Pith. sign in

REVIEW 3 major objections 6 minor 82 references

Forecasting Faculty Placement from Patterns in Co-authorship Networks

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A researcher's pre-hire position in the co-authorship network predicts elite faculty placement beyond what PhD rank and publication metrics capture.

desk verdict A credible first pass at predicting faculty placement from co-authorship networks, but the headline out-of-sample gains are overstated because the static GNNs see test nodes during training. read the letter →

arxiv 2507.14696 v1 pith:6AQO57CX submitted 2025-07-19 cs.SI cs.LG

Samantha Dies , David Liu , Tina Eliassi-Rad This is my paper
classification cs.SIcs.LG
keywords facultyplacementco-authorshipnetworkgraphneuralnetworkstemporalacademichiringprestigehierarchypredictioncomputerscience
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 aims to show that faculty placement—where a newly minted PhD lands their first tenure-track job—can be forecast at the individual level from pre-hire information, and that a candidate's position in the temporal co-authorship network carries predictive signal that PhD department prestige and bibliometric features do not. Using records for 4,656 US computer science faculty, the best model, a graph attention network trained on all three feature types, reaches a precision-recall area under the curve (PR-AUC) of 0.458 for placement at top-10 departments, and adding co-authorship features improves PR-AUC by 8.48% over PhD rank alone, 10.08% over bibliometrics alone, and 7.32% over both combined. The gains are largest at the top-10 threshold, where candidates look similar on paper, and fade as the definition of 'high-rank' broadens. If the claim holds, public collaboration records reveal part of what hiring committees respond to, which would give researchers a concrete way to study and potentially counter structural bias in academic hiring.

What carries the argument

The object that carries the argument is the temporal co-authorship network, a sequence of cumulative undirected graphs with edge weights equal to the number of joint publications, plus node feature tensors for PhD rank and 22 bibliometric features. Over this network the paper trains graph neural networks, most notably a graph attention network (GAT) that weights neighbor contributions during message passing, and a convolutional gated recurrent unit that unrolls across snapshots. The analytic device that isolates network structure is feature-set comparison combined with degree-preserving rewiring: as edges are progressively randomized, PR-AUC falls sharply, showing that the specific higher-order arrangement of co-authorship, not just node degree, is what the models exploit.

What would settle it

Train the leading model on hires from 2010-2016 and evaluate on hires from 2017-2019, rebuilding each year's co-authorship graph so that no test-year candidate or any edge incident to them appears in a training snapshot; if the co-authorship gain over PhD rank alone falls below the reported 7-10% and loses statistical significance, the central claim would be falsified.

Watch

Extended reading notes

Core claim

The central claim is that the structure of a candidate's pre-hire co-authorship network encodes information about the prestige of their first faculty placement that is not captured by PhD department rank or publication-based features. In the paper's terms, co-authorship acts as a proxy for the informal endorsement and social capital that hiring committees observe but that standard curriculum-vitae data omit. The evidence is a systematic model comparison: graph models that run message passing over the co-authorship network outperform all tabular models at every selectivity threshold, and the best configuration, a graph attention network using all three feature sets, raises PR-AUC for top-10 placement to 0.458 from 0.424 for the best non-network model. Mixed-effects regression attributes the gain to the co-authorship term, with the three relevant comparisons all statistically significant. The paper further shows that the effect is concentrated at the most elite departments and that degree-preserving rewiring of the network erodes performance, indicating that higher-order structure, not just degree, carries the signal.

Load-bearing premise

The claim that co-authorship features generalize to unseen candidates rests on treating the 2018-2020 hires as an out-of-sample test, yet the graph models pass messages over the full network that includes those test candidates, so a truly unseen candidate might not show the same gains.

Editorial extensions

If this is right

  • A candidate's pre-hire co-authorship context is a measurable predictor of elite placement, on top of productivity and pedigree.
  • Graph-based models outperform all tabular models at every high-rank threshold tested, so network architecture is not incidental to the result.
  • The co-authorship signal is strongest at the top-10 level and weakens as the threshold broadens, while PhD rank becomes more useful for excluding candidates from less selective definitions of high-rank.
  • Progressive rewiring degrades performance, so the predictive content lies in higher-order co-authorship structure rather than degree alone.

Reading between the lines

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

  • The paper leaves implicit that public co-authorship data could be used as an audit tool: if network position predicts placement beyond merit proxies, hiring outcomes could be checked against collaboration patterns without access to confidential letters or committee deliberations.
  • A natural test of the 'social endorsement' interpretation is to add citation-based quality measures or explicit advisor-advisee and mentor ties; if those variables absorb the network effect, the recommendation-letter proxy story would need revision.
  • A stricter inductive evaluation, in which test-year candidates and their incident edges are absent from every training snapshot, would directly test how much of the reported co-authorship gain survives true out-of-sample prediction.
  • Porting the same framework to disciplines or countries with different collaboration norms would reveal whether the top-10 concentration of network effects is a general feature of prestige hierarchies or a computer-science-specific pattern.
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

