Production of Knowledge and Evolution of Culture

Transcription

1 Production of Knowledge and Evolution of Culture Science Law How Does Science Progress? A Statistical Approach to Postmodern Theories of Knowledge The Geneology of Ideology: Identifying Persuasive Memes in the U.S. Circuit Courts Does Familiarity Breed Agreement? Evidence from Repeated Random Assignment in U.S. Circuit Courts Normative Commitments

2 Google Ngrams The word "originalism" was coined by Paul Brest in Here is the famous passage: "By originalism I mean the familiar approach to constitutional adjudication that accords binding authority to the text of the Constitution or the intentions of its adopters."

3 Originalism mean_distyear year

4 Law: Genealogy of Ideology Challenging to estimate peer effects Homophily - sorting/selection Confounding - external stimuli Simultaneity - reverse causality Common law, wikipedia, science, facebook, twitter Random assignment Shapes social and economic outcomes Clear ideological component Law evolves towards efficiency (Posner) Evolution leads towards Kantian morality (Alger and Weilbull)

5 Data

6 Data 388,000 cases since 1880 Citation graph 5 million edges denoting a case that cites another case Vote alignment 2.2 million pairs N-grams 2 billion N-grams up to length 8 5% random sample 426 hand-coded features 268 judges 269 features 6,000 cases in 25 legal areas hand-coded for meaning

7 Circuit Courts Circuit Map.ai -- Page 1 3/2

8 Peer Effects Figure 1: Dissent Rate across the Electoral Cycle ( ) 13% Presiden:al Elec:on 12% 11% Dissent Rate (Panel-Level) 10% 9% 8% 7% 6% 5% 4% Number of Quarters Un:l Next Presiden:al Elec:on

9 Peer Effects Figure 2: Judicial Panel Characteris<cs and Electoral Cycles in Dissents Last Quarter - Divided Last Quarter - Not Divided Not Last Quarter - Divided Not Last Quarter - Not Divided 0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% 3.5% 4.0% 4.5% Dissent Rate (Judge-Level)

10 Peer Effects Judicial Peer Effects: The Effect of Being on Unified Panels (1) (2) (3) (4) (5) Dissent Vote Mean of dep. var Majority *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) Unified *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) % Unified in Last Quarter * Minority ( ) ( ) ( ) ( ) ( ) % Unified in Last Quarter *** *** *** * *** * Majority ( ) ( ) ( ) ( ) ( ) % Unified in Last Quarter *** *** *** *** * Unified ( ) ( ) ( ) ( ) ( ) Circuit-Year Quarter-Party FE Yes No Yes Yes Yes Judge FE No Yes No No No Legal Issues FE No No Yes Yes Yes Sample All All All Experience 10 y Experience > 10 y Observations R-squared Notes: Robust OLS standard errors clustered at the circuit-year level in parentheses (* p < 0.10; ** p < 0.05; *** p < 0.01). Polarization Disassimilation

11 Peer Effects Judicial Peer Effects: The Effect of Being Minority on Divided Panels (1) (2) (3) (4) (5) Dissent Vote Mean of dep. var Majority *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) Unified *** *** *** *** *** ( ) ( ) ( ) ( ) ( ) % Minority in Last Quarter * Minority ( ) ( ) ( ) ( ) ( ) % Minority in Last Quarter * Majority ( ) ( ) ( ) ( ) ( ) % Minority in Last Quarter *** *** *** *** * Unified ( ) ( ) ( ) ( ) ( ) Circuit-Year Quarter-Party FE Yes No Yes Yes Yes Judge FE No Yes No No No Legal Issues FE No No Yes Yes Yes Sample All All All Experience 10 y Experience > 10 y Observations R-squared Notes: Robust OLS standard errors clustered at the quarter-year level in parentheses (* p < 0.10; ** p < 0.05; *** p < 0.01). Repulsion

12 Validating the N-grams Predict hand-coded cases Delete N-grams which appear in no more than a threshold based on the number of observations in each individual legal field Logistic AUC (better for imbalanced data - plots true positive over false positive rate)

13 Validating the N-grams

14 Validating the N-grams

15 Validating the N-grams

16 Validating the N-grams

17 Validating the N-grams

18 Constructing Memes Legal terminology 2-grams: AN, NN, VN, VV, NV, VP. 3-grams: NNN, AAN, ANN, NAN, NPN, VAN, VNN, AVN,VVN, VPN,ANV,NVV,VDN, VVV, NNV, VVP,VAV,VVN, NCN,VCV, ACA, PAN. 4-grams: NCVN, ANNN, NNNN, NPNN, AANN, ANNN, ANPN, NNPN, NPAN, ACAN, NCNN, NNCN, ANCN, NCAN, PDAN, PNPN, VDNN, VDAN, VVDN. Where, A: Adjective, N: Noun, V: Verb, P: Preposition, D: Determinant, C: Conjunction. 350,000 N-grams Phrase that propogates along the citation graph or seating graph or citation meme that propogates

19 Scoring Memetic Phrases

20 Memetic Phrases

21 Agreement

22 Agreement

23 Agreement: Case Level

24 Agreement: Judge Pair

25 Agreement: Logistic

26 Production of Knowledge and Evolution of Culture How Does Science Progress? Objective and subjective bodies of knowledge have different citation graphs Is Ideology Infectious? Identify three types of peer effects Polarization - sitting with like-minded judges causes unity against opposing-minded Disassimilation - sitting with like-minded judges causes disagreement with each other Repulsion - sitting with opposing-minded causes unity with like-minded Certain phrases predict opinion ideology Certain phrases appear memetic Identify causal effects of memes via random assignment of judges Normative Commitments Automatic coding of legal cases makes feasible rapid impact analyses of court decisions