effect sizes and uncertainty for all main comparisons and avoid interpreting non-significant results as confirmatory.
  4. [§4.2; §7.1] The repository subset used for the DOA and labor-distribution analysis is selected by popularity and activity: the authors take the 100 most-starred repositories with 100 to 2000 commits for each group. This post-hoc filter on success measures is likely to distort the comparison, since it conditions on the very outcomes later interpreted as group differences, and it restricts generalizability to all academic and industrial repositories. The supplementary pipeline (Fig. 9) should report how many repositories survive each filter, and the analysis should include sensitivity checks with different thresholds. Additionally, mixed repositories are omitted from several maintenance metrics (§4.2), which is surprising given the paper's central interest in mixed teams.
minor comments (4)
  1. [§4.2] The text in §4.2 refers to a "Quartile to Quartile plot" while Fig. 7B and the supplementary material use "quantile-quantile"; please standardize and correct the spelling.
  2. [Eq. (1)] Equation (1) is typeset incorrectly: "nX" should be a summation symbol, and the base of the logarithm should be stated explicitly.
  3. [Table 1] The caption "Field of AI F requency" contains a typo, and the table would benefit from a clear column header for the frequency counts.
  4. [§4.1; §4.2] The manuscript alternates between "industrial" and "private" when referring to the same group (e.g., "purely private and mixed teams" in §4.1 vs. "purely industrial" elsewhere); please choose one term for consistency.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the study's comparisons are empirical and externally grounded.

full rationale

The paper's central claim — that mixed academic-industrial teams lean toward industrial practices and achieve greater success in both papers and repositories — is supported by cross-sectional comparisons of externally collected data (Papers with Code, OpenAlex, GitHub). The quantities used (Shannon entropy, topic-pairing z-scores, language prevalence, Gini index, DOA authorship scores, citation counts, GitHub stars, time-series clusters) are defined independently of the paper's conclusions. The DOA weights and the 0.75 authorship threshold are adopted from prior published work (Fritz et al.; Avelino et al.) rather than fitted to force the reported results. No parameter is fitted to a subset of the data and then renamed as a prediction, and no conclusion is defined in terms of its own outcome. The 'fully published' metric is journal-only, which may bias the publication-gap finding as a measurement or construct-validity concern, but it is not circular: the paper does not define 'industrial authors publish less' into existence through the metric, and the criticism is about operationalization rather than derivation-by-definition. The main weakness is inferential — team composition is self-selected and success comparisons lack controls for team size, resources, or field — but confounding and selection are threats to causal validity, not circularity. Citations to prior literature (e.g., Moore and Frickel on asymmetrical convergence) are used interpretively and are not load-bearing in a way that reduces the empirical results to the cited claims. The derivation chain is therefore self-contained with respect to circularity.

Assumptions & free parameters 2 free parameters · 6 assumptions · 0 invented entities

The central claim depends on measurement choices inherited from prior work (DOA3, Uzzi, OpenAlex topics) and on the representativeness of the Papers with Code platform. The paper inherits these instruments without independent validation in the AI context, and it does not release code or data that would allow readers to audit the pipeline.

free parameters (2)
  • k (number of time-series clusters) = 4
    Chosen via the beta-CV heuristic and a visual elbow in Supplementary 7.6; the overrepresentation of industrial/mixed repositories in 'supra-linear' growth clusters depends on this choice.
  • DOA authorship threshold = 0.75
    Adopted from Avelino et al. (2019) to decide whether a contributor is an author of a file; this threshold directly affects the reported author counts (academic mean 16.9, industrial mean 29.2) that support the labor-distribution claim.
assumptions (6)
  • domain assumption Papers with Code is a representative source for AI research that ships code.
    The entire sample is drawn from this platform; if the platform skews toward certain fields or team types, all comparisons inherit that skew (Section 3).
  • domain assumption OpenAlex topic assignments, exactly three per paper, accurately represent the content of each paper.
    Topic diversity and typicality scores are computed from these assignments; errors in topic labeling would directly affect the diversity results (Section 4.1, Equation 1).
  • domain assumption The Uzzi rewiring method, designed for article-journal bipartite networks, transfers to a topic-topic co-occurrence network.
    Used to compute typicality z-scores for topic combinations; no validation of the transfer is provided (Section 4.1).
  • domain assumption GitHub stars and citations are valid proxies for the 'success' of code and papers respectively.
    The central claim that mixed teams are 'more successful' rests on these proxies; no evidence is given that they measure the intended constructs (Section 4.3).
  • domain assumption The DOA3 model and its coefficients, developed by Fritz et al. and Avelino et al., accurately quantify code authorship investment in AI repositories.
    The labor-distribution analysis relies on this metric, including the fixed weights in Equation 3 (Section 4.2).
  • domain assumption The fuzzy matching of institution names against university and Crunchbase lists, followed by manual review, correctly classifies team types.
    The grouping into academic/industrial/mixed depends on this step; misclassification could bias all comparisons (Section 7.1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Towards Industrial Convergence : Understanding the evolution of scientific norms and practices in the field of AI." pith.science (2026). https://pith.science/paper/7IUMQ2U3