3 major / 6 minor

Summary. This paper frames faculty placement as an individual-level prediction task and asks whether temporal co-authorship network structure adds predictive signal beyond PhD department rank and bibliometric features. The dataset consists of 4,656 US computer science faculty, with 1,974 hired between 2010 and 2020; the test set comprises hires from 2018-2020. The authors benchmark tabular classifiers, static GNNs (GCN, GAT, GraphSAGE), and a temporal GNN (GConvGRU), and use linear mixed-effects models to test whether adding co-authorship features improves PR-AUC for placement in top-X departments. They report significant gains for top-10 placement: 8.48% over PhD rank alone, 10.08% over bibliometric features alone, and 7.32% over the combined tabular baseline, with the best model (GAT with all features) reaching PR-AUC 0.458. They also report that co-authorship gains diminish as the high-rank threshold broadens, while PhD rank becomes more predictive.

Significance. If the main claims survive scrutiny, the paper would be a useful contribution to the science-of-science and academic-stratification literature: it moves from population-level correlations to an individual-level predictive framing, releases code and uses public data, and compares multiple model families. The finding that co-authorship structure helps most at the most selective placement thresholds is an interesting and falsifiable pattern. However, the headline quantitative claim is currently threatened by the transductive evaluation used for the static GNNs and by the degenerate random effect in the mixed-effects model. Both issues are addressable, and the paper’s strengths—temporal split design, benchmark breadth, public data and code—make it worthwhile to revise.

major comments (3)
  1. [Section 3.2.1 (static GNNs) and Section 3.1.3] For the static GNNs, the paper states that message passing is performed over the full node set V and that predictions and loss are restricted to Vhire via binary masks. Since V includes the 2018-2020 test hires, the test nodes’ features and edges are visible to the model during training. The claim in Section 3.1.3 of “strict temporal separation” therefore separates labels but not network context. This makes the comparison against tabular models asymmetric: tabular models never see test features during training, while static GNNs do. The reported co-authorship gains in Table 3 (e.g., 7.32%, p=0.003 over PhD+Bib) may reflect the transductive access of static GNNs rather than the predictive value of co-authorship features for genuinely unseen candidates. I request an inductive evaluation in which test nodes are excluded from the message-passing graph during training (e.g., train on snapshots built from Vtrain and Vval only, then apply to a graph that includes test nodes only at inference time), and a within-architecture comparison (e.g., GAT with and without co-author features) to isolate the feature-set contribution from the transductive effect.
  2. [Table 3 and Eq. (4)] The linear mixed-effects model used to support the significance claims reports Group Variance = 0.000 (SE 0.003) in every specification. A zero variance estimate means the random intercept for model architecture carries no information, and the model effectively reduces to ordinary linear regression without an architecture effect. The paper’s statement that the model “controll[s] for model type” via the random effect is therefore not supported by the reported fit. The p-values for adding co-authorship should be re-derived with a method that actually accounts for repeated observations per architecture (e.g., cluster-robust standard errors or a likelihood-ratio test against a fixed-effects-only model), and the boundary estimate should be discussed and interpreted.
  3. [Section 3.1.3 and Section 5] The evaluation uses a single temporal split with hires from 2010-2017 as train/validation and 2018-2020 as test. All conclusions about out-of-sample generalization rest on this one window, which is an untested assumption about representativeness; market dynamics, the COVID-era academic job market, or DBLP data completeness could affect this period. The authors should either run a rolling-origin evaluation (e.g., multiple cut points: train on 2010-Y, test on Y+1 for a range of Y) or explicitly qualify the claims as specific to this split. This is especially relevant for the threshold analysis in Section 5.2, where the confidence intervals in Figure 4 are computed within a single test set.
minor comments (6)
  1. [Figure 3 caption] The caption states that the GAT achieves an average PR-AUC of 0.424 and the transformer achieves 0.458, but Table 2 reports the reverse (GAT 0.458, Transformer 0.424); please correct the caption.
  2. [Section 3.2.1] “Spaciotemporal” should be “spatiotemporal” in the description of GConvGRU.
  3. [Table A1 and Tables A6-A9 captions] There are typographical errors: “co-authorhship” appears in the Table A1 caption, and “function fo” appears in the Table A6-A9 captions; please fix these.
  4. [Section 5.2] The percentages given in the text mix units: “top-10 (approximately 5% of the 186 departments)” refers to departments, while the immediately following percentages (22%, 36%, 49%, 67%, 77%) refer to researchers hired at those thresholds; please clarify the units to avoid confusion.
  5. [Section 4.2] The sentence “Features 2 and 6-11 use attributes of a researcher’s co-authors” should be “Features 6-11 use attributes of a researcher’s co-authors”, since feature 2 (average number of authors per paper) is a property of the paper rather than of co-author attributes.
  6. [Introduction, Contribution 1] The claim that no prior work incorporates co-authorship networks into predictive models of faculty placement should be narrowed to “no prior work evaluates such models on temporally held-out individuals,” given that Zuo et al. and Barnes et al. are cited as examining co-authorship and placement; the current wording overstates the novelty.

