Embed Size (px)
Citation preview
Quantifying the evolution of individual scientific impact
Roberta Sinatra @robysinatra
Is a scientific career predictable?
Who will make an outstanding discovery, and when?
Radicchi et al, Phys. Rev. E 80 (2009) Petersen et al, PNAS, 108 (2011)
500k over 110 years
5M citations
Physics
6 fields
~25,000 careers
~500,000 papers
Deville, Wang, Sinatra, Song, Barabási, Sci. Rep. 4:4770, (2014)
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
We study scientific impact through publications
0 10 20 300
100
200
300
400
time (in years)
c (c
itatio
ns)
Career of K.G. Wilson (Nobel in Physics, 1982)
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
0 10 20 300
100
200
300
400
time (in years)
c (c
itatio
ns)
Career of K.G. Wilson (Nobel in Physics, 1982)
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
0 10 20 300
100
200
300
400
time (in years)
c (c
itatio
ns)
t⇤
Career of K.G. Wilson (Nobel in Physics, 1982)
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
0 10 20 300
100
200
300
400
time (in years)
c (c
itatio
ns)
t⇤
c⇤
Career of K.G. Wilson (Nobel in Physics, 1982)
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
t⇤
P(t
⇤ )
0 10 20 30 400
0.05
0.1
0.15
0.2
DataRandomized
Simonton, D. F. (1997). Psychological Review 104, 66.
Jones, B. F., & Weinberg, B. A. (2011). PNAS, 108(47), 18910-18914.
Timing of the hit
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
0 10 20 300
100
200
300
400
time (in years)
c (c
itatio
ns)
Randomization: we shuffle the impact of papers
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
0 10 20 30 400
0.05
0.1
0.15
0.2
DataRandomized
t⇤
P(t
⇤ )
The hit is random in a scientist’s sequence of publications
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
The hit is random in a scientist’s sequence of publications
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
Impact is random within a scientific career
There is always hope!
Frank G. Wilczek Physics Nobel, 2004
John B. Fenn Chemistry Nobel, 2002
Careers look different
time (in years)1 10 20 30
050
100150200250300 S=1.49
c 10
time (in years)1 10 20 30
050
100150200250300 S=3.31
c 10
1 10 20 300
50100150200250300
time (in years)
S=7.24
c 10
We want to untangle productivity, individual ability, and luck
impact of j’s paper = luck * researcher Qcj,↵ = p↵Qj
Modelling individual careers: Q-model
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
c 10
0 10 20 300
100
200
300
400
500
c 10
0 10 20 300
200
300
400
500
100
cj,↵ = p↵Qj
Q = 10
Q = 2
Modelling individual careers: Q-model
Sinatra, Wang, Deville, Song, Barabási, in review (2016)
are decoupled
P (p)
cj,↵ = p↵Qj
P (Q)
Modelling individual careers: Q-model
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
1 10 20 300
50100150200250300
time (in years)
S=7.24
c 10
time (in years)1 10 20 30
050
100150200250300 S=1.49
c 10
time (in years)1 10 20 30
050
100150200250300 S=3.31
c 10
S(∆
N)
career0 0.25 0.5 0.75 10
246810
S(∆
N)
career0 0.25 0.5 0.75 10
246810
S(∆
N)
career0 0.25 0.5 0.75 10
246810
Prediction 1: the Q-parameter is stable throughout a career
Q = 1.49 Q = 3.31 Q = 7.24Q(�
N)
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
is universalP (p)
P(p)
100 101 1020
0.2
0.4
0.6
0.8
1
P(�
c)
DATA
P(p)
100 101 1020
0.2
0.4
0.6
0.8
1
RESCALED DATA
Prediction 2: the Q-model untangles luck and researcher Q
c
Q
P
✓�
c Q
◆Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
0 100 200 300050
100150200250300350
⟨c∗ 10⟩
S = 1.2S = 4.0S = 10.0
N0 100 200
101
102
103
104
S = 1.2
0 100 200
S = 5.3
NN
ObservedPredicted
Ctot(N
)
Q = 1.2 Q = 5.3
Q =
Q =
Q =
Prediction 3: the Q-model analytically predicts the dynamic of impact indicators
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
Prediction 4: the Q-parameter has early predictive power
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
Prediction 5: the Q-parameter detects nobel laureates better than other metrics
c 10
c 10
c 10
c∗10
1 10 5000
0.2
0.4
0.6
0.8
1
1.6
Q
Increasing QA
B C D
2.7
4.5
7.8
13.3
100 101 1020
0.2
0.4
0.6
0.8
1
100 101 1020
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 10
0.2
0.4
0.6
0.8
1
Other scientists
Nob
el La
urea
tes
scientist , 0.94h−index, 0.90highest impact , 0.88
productivity N, 0.65total citations , 0.86
1 10 20 300
50100150200250300
1 10 20 300
50100150200250300
1 10 20 300
50100150200250300
time (in years)
Q=9.99
Q=3.31
Q=1.49
Ctot
P(≥
c 10,i/Q
i)
c10,i/Qic10,i
P(≥
c 10,i)
c10,i
P(≥
c 10,i)
P(p)
P(p)
Q
Sinatra, Wang, Deville, Song, Barabási, Science, 354, 6312 (2016)
Who will make an outstanding discovery, and when?
Is a scientific career predictable?
Who will make an outstanding discovery, and when?
Lucky scientists with high Q, randomly in their career
Is a scientific career predictable?
Who will make an outstanding discovery, and when?
Lucky scientists with high Q, randomly in their career
Yes, after we quantify the luck component
Is a scientific career predictable?
https://www.youtube.com/watch?v=qlnxM-ld4BU
Acknowledgements
Pierre Deville
Albert-László Barabási
Yifang Ma
Junming Huang
Michael SzellChaoming Song
Dashun Wangwith
thanks to
refs
Sinatra et al. Nature Physics 11, 791-796 (2015)
Deville et al. Sci. Rep. 4:4770, (2014) Wang, et al. Science 342, 6154 (2013)
Szell and Sinatra, PNAS 112:48, 14749-14750 (2015)
Kim Albrecht
Mauro Martino
Sinatra, et. al, Science, 354, 6312 (2016)
Shameless plug
More info on: cns.ceu.edu
Join us!
Open Call for PhD in Network Science Deadline: February 1st, 2017