@misc{pith2026250517945,
  author       = {Pith},
  title        = {Pith review of: Towards Industrial Convergence : Understanding the evolution of scientific norms and practices in the field of AI},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7IUMQ2U3}},
  note         = {Machine review of arXiv:2505.17945}
}
read the original abstract

In the field of artificial intelligence (AI) research, there seems to be a rapprochement between academics and industrial forces. The aim of this study is to assess whether and to what extent industrial domination in the field as well as the ever more frequent switch between academia and industry resulted in the adoption of industrial norms and practices by academics. Using bibliometric information and data on scientific code, we aimed to understand academic and industrial researchers' practices, the way of choosing, investing, and succeeding across multiple and concurrent artifacts. Our results show that, although both actors write papers and code, their practices and the norms guiding them differ greatly. Nevertheless, it appears that the presence of industrials in academic studies leads to practices leaning toward the industrial side, but also to greater success in both artifacts, suggesting that if convergence is, then it is passing through those mixed teams rather than through pure academic or industrial studies.

Figures

Figures reproduced from arXiv: 2505.17945 by the authors.

Figure 1
Figure 1. Institution type and labels in the dataset A) Top 25 most represented industrial institution in the sample ; B) Top 25 most represented academic institution in the sample ; C) Study repartition within each group : ”Company” represent studies conducted only by industrial actor, ”Public” represents studies only conducted by academic or other non-profit / governmental actors and ”Mixed” studies conducted with researche… view at source ↗
Figure 2
Figure 2. Topical diversity in Academic and Industrial papers : A) Shannon Entropy of journal’s (excluding pre-publication venues) OpenAlex topics, higher value means higher topical diversity. The Shannon entropy is computed for each AI task subgroup and across groups. B) Shannon entropy for papers topic using OpenAlex Topics. C) Topic pairing z-score in academic and industrial paper highlighting the different in topic combin… view at source ↗
Figure 3
Figure 3. Programming language in academic and industrial repositories : A) Frequency of file programming language type for purely academic, mixed and purely industrial repositories. B) Quartile to Quartile plot for the number of programming language within the repositories for purely academic, mixed and purely industrial repositories, each points represents the group quartile and the dashed line the quartile for the entire p… view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Venue choice for academic and industrial papers : A) Density distribution function for the academic specific, industrial specific and common venues (excluding pre-publication venues). The inset shows the reparation of public and private articles within those categories…
Figure 5
Figure 5. Figure 5: Repository presentation metrics: A) The figure shows the frequency/probability of encountering different elements within academic, mixed and industrial repositories. All repositories are considered except for the "install" variable which is filtered for repositories ut…
Figure 6
Figure 6. Figure 6: Repository maintenance by academic and industrial team : A) The figure shows, for each file in industrial and academic repositories, the fraction of contributors considered as "au￾thor" according to the threshold. B) Entropy of line modification for academic and indust…
Figure 7
Figure 7. Figure 7: Artifact performance : A) Box plot for the number of stars and forks for Github repositories linked to purely public, purely private and mixed studies. B) The figure shows a quantile–quantile plot for the purely industrials, purely academics and mixed teams for the num…
Figure 8
Figure 8. Figure 8: Artifact performance over time : A) K-Means clustering of the cumulative time series of stars for a sample of repositories. The figure is showing each cluster with the dashed grey line being the general trend of the cluster and the transparent green line the time serie…
Figure 9
Figure 9. Figure 9: Data collection pipeline. 23 [PITH_FULL_IMAGE:figures/full_fig_p023_9.png]
Figure 10
Figure 10. Figure 10: Programming languages frequency across group at the file level. [PITH_FULL_IMAGE:figures/full_fig_p024_10.png]
Figure 11
Figure 11. Figure 11: Programming languages frequency across group at the file level. [PITH_FULL_IMAGE:figures/full_fig_p024_11.png]
Figure 12
Figure 12. Figure 12: QQ (Quartile to Quartile) plot of the publication venue for Academic and Industrial [PITH_FULL_IMAGE:figures/full_fig_p025_12.png]
Figure 13
Figure 13. Figure 13: Repositories maintenance time : A) Fraction of issues closed after n days in academic and industrial (at least one industrial authors) repositories B) Time between commits in academic and industrial repositories The figure shows conflicting trends, with some maintenan…
Figure 14
Figure 14. Figure 14: Elbow plot for the time series K-Means : A) Elbow plot for the GitHub stars time series. B) Elbow plot for the citation time series 26 [PITH_FULL_IMAGE:figures/full_fig_p026_14.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references · 52 canonical work pages