Circularity Check

2 steps flagged · score 6.0 of 10

The claimed co-authorship improvement partially reduces to architecture and transductive evaluation: for GNNs, the 'PhD' and 'PhD+Co-authorship' feature sets are the same input by the paper's own definition, and static GNNs see test nodes during training.

  1. fitted input called prediction [Sections 3.1.2, 3.1.3, 3.2.1]
    "The node set V is fixed and consists of the 4656-researcher cohort... Message passing is performed over the full node set V, but predictions and loss computation are restricted to Vhire using binary node masks. ... This protocol ensures a strict temporal separation between observed (training/validation) and unobserved (test) researchers, enabling robust evaluation of predictive generalization to truly out-of-sample cases."

    For the static GNNs that dominate the top results, training on snapshot G_{t-1} includes test nodes as part of V: their features are in the input tensor and their edges participate in message passing during every training epoch. The claimed 'strict temporal separation' separates only labels, not node features or network context. The model's parameters are therefore fit with test-node inputs, and the reported out-of-sample PR-AUC gains for co-authorship are transductive scores. The prediction is presented as generalization to unseen individuals, but the network context of those individuals is observed during training, so the measured co-authorship advantage is not evidence of transfer to genuinely unseen candidates.

  2. self definitional [Section 3.2.1 and Table 1]
    "Each model receives node features from XPhD, XBib, both, or, in the case of no features, the constant vector 1 as is standard in graph machine learning77. ... Message passing is performed over the full node set V, but predictions and loss computation are restricted to Vhire using binary node masks."

    For any GNN, the co-authorship graph is always the substrate of message passing, regardless of the node-feature set. Thus a GNN with node features XPhD ('PhD' feature set) and a GNN with node features XPhD plus the co-authorship graph ('PhD+Co-authorship' feature set) are identical by the paper's stated input definition. The linear mixed-effects regression (Table 3) reports a significant coefficient for adding co-authorship, but within graph architectures the two feature sets collapse, so the coefficient can only be identified by comparing graph architectures (which always use co-authorship) with tabular architectures (which never do). The claimed co-authorship improvement therefore reduces to an architecture comparison rather than to a separately varied co-authorship input.

full rationale

The paper contains no load-bearing self-citations, and the outcome labels are genuinely held out for all models, so the individual placement predictions are not tautological. However, the headline contribution—that co-authorship features improve prediction—is undermined by two construction-level problems: for GNNs the 'PhD' and 'PhD+Co-authorship' feature sets coincide, making the mixed-model feature-set coefficient an architecture confound; and static GNNs include test nodes in the message-passing graph during training, making the 'out-of-sample' evaluation transductive. These issues mean that part of the central quantitative claim reduces by construction rather than by an independent network-signal comparison. No evidence of self-citation chains, imported uniqueness theorems, or renamed known results was found.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The paper is an empirical ML study, so the main hand-chosen quantities are the outcome threshold, the co-author rank cutoffs, the temporal split, and the sampling probability. The domain assumptions about data quality and network scope are load-bearing for the predictive claim. No new theoretical entities or mediators are introduced.

free parameters (4)
  • High-rank threshold (primary: top-10) = 10
    The binary outcome is defined by whether the first faculty appointment is at a CSRankings top-10 department; the paper also evaluates top-20/30/40/50. This hand-chosen threshold determines the class balance and the measured effect sizes.
  • Co-author high-rank cutoffs in features = top-10 and top-50
    Bibliometric features 8-11 count co-authors employed at top-10 or top-50 departments; these cutoffs are chosen without a stated data-driven justification.
  • Temporal split boundary = test on 2018-2020 hires
    Hires from 2018-2020 form the single test set; this boundary is chosen by the authors and affects the evaluation.
  • Train sampling probability = 0.8
    Within each training year, 80% of candidates are sampled for training and 20% for validation, adding stochasticity to the split.
assumptions (5)
  • domain assumption CSRankings publication-based ranks are a valid proxy for department prestige and for the outcome label.
    Used throughout to define both PhD and faculty department ranks (Section 3.1.1).
  • domain assumption The CS Professors dataset correctly records each researcher's first tenure-track placement and hire year.
    Labels and temporal splits are built on this dataset (Section 3.1.1).
  • domain assumption Restricting the co-authorship network to the 4,656-faculty cohort preserves the predictive signal of collaboration structure.
    The network contains only nodes from the faculty cohort, excluding non-faculty collaborators (Section 3.1.2).
  • domain assumption The 2018-2020 hiring window is representative of the faculty placement process.
    The single temporal test split is used to estimate generalization (Section 3.1.3).
  • standard math Standard supervised learning assumptions: the training distribution matches the test distribution for the binary placement task.
    The ML models and evaluation assume exchangeability of candidates conditional on features.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Forecasting Faculty Placement from Patterns in Co-authorship Networks." pith.science (2026). https://pith.science/paper/6AQO57CX

@misc{pith2026250714696,
  author       = {Pith},
  title        = {Pith review of: Forecasting Faculty Placement from Patterns in Co-authorship Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6AQO57CX}},
  note         = {Machine review of arXiv:2507.14696}
}
read the original abstract

