Implications of Distance over Redistricting Maps: Central and Outlier Maps
Authors: Seyed A. Esmaeili, Darshan Chakrabarti, Hayley Grape, Brian Brubach
AAAI 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct experiments on three states, North Carolina (NC), Maryland (MD), and Pennsylvania (PA), which have featured in major court cases on partisan gerrymandering (LWV vs Commonwealth of Pennsylvania No. 159 MM 2018; Rucho v. Common Cause No. 18-422, 588 U.S. 2019). ... Figure 2 shows the unweighted distance histogram for NC. ... Using similar logic, we can also detect gerrymandered maps. Specifically, the 2011 and 2016 enacted maps of NC were widely considered to be gerrymandered, and both maps are at the right tail of the histogram, far from the centroid. |
| Researcher Affiliation | Academia | 1University of Chicago Data Science Institute 2Columbia University 3Wellesley College esmaeili@uchicago.edu, darshan.chakrabarti@columbia.edu, hg3@wellesley.edu, bb100@wellesley.edu |
| Pseudocode | Yes | Algorithm 1: Finding the Sample Medoid Input: MT = {A1, . . . , AT }, = { (i, j) > 0, 8i, j 2 V }. 1: Calculate the centroid map Ac = 1 T PT t=1 At. 2: Pick the map A 2 MT which minimizes the d2, distance from the centroid Ac, i.e. A = arg min A2MT d2, (A, Ac). return A |
| Open Source Code | No | The paper mentions that the 'Re Com' algorithm (a third-party tool) is available online, but there is no explicit statement or link indicating that the authors' own code for their proposed framework is open-source or publicly available. |
| Open Datasets | No | The paper describes characteristics of the data used (e.g., number of voting units for NC, MD, PA) and references their origin in redistricting cases, but it does not provide concrete access information such as specific links, DOIs, repositories, or formal citations for the raw geographical or election data used as input for the Re Com algorithm. |
| Dataset Splits | No | The paper describes sampling maps and using burn-in samples, but it does not specify explicit training, validation, and test dataset splits as typically defined for model reproduction or evaluation. |
| Hardware Specification | No | The paper does not provide specific hardware details such as GPU/CPU models, processor types, or memory amounts used for running its experiments. |
| Software Dependencies | No | The paper mentions the use of 'Re Com' (a third-party tool) but does not provide its version number or any other specific software dependencies with version details needed to replicate the experiment. |
| Experiment Setup | Yes | We sample 200,000 maps instead to estimate the centroid. ... we always exclude the first 2,000 samples from any calculation as these are considered to be burn-in samples. ... (1) Sample 200,000 maps and pick the one closest to the centroid Aclosest. (2) Start the Re Com chain from Aclosest but given a specific state (redistricting map) we only allow transitions to new states (maps) that are closer to the centroid, and we do this for 200,000 steps to obtain the final estimated medoid ˆA . |