DISCIPLINARY AND INTERDISCIPLINARY RESEARCH AT NSF Myron Gutmann Leah Nichols COSSA Colloquium 2012 November 2012 1
ACKNOWLEDGEMENTS Dave Newman, University of California, Irvine Julia Lane, American Institutes for Research (previous SciSIP Program Director) Science of Science (Sci 2 ) Tools. Indiana University and SciTech Strategies, http://sci2.cns.iu.edu Science of Science and Innovation Policy (SciSIP) Program Amber Story, Amy Friedlander, and many others at NSF 2
STARTING POINTS Science increasingly operates at intersections Most significant societal problems cannot be solved inside a single disciplinary framework Health, energy-environment, education, etc. Broad recognition of need for change within government, universities, media Lots of change is underway 3
Future research in the SBE sciences will be multidisciplinary, collaborative, data intensive, and frequently problem oriented. 4 4
TWO KEY DISTINCTIONS Science as a representation of an interaction of Ideas People Institutions (sponsors, journals, departments, labs, centers, universities) Interdisciplinarity at NSF as a reflection of Concepts in proposals Investments by programs/divisions/directorates through co-funding 5
6 BACKGROUND: HOW NSF STRUCTURES ITS ACTIVITIES Core Disciplinary Programs Economics, Political Science, Sociology Multi-Disciplinary Programs Decision, Risk, and Management Science (DRMS), Science and Technology Studies (STS), Science of Science and Innovation Policy (SciSIP) Targeted Interdisciplinary Programs Dynamics of Coupled Natural Human Systems (CNH), Methodology, Measurement, and Statistics (MMS) Open Ended Interdisciplinary Programs INSPIRE, Postdoctoral Research Fellowships, Interdisciplinary Research in Behavioral & Social Sciences
METHODS: DATA 14,225 Awards in SBE s portfolio, 2000-2011 NSF Budget Data NSF Topic Model (Portfolio Explorer) 7
METHODS: COFUNDING ANALYSIS Research directorates and SBE Core Programs equated with disciplines IDR inferred when award is funded by more than one directorate or program Degree of interdisciplinarity calculated by weight of contributions from each directorate or program Non-Disciplinary Specific = Non-research or interdisciplinary directorates and offices 8
PROGRAM COFUNDING: 2000-2011 (BY AWARD COUNT) 100% 90% 80% 70% 60% 50% None None None None None None 40% 30% 20% 10% 0% 9 External External External Perception, Action and Cognition Decision Risk & Management Science Economics Program Geography and Spatial Sciences External Sociology Political Science
METHODS: TOPIC MODEL ANALYSIS (I) Topic Model: Statistical text-mining extracted 1,000 latent categories from NSF 2000-2011 awards and proposals Topics classified into disciplines according to normalized frequency of appearance in each directorate or SBE program. IDR inferred when the topic tags on an award fall into more than one discipline Degree of interdisciplinarity in each award calculated by weighting disciplines by topic tag 10 order
METHODS: TOPIC MODEL ANALYSIS (II) Non-disciplinary specific = topics found with similar frequency in multiple directorates (e.g. Hypothesis, Imaging, Students) SBE Non Specific = Topics found predominantly in SBE, but with similar frequency throughout SBE s Programs (e.g. Public Policy, Gender) 11
$ (Billions) 4.5 4 3.5 3 2.5 2 1.5 1 0.5 Disciplines in the SBE Portfolio: 2000-2011 (by funding amount) Engineering CISE Biology SBE Not Disciplinary Specific Geosciences Engineering CISE Biology SBE Lots of scientific interaction *** More when measured by words than institutional arrangements Biology SBE Not Disciplinary Specific Physics Mathematics Material Sciences Geosciences Engineering Computer and Information Sciences Chemistry Biology 0 Cofund Analysis (SBE + Cofund $) $4.2 Billion Topic Model Analysis (SBE + Cofund $) $4.2 Billion Topic Model Analysis (SBE $ Only) $2.4 Billion Astronmoical Sciences SBE 12 14,225 Awards for all three analysis
Disciplinary Co-occurrence within the SBE Portfolio (2000-2011) Not Disciplinary Specific (6998) Physics (223) Social, Behavioral, and Economic Sciences Astronomical Sciences (106) Biology (2564) Not Disciplinary Specific (966) Physics (28) Social, Behavioral, and Economic Sciences Astronomical Sciences (11) Biology (422) Mathematics (623) Chemistry (83) Mathematics (175) Chemistry (47) Material Sciences (132) Geosciences (746) Engineering (1647) Topic Model Analysis Computer and Information Sciences (1519) # of Awards 14,225 7,800 3,500 7,000 Material Sciences (27) # of Co-Occurrences 2,500 1,000 1 n = 14,225 (for both analyses) Geosciences (308) Co-fund Analysis Engineering (437) Computer and Information Sciences (403) 13
Interdisciplinarity by Award Size in the SBE Portfolio: 2000-2011 (by award count) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Topic Model Analysis None None None External External External Small Medium Large IDR Type None External 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Cofund Analysis None None None External External External Small Medium Large Award Size Definition % of portfolio (by count) Small < $100,000 50% Medium $100,000-$500,000 43% Large > $500,000 7% Includes full SBE Portfolio: 14,225 awards Award Size (based on full awarded amount) IDR Type: None = Single Discipline = More than one SBE discipline External = At least one non-sbe discipline