Faculty hiring shapes the flow of ideas, resources, and opportunities in academia, influencing not only individual career trajectories but also broader patterns of institutional prestige and scientific progress. While traditional studies have found strong correlations between faculty hiring and attributes such as doctoral department prestige and publication record, they rarely assess whether these associations generalize to individual hiring outcomes, particularly for future candidates outside the original sample. Here, we consider faculty placement as an individual-level prediction task. Our data consist of temporal co-authorship networks with conventional attributes such as doctoral department prestige and bibliometric features. We observe that using the co-authorship network significantly improves predictive accuracy by up to 10% over traditional indicators alone, with the largest gains observed for placements at the most elite (top-10) departments. Our results underscore the role that social networks, professional endorsements, and implicit advocacy play in faculty hiring beyond traditional measures of scholarly productivity and institutional prestige. By introducing a predictive framing of faculty placement and establishing the benefit of considering co-authorship networks, this work provides a new lens for understanding structural biases in academia that could inform targeted interventions aimed at increasing transparency, fairness, and equity in academic hiring practices.

Figures

Figures reproduced from arXiv: 2507.14696 by the authors.

Figure 1
Figure 1. Overview of related tasks. We summarize the existing literature landscape according to the type of task, i.e., predictive or not. We also consider whether the models predict faculty placement and if they incorporate placement in the co-authorship network. To our knowledge, our paper is the first to frame the prediction of faculty placement using temporal co-authorship networks as a machine learning task that general… view at source ↗
Figure 2
Figure 2. Temporal data splits and the modeling pipeline. We depict how we split the data into train, validation, and test sets and outline how the datasets are used by different models. Here, (a) displays the temporal data splitting process. Data from years t0 to tf −4 are used as the train (red) and validation (yellow) input, while data from years tf −3 to tf −1 serve as the test (blue) input. In all cases, we define y, the… view at source ↗
Figure 3
Figure 3. Predictive performance for top-10 department placement by feature set and model type. We plot the average PR-AUC and standard deviation across 10 runs for each of the model trained on our feature sets. Higher values of PR-AUC indicate better performance, and we compare against our three heuristics: random guessing (dotted), the average co-author rank (dashed), and PhD rank (dash-dotted). The performance of the white… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: The impact of the co-authorship network on performance and the usefulness of PhD rank vs. bibliometric features for different thresholds for high-rank. Each line shows the average difference in PR-AUC (with 90% confidence intervals) between two feature sets. In (a), co…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

82 extracted references · 51 canonical work pages

  1. [1]

    Abramo, G., D’angelo, C. A. & Caprasecca, A. The contribution of star scientists to overall sex differ- ences in research productivity.Scientometrics 81, 137–156, DOI: 10.1007/s11192-008-2131-7 (2009)

    work page doi:10.1007/s11192-008-2131-7 2009

  2. [2]

    H., Larremore, D

    Zhang, S., Wapman, K. H., Larremore, D. B. & Clauset, A. Labor advantages drive the greater productivity of faculty at elite universities.Sci. Adv. 8, abq7056, DOI: 10.1126/sciadv.abq7056 (2022)

    work page doi:10.1126/sciadv.abq7056 2022

  3. [3]

    Petersen, A. M. Quantifying the impact of weak, strong, and super ties in scientific careers. Proc. Natl. Acad. Sci.112, 4671–4680, DOI: doi.org/10. 1073/pnas.1501444112 (2015)

    2015

  4. [4]

    Fernandes, J. D.et al. A survey-based analysis of the academic job market. Elife 9, e54097, DOI: 10.7554/eLife.54097 (2020)

    work page doi:10.7554/elife.54097 2020

  5. [5]

    Cole, J. R. & Cole, S.Social Stratification in Science (University of Chicago Press, Chicago, 1973)

    1973

  6. [6]

    Manis, J. G. Some academic influences upon publi- cation productivity.Soc. Forces29, 267–272, DOI: 10.2307/2572416 (1951)

    work page doi:10.2307/2572416 1951

  7. [7]

    Hagstrom, W. O. Departmental prestige and scien- tific productivity.Sociol. Abstr.16, 16 (1968)

    work page 1968

  8. [8]

    Morgan, A. C.et al. Socioeconomic roots of aca- demic faculty.Nat. Hum. Behav.6, 1625–1633, DOI: 10.31235/osf.io/6wjxc (2022)

    work page doi:10.31235/osf.io/6wjxc 2022

Show all 82 references
  1. [9]

    & Kim, J

    Celis, S. & Kim, J. The making of homophilic net- works in international research collaborations: A global perspective from chilean and korean engineer- ing. Res. Policy47, 573–582, DOI: 10.1016/j.respol. 2018.01.001 (2018)

    2018 doi

  2. [10]

    Zuo, Z., Zhao, K. & Ni, C. Standing on the shoulders of giants?—Faculty hiring in information schools.J. Informetrics 13, 341–353, DOI: 10.1016/j.joi.2019. 01.007 (2019)

    2019 doi

  3. [11]

    Posselt, J. R. Equity in science: Representation, culture, and the dynamics of change in graduate education (Stanford University Press, 2020)