  1. [1]

    The de-democratization of ai: Deep learning and the compute divide in artificial intelligence research

    Ahmed, N., and Wahed, M. The de-democratization of ai: Deep learning and the compute divide in artificial intelligence research. arXiv preprint arXiv:2010.15581\/ (2020)

    arXiv 2020

  2. [2]

    Ahmed, N., Wahed, M., and Thompson, N. C. The growing influence of industry in ai research. Science 379 , 6635 (Mar. 2023), 884–886

    work page 2023

  3. [3]

    Des logiciels libres au contr \^o le du code: l'industrialisation de l' \'e criture informatique

    Alcaras, G. Des logiciels libres au contr \^o le du code: l'industrialisation de l' \'e criture informatique . PhD thesis, Paris, EHESS, 2022

    work page 2022

  4. [4]

    Avelino, G., Passos, L., Hora, A., and Valente, M. T. Measuring and analyzing code authorship in 1 + 118 open source projects. Science of Computer Programming 176\/ (May 2019), 14–32

    work page 2019

  5. [5]

    Partenaires particuliers : financements sur projet et travail relationnel dans les réseaux de collaboration science-industrie

    Barrier, J. Partenaires particuliers : financements sur projet et travail relationnel dans les réseaux de collaboration science-industrie. Genèses n° 94 , 1 (May 2014), 55–80

    work page 2014

  6. [6]

    A firm scientific community: Industry participation and knowledge diffusion

    Baruffaldi, S., and P\" o ge, F. A firm scientific community: Industry participation and knowledge diffusion. SSRN Electronic Journal\/ (2020)

    work page 2020

  7. [7]

    What’s in a github star? understanding repository starring practices in a social coding platform

    Borges, H., and Tulio Valente, M. What’s in a github star? understanding repository starring practices in a social coding platform. Journal of Systems and Software 146\/ (Dec. 2018), 112–129

    work page 2018

  8. [8]

    u ggemann, M., L\

    Br\" u ggemann, M., L\" o rcher, I., and Walter, S. Post-normal science communication: exploring the blurring boundaries of science and journalism. Journal of Science Communication 19 , 03 (June 2020), A02

    work page 2020

Show all 53 references
  1. [9]

    les brevets sont à peine au rang d’une publication

    Brun, V. "les brevets sont à peine au rang d’une publication" : Projets de valorisation et cycle de crédibilité au cnrs. Revue d’anthropologie des connaissances 17 , 2 (May 2023)

    2023

  2. [10]

    Cellules souches et technoscience : sociologie de l’émergence et de la régulation d’un domaine de recherche biomédicale en france

    Brunet, P., and Dubois, M. Cellules souches et technoscience : sociologie de l’émergence et de la régulation d’un domaine de recherche biomédicale en france. Revue fran c aise de sociologie Vol. 53 , 3 (Sept. 2012), 391–428