Decision, Risk and Management Science (2) Geosciences (7) Computer and Info. Sciences (0) Chemistry (2) Mathematics Physics (1) Anthropology (5) (10) Engineering Archaeology (0) (11) Biology Cognitive Science (8) Decision Sciences Economics (4) Geography (1) Law (3) (2) Engineering (0) Geosciences (3) Computer and Info. Sciences (0) Chemistry (2) Biology Sociology Physics (1) Anthropology (6) Archaeology (0) Cognitive Science (1) Decision Sciences (1) Economics (5) Geography (2) Law (3) (29) Not Disciplinary Specific (12) SBE Non Specific (1) Sociology (17) Social Psychology Learning Sciences (3) Linguistics (0) MMS (3) Political Science (2) STS (6) (15) Not Disciplinary Specific (18) SBE Non Specific Sociology (2) Social Psychology Learning Sciences (1) Linguistics (0) MMS (2) Political Science (7) STS (2) (1) Geosciences (28) Computer and Info. Sciences (17) Biology Perception, Action, Cognition (7) Mathematics Physics (2) (1) Material Science Anthropology (1) (11) Engineering Cognitive Science Decision Sciences (1) Economics (1) Geography and Spatial Sciences (0) Chemistry (2) Engineering (2) Geosciences (2) Computer and Info. Sciences (3) Biology Physics (0) (0) Mathematics Anthropology (3) Archaeology (0) Cognitive Science (0) Decision Sciences (0) Economics (2) Geography (40) Not Disciplinary Specific (9) SBE Non Specific 15 (1) Sociology (4) Social Psychology STS (4) Learning Sciences (6) Linguistics (14) MMS (0) (7) Not Disciplinary Specific (5) SBE Non Specific (1) Sociology Learning Sciences (0) Linguistics (1) MMS (1) Political Science (1) STS (2)
(2) Geosciences (4) Computer and Info. Sciences (5) Engineering Political Science (2) Mathematics Anthropology (4) Archaeology (1) Cognitive Science (2) Decision Sciences (8) Economics (21) (9) Computer and Info. Sciences (6) Mathematics (1) Material Science (7) Engineering (2) Geosciences Economics Physics (1) Anthropology (2) Archaeology Cognitive Science (1) Decision Sciences (9) Economics (5) Biology (44) Not Disciplinary Specific (31) SBE Non Specific (10) Sociology (3) Social Psychology STS (7) Geography (1) Law (6) Learning Sciences (1) MMS (12) Political Science (6) Biology (1) Astronomical Sciences (34) Not Disciplinary Specific (28) SBE Non Specific (2) Sociology (1) Social Psychology Geography (3) Law (2) Learning Sciences (1) MMS (13) Political Science (8) STS (6) 16 Network Characteristics DRMS Sociology PAC GSS Political Science Economics # of Nodes 23 22 19 21 20 23 Max. Node Weight 58 65 71 48 88 81 # of Edges 112 108 83 82 82 82 Max. Edge Weight 29 27 40 23 44 34 Average Degree 9.74 9.82 8.74 7.81 8.20 7.13 Density 0.345 0.332 0.255 0.252 0.252 0.252 % Interdisciplinary 91 93 94 99 78 76 (by Award Count) Number of Awards 478 1064 362 1118 845 1576
PROGRAM COFUNDING: 2000-2011 (BY AWARD COUNT) 100% 90% 80% 70% 60% 50% None None None None None None 40% 30% 20% 10% 0% 17 External External External Perception, Action and Cognition Decision Risk & Management Science Economics Program Geography and Spatial Sciences External Sociology Political Science
IDR RANKING BY ANALYSIS METHOD Decision, Risk, and Management Science Topic Model Analysis (Networks) Topic Model Analysis (% Award Count) Cofunding Analysis 1 4 2 Sociology 2 3 5 Perception, Action, Cognition Geography and Spatial Sciences 3 2 1 4 1 4 Political Science 5 5 6 Economics 6 6 3 18
Cognitive Science and Neuroscience at NSF (2007-2011) Measuring & Modeling Cognition Environmental Interactions, Behavior & the Nervous System Higher Order Cognition Language & Meaning Linguistics Language Speech Cognitive Science Human Motion Computer Vision Olfactory System fmri Stimulus & Response Neuroscience Human Vision Vision & Perception Hormones Neuroscience Electrophysiology Neurite Outgrowth Mutation Neural Development & Signaling Mechanisms Neural Circuits & Systems 355 100 Number of Awards Number of Co-occurrences 90 10 Perception & Action in Nature & Machines
Directorate Contributions to Cognitive and Neuroscience Clusters (2007-2011) 100% 90% 80% 70% 60% 50% MPS ENG EHR SBE ENG EHR EHR SBE ENG SBE MPS ENG EHR SBE CISE ENG SBE MPS ENG 40% 30% SBE CISE BIO BIO 20% 10% CISE CISE CISE BIO 0% Measuring and Modeling Cognition Complex Cognition Language and Meaning Perception and Action in Nature and Machines Neural Circuits and Systems Environmental Interactions, Behavior, and the Nervous System Neural Development and Signaling Mechanims BIO CISE SBE EHR ENG GEO MPS O/D
CONCLUSIONS Developing a single measure for interdisciplinarity masks the complexity of the concept. By any measure, the SBE portfolio is highly interdisciplinary. The topic model tool provides a novel way to assess IDR in the textual content of award proposals. Cofunding reflects IDR, but is affected by institutional realities. It underestimates IDR, especially in small- and medium-sized awards. 21
SO WHAT IS A DISCIPLINE? A set of topics to be studied, or a specific set of theoretical or methodological approaches? Research done by people in a specific departments? Research that can be funded by a specific NSF program? Research that is acceptable for publication in specific journals? 22 and what do we want disciplines to be in an evolving scientific world?
Thank You! mgutmann@nsf.gov lenichol@nsf.gov 23