    2020

  4. [12]

    Holley, K. A. & Gardner, S. Navigating the pipeline: How socio-cultural influences impact first-generation doctoral students. J. Divers. High. Educ.5, 112, DOI: 10.1037/a0026840 (2012)

    2012 doi

  5. [13]

    H., Zhang, S., Clauset, A

    Wapman, K. H., Zhang, S., Clauset, A. & Larremore, D. B. Quantifying hierarchy and dynamics in us faculty hiring and retention.Nature 610, 120–127, DOI: 10.1038/s41586-022-05222-x (2022)

    2022 doi

  6. [14]

    G., Cavazos, D

    Bedeian, A. G., Cavazos, D. E., Hunt, J. G. & Jauch, L. R. Doctoral degree prestige and the academic marketplace: A study of career mobility within the management discipline. Acad. Manag. Learn. & Educ. 9, 11–25, DOI: 10.5465/amle.9.1.zqr11 (2010). 17/34

    2010 doi

  7. [15]

    FitzGerald, C., Huang, Y., Leisman, K. P. & Topaz, C. M. Temporal dynamics of faculty hiring in math- ematics. Humanit. Soc. Sci. Commun.10, 1–10, DOI: 10.31235/osf.io/3bgyq (2023)

    2023 doi

  8. [16]

    Li, W., Zhang, S., Zheng, Z., Cranmer, S. J. & Clauset, A. Untangling the network effects of produc- tivity and prominence among scientists.Nat. Com- mun. 13, 4907, DOI: 10.1038/s41467-022-32604-6 (2022)

    2022 doi

  9. [17]

    Long, J. S. Productivity and academic position in the scientific career.Am. Sociol. Rev.43, 889–908, DOI: 10.2307/2094628 (1978)

    1978 doi

  10. [18]

    The academic caste system: Prestige hier- archies in PhD exchange networks.Am

    Burris, V. The academic caste system: Prestige hier- archies in PhD exchange networks.Am. Sociol. Rev. 69, 239–264, DOI: 10.1177/000312240406900205 (2004)

    2004 doi

  11. [19]

    Barnes, K.et al. Edge interventions can mitigate demographic and prestige disparities in the com- puter science coauthorship network.arXiv preprint arXiv:2506.04435 DOI: 10.48550/arXiv.2506.04435 (2025)

    2025 doi arXiv

  12. [20]

    Prediction models for diagno- sis and prognosis of covid-19: Systematic review and critical appraisal

    Wynants, L.et al. Prediction models for diagno- sis and prognosis of covid-19: Systematic review and critical appraisal. BMJ 369, DOI: 10.3410/f. 737693696.793573809 (2020)

    2020

  13. [21]

    The impact of machine learning on economics

    Athey, S. The impact of machine learning on economics. In The economics of artificial intelli- gence: An agenda, 507–547, DOI: 10.7208/chicago/ 9780226613475.003.0021 (University of Chicago Press, 2018)

    2018 doi

  14. [22]

    & Brunet, G

    Bauer, P., Thorpe, A. & Brunet, G. The quiet revolution of numerical weather prediction.Nature 525, 47–55, DOI: 10.1038/nature14956 (2015)

    2015 doi

  15. [23]

    F., Larremore, D

    Way, S. F., Larremore, D. B. & Clauset, A. Gen- der, productivity, and prestige in computer science faculty hiring networks. InProceedings of the 25th International Conference on World Wide Web, vol. 1, 1169–1179, DOI: 10.1145/2872427.2883073 (2016)

    2016

  16. [24]

    & Larremore, D

    Clauset, A., Arbesman, S. & Larremore, D. B. Sys- tematic inequality and hierarchy in faculty hiring networks. Sci. Adv. 1, e1400005, DOI: 10.1126/ sciadv.1400005 (2015)

    2015

  17. [25]

    & Schweitzer, F

    Sarigöl, E., Pfitzner, R., Scholtes, I., Garas, A. & Schweitzer, F. Predicting scientific success based on coauthorship networks.EPJ Data Sci.3, 1–16, DOI: 10.1140/epjds/s13688-014-0009-x (2014)

    2014 doi

  18. [26]

    & Van der Leij, M

    Ductor, L., Fafchamps, M., Goyal, S. & Van der Leij, M. J. Social networks and research output. Rev. Econ. Stat.96, 936–948, DOI: 10.1162/rest_ a_00430 (2014)

    2014 doi

  19. [27]

    Du, W., Xie, Z. & Lv, Y. Predicting publication productivity for authors: Shallow or deep archi- tecture? Scientometrics 126, 5855–5879, DOI: 10.1007/s11192-021-04027-5 (2021)

    2021 doi

  20. [28]

    & Davies, B

    Grodzinski, N., Grodzinski, B. & Davies, B. M. Can co-authorship networks be used to predict author research impact? A machine-learning based analysis within the field of degenerative cervical myelopathy research. PLOS One16, e0256997, DOI: 10.1371/ journal.pone.0256997 (2021)

    2021

  21. [29]