    2012

  3. [11]

    La revanche des neurones: L’invention des machines inductives et la controverse de l’intelligence artificielle

    Cardon, D., Cointet, J.-P., and Mazières, A. La revanche des neurones: L’invention des machines inductives et la controverse de l’intelligence artificielle. Réseaux n° 211 , 5 (Nov. 2018), 173–220

    2018

  4. [12]

    The strategic drivers and objectives of communities of practice as vehicles for knowledge management in small and medium enterprises

    du Plessis, M. The strategic drivers and objectives of communities of practice as vehicles for knowledge management in small and medium enterprises. International Journal of Information Management 28 , 1 (Feb. 2008), 61–67

    2008

  5. [13]

    Analyzing the impact of companies on ai research based on publications

    F\" a rber, M., and Tampakis, L. Analyzing the impact of companies on ai research based on publications. Scientometrics 129 , 1 (Nov. 2023), 31–63

    2023

  6. [14]

    Of Timescapes and Knowledgescapes

    Felt, U. Of Timescapes and Knowledgescapes . Oxford University Press, Dec. 2016, p. 129–148

    2016

  7. [15]

    R., Wang, D., Cebrian, M., and Rahwan, I

    Frank, M. R., Wang, D., Cebrian, M., and Rahwan, I. The evolution of citation graphs in artificial intelligence research. Nature Machine Intelligence 1 , 2 (Feb. 2019), 79–85

    2019

  8. [16]

    Frickel, S., and Moore, K. , Eds. The new political sociology of science . Science & Technology in Society. University of Wisconsin Press, Madison, WI, Mar. 2006

    2006

  9. [17]

    C., Murphy-Hill, E., Ou, J., and Hill, E

    Fritz, T., Murphy, G. C., Murphy-Hill, E., Ou, J., and Hill, E. Degree-of-knowledge: Modeling a developer's knowledge of code. ACM Transactions on Software Engineering and Methodology (TOSEM) 23 , 2 (2014), 1--42

    2014

  10. [18]

    C., and Murphy-Hill, E

    Fritz, T., Ou, J., Murphy, G. C., and Murphy-Hill, E. A degree-of-knowledge model to capture source code familiarity. In Proceedings of the 32nd ACM/IEEE International Conference on Software Engineering-Volume 1\/ (2010), pp. 385--394

    2010

  11. [19]

    O., and Ravetz, J

    Funtowicz, S. O., and Ravetz, J. R. Science for the post-normal age. Futures 25 , 7 (Sept. 1993), 739–755

    1993

  12. [20]

    A meso-scale cartography of the ai ecosystem

    Gargiulo, F., Fontaine, S., Dubois, M., and Tubaro, P. A meso-scale cartography of the ai ecosystem. Quantitative Science Studies 4 , 3 (2023), 574–593

    2023

  13. [21]

    Resource democratization: Is compute the binding constraint on ai research? Proceedings of the AAAI Conference on Artificial Intelligence 38 , 18 (Mar

    Gelles, R., Kinoshita, V., Musser, M., and Dunham, J. Resource democratization: Is compute the binding constraint on ai research? Proceedings of the AAAI Conference on Artificial Intelligence 38 , 18 (Mar. 2024), 19840–19848

    2024

  14. [22]

    The new production of knowledge

    Gibbons, M., Limoges, C., Nowotny, H., Schwartzman, S., Scott, P., and Trow, M. The new production of knowledge . SAGE Publications, London, England, July 1994

    1994

  15. [23]

    This ai researcher is trying to ward off a reproducibility crisis

    Gibney, E. This ai researcher is trying to ward off a reproducibility crisis. Nature 577 , 7788 (Dec. 2019), 14–14

    2019

  16. [24]

    Big tech influence over ai research revisited: memetic analysis of attribution of ideas to affiliation, 2023

    Giziński, S., Kaczyńska, P., Ruczyński, H., Wiśnios, E., Pieliński, B., Biecek, P., and Sienkiewicz, J. Big tech influence over ai research revisited: memetic analysis of attribution of ideas to affiliation, 2023

