Ideas worth spreading: How does network position influence the spread of research topics? Allison Morgan, Dimitrios Economou, Samuel Way, Aaron Clauset
Science is a meritocracy… right? https://pxhere.com/en/photo/950021 (CC 2.0)
Yet, we know that some scientists and institutions are far more influential than others. Proc. Natl. Acad. Sci. U.S.A 111(43) 15316-15321(2014) Am. J. Soc. 76(2), 286-306 (1970) Science 159.3810, 56-63 (1968) Sociol. Educ. 375-397 (1971) Am. Soc. Rev. 55, 469-478 (1990) Critical Inquiry (2017)
Three explanations (1) genuine differences in merit (2) non-meritocratic social processes (3) non-meritocratic structural factors https://www.olympic.org/news/1932-the-podium-makes-its-olympic-debut
Three explanations (1) genuine differences in merit (2) non-meritocratic social processes (3) non-meritocratic structural factors https://www.olympic.org/news/1932-the-podium-makes-its-olympic-debut
Faculty hiring as a mechanism American Scientist 55, 156-165 (2005) R1: Are research ideas carried by Proc. 11th Conf. on Web and Social Media ( 2017 ) faculty hiring? MIT Cornell Caltech UC Berkeley Harvard R2: Does the structure of the faculty hiring network affect the spread of ideas? Stanford Washington Princeton Carnegie Mellon Yale Sci. Adv. 1(1), e1400005, 2015.
Data Faculty hiring networks MIT Cornell Caltech Education & employment for faculty from 205 UC Berkeley Harvard U.S. and Canadian CS departments - Institution (node) u with unique prestige π Stanford Washington - Edge ( u, v ) represents a PhD candidate from u Princeton Carnegie Mellon Yale who got an assistant faculty position at v Science Advances 1(1), e1400005, 2015. Publication records Over 200K publication records for 2.6K tenure- track faculty - Title, author list, venue, and date - Matched with employment start dates Proc. 25th Int'l World Wide Web Conf. (WWW), ( 2016 )
R1: Are research ideas carried by faculty hiring?
R1: Are research ideas carried by faculty hiring? For each department that has adopted a research idea, either: (a) the department hired a scientist who works on that idea [hiring], or (b) some scientist at the department begins working on the idea [non-hiring] Test: choose 3 research topics and evaluate the fraction of times those topics spread via (a) in real life, compared to the expected fraction under (b)
R1: Are research ideas carried by faculty hiring? For each department that has adopted a research idea, either: (a) the department hired a scientist who works on that idea [hiring], or (b) some scientist at the department begins working on the idea [non-hiring] Test: choose 3 research topics and evaluate the fraction of times those topics spread via (a) in real life, compared to the expected fraction under a permutation of publication titles
<latexit sha1_base64="iAqWNq9w5glWEuGpFKBOjaPT7j0=">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</latexit> <latexit sha1_base64="iAqWNq9w5glWEuGpFKBOjaPT7j0=">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</latexit> <latexit sha1_base64="iAqWNq9w5glWEuGpFKBOjaPT7j0=">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</latexit> <latexit sha1_base64="iAqWNq9w5glWEuGpFKBOjaPT7j0=">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</latexit> R1: Are research ideas carried by faculty hiring? For each department that has adopted a research idea, either: (a) the department hired a scientist who works on that idea [hiring], or (b) some scientist at the department begins working on the idea [non-hiring] Test: choose 3 research topics and evaluate the fraction of times those topics spread via (a) in real life, compared to the expected fraction under a permutation of publication titles 2/3 of research topics were topic X h o h e p significantly more likely to be ❌ deep learning 0.34 0.30 0 . 16 ± 0 . 01 transmitted via hiring than at ✅ topic modeling 0.33 0.22 0 . 01 ± 0 . 01 random ✅ incremental computing 0.39 0.19 0 . 01 ± 0 . 01
R2: Does the structure of the faculty hiring network affect the spread of ideas?
R2: Does the structure of the faculty hiring network affect the spread of ideas? Northwestern To simulate the diffusion of ideas, use a Susceptible-Infected (SI) model Columbia Seed an epidemic at a single Harvard Stanford university with unique prestige π Colorado (network location) Slope: 0.0163 5.0 Varying the transmissibility p (quality 4.5 Average Path Length ( 〈 l 〉 ) 4.0 of an idea) 3.5 3.0 MIT 2.5 New Mexico Measure the fraction of universities State University 2.0 University of Colorado, which have adopted the idea 1.5 Boulder 1.0 20 40 60 80 100 120 140 160 Universities Sorted by Prestige ( π ) Increasing Prestige
R2: Does the structure of the faculty hiring network affect the spread of ideas? 1.0 Fraction of Network Infected 0.8 p Epidemic Size 0.10 0.6 0.30 0.50 0.4 0.70 0.90 0.2 0.0 0 50 100 150 University Prestige ( π ) Increasing Prestige
R2: Does the structure of the faculty hiring network affect the spread of ideas? 1.0 Fraction of Network Infected 0.8 p Epidemic Size 0.10 0.6 0.30 0.50 0.4 0.70 0.90 0.2 0.0 0 50 100 150 University Prestige ( π ) Increasing Prestige
R2: Does the structure of the faculty hiring network affect the spread of ideas? 1.0 Fraction of Network Infected 0.8 p Epidemic Size 0.10 0.6 Stanford 0.30 0.50 0.4 0.70 Montana State 0.90 0.2 Colorado 0.0 0 50 100 150 University Prestige ( π ) Increasing Prestige
R2: Does the structure of the faculty hiring network affect the spread of ideas? 1.0 Fraction of Network Infected 0.8 p Epidemic Size 0.10 0.6 0.30 0.50 [Expected] 0.4 0.70 0.90 0.2 0.0 0 50 100 150 University Prestige ( π ) Increasing Prestige
R2: Does the structure of the faculty hiring network affect the spread of ideas? 1.0 Fraction of Network Infected 0.8 Infection Rate p (“Idea Quality”) p Epidemic Size 0.10 0.6 0.1 0.30 0.3 0.50 0.5 0.4 0.70 0.7 0.90 0.9 0.2 0.0 0 50 100 150 University Prestige ( π ) Increasing Prestige
R2: Does the structure of the faculty hiring network affect the spread of ideas? 1.0 Fraction of Network Infected Great ideas can spread 0.8 regardless of starting place Infection Rate p (“Idea Quality”) p Epidemic Size 0.10 0.6 0.1 0.30 0.3 0.50 0.5 0.4 0.70 0.7 0.90 0.9 0.2 0.0 0 50 100 150 University Prestige ( π ) Increasing Prestige
R2: Does the structure of the faculty hiring network affect the spread of ideas? 1.0 Fraction of Network Infected Great ideas can spread 0.8 regardless of starting place Infection Rate p (“Idea Quality”) p Epidemic Size 0.10 0.6 0.1 0.30 0.3 0.50 0.5 0.4 0.70 0.7 0.90 0.9 0.2 Good ideas spread more easily from high-prestige 0.0 universities 0 50 100 150 University Prestige ( π ) Increasing Prestige
Conclusion Ideas spread in academia via faculty hiring. The structure of this network privileges elite institutions. Good ideas can spread further and faster from prestigious universities, but great ideas can spread from any university. Future work should consider other (non-meritocratic) mechanisms, as well as the full text of research papers or other research ideas. Remaining questions: How should we address this inequality?
Recommend
More recommend