    & Vazirgiannis, M

    Nikolentzos, G., Panagopoulos, G., Evdaimon, I. & Vazirgiannis, M. Can author collaboration reveal impact? The case of h-index. Predict. Dyn. Res. Impact 177–194, DOI:10.1007/978-3-030-86668-6_8 (2021)

    2021 doi

  22. [30]

    Daud, A. et al. Finding rising stars in co-author networks via weighted mutual influence. In Pro- ceedings of the 26th International Conference on World Wide Web Companion, vol. 1, 33–41, DOI: 10.1145/3041021.3054137 (2017)

    2017

  23. [31]

    & Chen, F

    Ding, F., Liu, Y., Chen, X. & Chen, F. Rising star evaluation in heterogeneous social network.IEEE Access6, 29436–29443, DOI: 10.1109/access.2018. 2812923 (2018)

    2018 doi

  24. [32]

    Billah, S. M. & Gauch, S. Social network analysis for predicting emerging researchers. In2015 7th In- ternational Joint Conference on Knowledge Discov- ery, Knowledge Engineering and Knowledge Manage- ment, vol. 1, 27–35, DOI: 10.5220/0005593500270035 (IEEE, 2015)

    doi

  25. [33]

    S., Tew, K

    Li, X.-L., Foo, C. S., Tew, K. L. & Ng, S.-K. Search- ing for rising stars in bibliography networks. In Database Systems for Advanced Applications: 14th International Conference, DASFAA 2009, Brisbane, Australia, April 21-23, 2009. Proceedings 14, 288– 292, DOI: 10.1007/978-3-...

    2009 doi

  26. [34]

    The chaperone effect in scientific publishing

    Sekara, V.et al. The chaperone effect in scientific publishing. Proc. Natl. Acad. Sci.115, 12603–12607, DOI: 10.1073/pnas.1800471115 (2018)

    2018 doi

  27. [35]

    Y., Liao, C

    Li, E. Y., Liao, C. H. & Yen, H. R. Co-authorship networks and research impact: A social capital perspective. Res. Policy 42, 1515–1530, DOI: 10.1016/j.respol.2013.06.012 (2013)

    2013 doi

  28. [36]

    & Boardman, C

    Ponomariov, B. & Boardman, C. What is co- authorship? Scientometrics 109, 1939–1963, DOI: 10.1007/s11192-016-2127-7 (2016)

    2016 doi

  29. [37]

    P., Jones, C

    Borgatti, S. P., Jones, C. & Everett, M. G. Network measures of social capital.Connections 21, 27–36 (1998)

    1998

  30. [38]

    F., Morgan, A

    Way, S. F., Morgan, A. C., Clauset, A. & Larremore, D. B. The misleading narrative of the canonical 18/34 facultyproductivitytrajectory. Proc. Natl. Acad. Sci. 114, E9216–E9223, DOI: 10.1073/pnas.1702121114 (2017)

    2017 doi

  31. [39]

    Aguinis, H., Ji, Y. H. & Joo, H. Gender productiv- ity gap among star performers in stem and other scientific fields. J. Appl. Psychol.103, 1283, DOI: 10.1037/apl0000331 (2018)

    2018 doi

  32. [40]

    Ceci, S. J. & Williams, W. M. Women have substan- tial advantage in stem faculty hiring, except when competing against more-accomplished men.Front. psychology 1532, DOI: 10.3389/fpsyg.2015.01532 (2015)

    2015

  33. [41]

    & Zeng, A

    Xing, Y., Ma, Y., Fan, Y., Sinatra, R. & Zeng, A. Academic mentees succeed in big groups, but thrive in small groups.Nat. Hum. Behav.9, 902–916, DOI: 10.1038/s41562-025-02114-8 (2025)

    2025 doi

  34. [42]

    F., Larremore, D

    Zhang, S., LaBerge, N., Way, S. F., Larremore, D. B. & Clauset, A. Scientific productivity as a random walk. arXiv preprint arXiv:2309.04414 DOI: 10. 48550/arXiv.2309.04414 (2023)

    2023 doi

  35. [43]

    H., Clauset, A

    LaBerge, N., Wapman, K. H., Clauset, A. & Lar- remore, D. B. Gendered hiring and attrition on the path to parity for academic faculty.Elife 13, RP93755, DOI: 10.7554/elife.93755.2.sa2 (2024)

    2024 doi

  36. [44]

    J., Sinatra, R

    Huang, J., Gates, A. J., Sinatra, R. & Barabási, A.-L. Historical comparison of gender inequality in scientific careers across countries and disciplines. Proc. Natl. Acad. Sci.117, 4609–4616, DOI: 10. 1073/pnas.1914221117 (2020)

    2020

  37. [45]

    L., Melkers, J

    Pinheiro, D. L., Melkers, J. & Newton, S. Take me where I want to go: Institutional prestige, advisor sponsorship, and academic career place- ment preferences. PLOS One12, e0176977, DOI: 10.1371/journal.pone.0176977 (2017)

    2017 doi

  38. [46]