    2023

  17. [25]

    The state of the ml-universe: 10 years of artificial intelligence ; machine learning software development on github

    Gonzalez, D., Zimmermann, T., and Nagappan, N. The state of the ml-universe: 10 years of artificial intelligence ; machine learning software development on github. In Proceedings of the 17th International Conference on Mining Software Repositories\/ (June 2020), MSR ’20, ACM

    2020

  18. [26]

    Science d'ing \'e nieurs et sciences pour l'ing \'e nieur: l'exemple du g \'e nie chimique

    Grossetti, M., and Detrez, C. Science d'ing \'e nieurs et sciences pour l'ing \'e nieur: l'exemple du g \'e nie chimique. Sciences de la soci \'e t \'e : Les cahiers du LERASS\/ (2000), pp--63

    2000

  19. [27]

    Ethical considerations and statistical analysis of industry involvement in machine learning research

    Hagendorff, T., and Meding, K. Ethical considerations and statistical analysis of industry involvement in machine learning research. AI & SOCIETY 38 , 1 (Sept. 2021), 35–45

    2021

  20. [28]

    K., Franssen, T., Scholten, W., and de Rijcke, S

    Hessels, L. K., Franssen, T., Scholten, W., and de Rijcke, S. Variation in valuation: How research groups accumulate credibility in four epistemic cultures. Minerva 57 , 2 (Jan. 2019), 127–149

    2019

  21. [29]

    G., Brusoni, S., and Candelon, F

    Jacobides, M. G., Brusoni, S., and Candelon, F. The evolutionary dynamics of the artificial intelligence ecosystem. Strategy Science 6 , 4 (2021), 412--435

    2021

  22. [30]

    Reimagining Innovation

    Joly, P.-B. Reimagining Innovation . Springer Singapore, 2019, p. 25–45

    2019

  23. [31]

    The privatization of AI research(-ers): Causes and potential consequences -- from university-industry interaction to public research brain-drain?

    Jurowetzki, R., Hain, D., Mateos-Garcia, J., and Stathoulopoulos, K. The privatization of AI research(-ers): Causes and potential consequences -- from university-industry interaction to public research brain-drain?

  24. [32]

    C., and Stathoulopoulos, K

    Klinger, J., Mateos-Garcia, J. C., and Stathoulopoulos, K. A narrowing of ai research? SSRN Electronic Journal\/ (2020)

    2020

  25. [33]

    Forms and varieties of research and industry collaboration across disciplines

    Kotiranta, A., Tahvanainen, A., Kovalainen, A., and Poutanen, S. Forms and varieties of research and industry collaboration across disciplines. Heliyon 6 , 3 (Mar. 2020), e03404

    2020

  26. [34]

    Latour: Science in action - how to follow scient ists & engineers through society (cloth)

    Latour, B. Latour: Science in action - how to follow scient ists & engineers through society (cloth) . Harvard University Press, London, England, July 1987

    1987

  27. [35]

    Laboratory life: The construction of scientific facts

    Latour, B., and Woolgar, S. Laboratory life: The construction of scientific facts . Princeton university press, 2013 [1979]

    2013

  28. [36]

    L., Hess, D., and Frickel, S

    Moore, K., Kleinman, D. L., Hess, D., and Frickel, S. Science and neoliberal globalization: a political sociological approach. Theory and Society 40\/ (2011), 505--532

    2011

  29. [37]

    Protective or connective professionalism? how connected professionals can (still) act as autonomous and authoritative experts

    Noordegraaf, M. Protective or connective professionalism? how connected professionals can (still) act as autonomous and authoritative experts. Journal of Professions and Organization 7 , 2 (June 2020), 205–223

    2020

  30. [38]

    The idea of collaboration in the academy: Its epistemic and social potentials and risks for knowledge generation

    Papatsiba, V. The idea of collaboration in the academy: Its epistemic and social potentials and risks for knowledge generation. Policy Futures in Education 11 , 4 (Jan. 2013), 436–448

    2013

  31. [39]

    Academic engagement: A review of the literature 2011-2019

    Perkmann, M., Salandra, R., Tartari, V., McKelvey, M., and Hughes, A. Academic engagement: A review of the literature 2011-2019. Research Policy 50 , 1 (Jan. 2021), 104114