    F., Achakulvisut, T., Acuna, D

    Liénard, J. F., Achakulvisut, T., Acuna, D. E. & David, S. V. Intellectual synthesis in mentorship de- termines success in academic careers.Nat. Commun. 9, 4840, DOI: 10.1038/s41467-018-07034-y (2018)

    2018 doi

  39. [47]

    C., Economou, D

    Morgan, A. C., Economou, D. J., Way, S. F. & Clauset, A. Prestige drives epistemic inequality in the diffusion of scientific ideas.EPJ Data Sci.7, 40, DOI: 10.1140/epjds/s13688-018-0166-4 (2018)

    2018 doi

  40. [48]

    J., Kahn, S

    Ceci, S. J., Kahn, S. & Williams, W. M. Exploring gender bias in six key domains of academic science: An adversarial collaboration.Psychol. Sci. Public In- terest 24, 15–73, DOI: 10.1177/15291006231163179 (2023)

    2023 doi

  41. [49]

    & Livan, G

    Sun, Y., Caccioli, F. & Livan, G. Ranking mobility and impact inequality in early academic careers (2023). 2303.15988

    2023 arXiv

  42. [50]

    C., Glick, W

    Miller, C. C., Glick, W. H. & Cardinal, L. B. The allocation of prestigious positions in organizational science: Accumulative advantage, sponsored mobil- ity, and contest mobility.J. Organ. Behav. The Int. J. Ind. Occup. Organ. Psychol. Behav.26, 489–516, DOI: 10.1002/job.325 (2005)

    2005 doi

  43. [51]

    Hargens, L. L. & Hagstrom, W. O. Sponsored and contest mobility of american academic scientists.So- ciol. Educ.40, 24–38, DOI: 10.2307/2112185 (1967)

    1967 doi

  44. [52]

    & Conti, A

    Baruffaldi, S., Visentin, F. & Conti, A. The produc- tivity of science & engineering phd students hired from supervisors’ networks.Res. Policy45, 785–796, DOI: 10.1016/j.respol.2015.12.006 (2016)

    2016 doi

  45. [53]

    & O’Neill, S

    Yadav, A., McHale, J. & O’Neill, S. How does co- authoring with a star affect scientists’ productivity? Evidence from small open economies.Res. Policy52, 104660, DOI: 10.1016/j.respol.2022.104660 (2023)

    2023

  46. [54]

    & Wang, D

    Li, J., Yin, Y., Fortunato, S. & Wang, D. Scientific elite revisited: Patterns of productivity, collabora- tion, authorship and impact.J. Royal Soc. Interface 17, 20200135, DOI: 10.1098/rsif.2020.0135 (2020)

    2020

  47. [55]

    Co-authorship and the output of academic economists

    Hollis, A. Co-authorship and the output of academic economists. Labour Econ.8, 503–530, DOI: 10.1016/ S0927-5371(01)00041-0 (2001)

    2001

  48. [56]

    Xu, Q. A. & Chang, V. Co-authorship network and the correlation with academic performance.Internet Things 12, 100307, DOI: 10.1016/j.iot.2020.100307 (2020)

    2020

  49. [57]

    & Livan, G

    Li, W., Aste, T., Caccioli, F. & Livan, G. Early coauthorship with top scientists predicts success in academic careers. Nat. Commun. 10, 5170, DOI: 10.1038/s41467-019-13130-4 (2019)

    2019 doi

  50. [58]

    & Steffen-Fluhr, N

    Collins, R. & Steffen-Fluhr, N. Hidden patterns: Using social network analysis to track career tra- jectories of women stem faculty. Equal. Divers. Inclusion: An Int. J.38, 265–282, DOI: 10.1108/ edi-09-2017-0183 (2019)

    2019

  51. [59]

    M., Williams, H

    Jaramillo, A. M., Williams, H. T., Perra, N. & Menezes, R. The structure of segregation in co- authorship networks and its impact on scientific production. EPJ Data Sci.12, 47, DOI: 10.1140/ epjds/s13688-023-00411-8 (2023)

    2023

  52. [60]

    M., Macedo, M

    Jaramillo, A. M., Macedo, M. & Menezes, R. Reach- ing to the top: The gender effect in highly-ranked academics in computer science. Adv. Complex Syst.24, 2150008, DOI: 10.1142/s0219525921500089 (2021)

    2021 doi

  53. [61]

    & Zhao, K

    Zuo, Z. & Zhao, K. Understanding and predicting fu- ture research impact at different career stages—A so- cial network perspective.J. Assoc. for Inf. Sci. Tech- nol. 72, 454–472, DOI: 10.1002/asi.24415 (2021)

    2021 doi

  54. [62]

    Lee, D. H. Predicting the research performance of early career scientists. Scientometrics 121, 1481– 1504, DOI: 10.1007/s11192-019-03232-7 (2019). 19/34

    2019 doi

  55. [63]

    Berger, E. D. CSRankings. https://csrankings.org/ (2023)

    2023

  56. [64]

    Werbos, P. J. Backpropagation through time: What it does and how to do it. InProceedings of the IEEE, vol. 78, 1550–1560, DOI: 10.1109/5.58337 (IEEE, 1990)

    1990 doi

  57. [65]