    2011

  32. [40]

    Cadrage industriel et production de connaissances

    Raimbault, B. Cadrage industriel et production de connaissances. le cas de la biologie synthétique en france. Sociologie du travail 64 , 4 (Dec. 2022)

    2022

  33. [41]

    Faire avec l’industrie: Repenser la crédibilité scientifique par la preuve de concept

    Raimbault, B. Faire avec l’industrie: Repenser la crédibilité scientifique par la preuve de concept. Revue d’anthropologie des connaissances 17 , 2 (May 2023)

    2023

  34. [42]

    Rikap, C. Varieties of corporate innovation systems and their interplay with global and national systems: Amazon, facebook, google and microsoft’s strategies to produce and appropriate artificial intelligence. Review of International Political Economy 31 , 6 (June 2024), 1735–1763

    2024

  35. [43]

    The pressure to publish pushes down quality

    Sarewitz, D. The pressure to publish pushes down quality. Nature News 533 , 7602 (2016), 147

    2016

  36. [44]

    Schot, J., and Steinmueller, W. E. Three frames for innovation policy: R&d, systems of innovation and transformative change. Research Policy 47 , 9 (Nov. 2018), 1554–1567

    2018

  37. [45]

    The triple helix and new production of knowledge: Prepackaged thinking on science and technology

    Shinn, T. The triple helix and new production of knowledge: Prepackaged thinking on science and technology. Social Studies of Science 32 , 4 (Aug. 2002), 599–614

    2002

  38. [46]

    The transverse science and technology culture: Dynamics and roles of research-technology

    Shinn, T., and Joerges, B. The transverse science and technology culture: Dynamics and roles of research-technology. Social Science Information 41 , 2 (June 2002), 207–251

    2002

  39. [47]

    Paths of commercial knowledge: Forms and consequences of university–enterprise synergy in scientist-sponsored firms

    Shinn, T., and Lamy, E. Paths of commercial knowledge: Forms and consequences of university–enterprise synergy in scientist-sponsored firms. Research Policy 35 , 10 (Dec. 2006), 1465–1476

    2006

  40. [48]

    D., Sch\" a fer, S., and Bernstein, M

    Smith, R. D., Sch\" a fer, S., and Bernstein, M. J. Governing beyond the project: Refocusing innovation governance in emerging science and technology funding. Social Studies of Science 54 , 3 (Nov. 2023), 377–404

    2023

  41. [49]

    Tslearn, a machine learning toolkit for time series data

    Tavenard, R., Faouzi, J., Vandewiele, G., Divo, F., Androz, G., Holtz, C., Payne, M., Yurchak, R., Ru wurm, M., Kolar, K., and Woods, E. Tslearn, a machine learning toolkit for time series data. Journal of Machine Learning Research 21 , 118 (2020), 1--6

    2020

  42. [50]

    K., Pasquier, T., and Crosas, M

    Trisovic, A., Lau, M. K., Pasquier, T., and Crosas, M. A large-scale study on research code quality and execution. Scientific Data 9 , 1 (Feb. 2022)

    2022

  43. [51]

    Atypical combinations and scientific impact

    Uzzi, B., Mukherjee, S., Stringer, M., and Jones, B. Atypical combinations and scientific impact. Science 342 , 6157 (Oct. 2013), 468–472

    2013

  44. [52]

    G., Treude, C., Guo, J., and Matsumoto, K

    Wattanakriengkrai, S., Chinthanet, B., Hata, H., Kula, R. G., Treude, C., Guo, J., and Matsumoto, K. Github repositories with links to academic papers: Public access, traceability, and evolution. Journal of Systems and Software 183\/ (Jan. 2022), 111117

    2022

  45. [53]

    C., Sellitto, M., Shoham, Y., Clark, J., and Perrault, R

    Zhang, D., Mishra, S., Brynjolfsson, E., Etchemendy, J., Ganguli, D., Grosz, B., Lyons, T., Manyika, J., Niebles, J. C., Sellitto, M., Shoham, Y., Clark, J., and Perrault, R. The ai index 2021 annual report, 2021

    2021

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.