    DBLP computer science bibliogra- phy

    DBLP. DBLP computer science bibliogra- phy. https://dblp.org/xml/release/dblp-2023-1-03. xml.gz (2023)

    2023

  58. [66]

    Computer science open data

    Huang, J. Computer science open data. https:// jeffhuang.com/computer-science-open-data/ (2022)

    2022

  59. [67]

    & Vapnik, V

    Cortes, C. & Vapnik, V. Support-vector net- works. Mach. Learn.20, 273–297, DOI: 10.1007/ BF00994018 (1995)

    1995

  60. [68]

    Random forests.Mach

    Breiman, L. Random forests.Mach. Learn.45, 5–32, DOI: 10.1023/a:1010933404324 (2001)

    2001 doi

  61. [69]

    & Guestrin, C

    Chen, T. & Guestrin, C. XGBoost: A scalable tree boosting system. InProceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794, DOI: 10.1145/ 2939672.2939785 (2016)

    2016

  62. [70]

    E., Hinton, G

    Rumelhart, D. E., Hinton, G. E. & Williams, R. J. Learning representations by back-propagating er- rors. Nature 323, 533–536, DOI: 10.1038/323533a0 (1986)

    1986 doi

  63. [71]

    & Karnin, Z

    Huang, X., Khetan, A., Cvitkovic, M. & Karnin, Z. Tabtransformer: Tabular data modeling using contextual embeddings. arXiv preprint arXiv:2012.06678 DOI: 10.48550/arXiv.2012.06678 (2020)

    2020 doi arXiv

  64. [72]

    Kipf, T. N. & Welling, M. Semi-supervised classifi- cation with graph convolutional networks.arXiv preprint arXiv:1609.02907 DOI: 10.48550/arXiv. 1609.02907 (2016)

    2016 doi arXiv

  65. [73]

    Graph attention networks.Stat 1050, 10–48550, DOI: 10.48550/arXiv.1710.10903 (2017)

    Velickovic, P.et al. Graph attention networks.Stat 1050, 10–48550, DOI: 10.48550/arXiv.1710.10903 (2017)

    2017 doi arXiv

  66. [74]

    & Leskovec, J

    Hamilton, W., Ying, Z. & Leskovec, J. Inductive representation learning on large graphs. InAdvances in Neural Information Processing Systems, vol. 30, DOI: 10.5555/3294771.3294869 (Curran Associates, Inc., 2017)

    2017

  67. [75]

    & Bres- son, X

    Seo, Y., Defferrard, M., Vandergheynst, P. & Bres- son, X. Structured sequence modeling with graph convolutionalrecurrentnetworks. In Neural Informa- tion Processing: 25th International Conference, 362– 373, DOI: 10.1007/978-3-030-04167-0_33 (Springer, 2018)

    doi

  68. [76]

    & Vandergheynst, P

    Defferrard, M., Bresson, X. & Vandergheynst, P. Convolutional neural networks on graphs with fast localized spectral filtering. InAdvances in Neural Information Processing Systems, vol. 29, 3844–3852, DOI: 10.5555/3157382.3157527 (Curran Associates, Inc., 2016)

  69. [77]

    & Yang, C

    Cui, H., Lu, Z., Li, P. & Yang, C. On positional and structural node features for graph neural networks on non-attributed graphs. InProceedings of the 31st ACM International Conference on Information & Knowledge Management, 3898–3902, DOI: 10.1145/ 3511808.3557661 (2022)

    2022

  70. [78]

    Scikit-learn: Machine learning in python

    Pedregosa, F.et al. Scikit-learn: Machine learning in python. J. Mach. Learn. Res.12, 2825–2830, DOI: 10.5555/1953048.2078195 (2011)

    2011

  71. [79]

    PyTorch: An imperative style, high- performance deep learning library.arXiv preprint arXiv:1912.01703 DOI: 10.48550/arXiv.1912.01703 (2019)

    Paszke, A. PyTorch: An imperative style, high- performance deep learning library.arXiv preprint arXiv:1912.01703 DOI: 10.48550/arXiv.1912.01703 (2019)

    2019 doi arXiv

  72. [80]

    & Lenssen, J

    Fey, M. & Lenssen, J. E. Fast graph representation learning with PyTorch geometric.arXiv preprint arXiv:1903.02428 DOI: 10.48550/arXiv.1903.02428 (2019)

    2019 doi arXiv

  73. [81]

    Rozemberczki, B. et al. PyTorch geometric tem- poral: Spatiotemporal signal processing with neu- ral machine learning models. InProceedings of the 30th ACM International Conference on Informa- tion & Knowledge Management, 4564–4573, DOI: 10.1145/3459637.3482014 (2021)

    2021

  74. [82]

    Adam: Amethodforstochas- tic optimization

    Kingma, D.P.&Ba, J. Adam: Amethodforstochas- tic optimization. arXiv preprint arXiv:1412.6980 (2017). 20/34 A Rewiring Co-authorship Networks To evaluate the degree to which the specific edges in the co-authorship network carry meaningful predictive signals beyond degree, we t...

    2017 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.