Egalitarian Committee Scoring Rules
Authors: Haris Aziz, Piotr Faliszewski, Bernard Grofman, Arkadii Slinko, Nimrod Talmon
IJCAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Following Elkind et al. [2017a], we evaluate our rules on the 2D Euclidean domain and present the results graphically. To our surprise, we find that sometimes egalitarian rules behave very similarly to utilitarian ones, defined using different scoring functions (in our experiments this happens, e.g., for k-Borda and egalitarian-Pessimist). For each of the settings and for each of our rules (both utilitarian and egalitarian), we drew 10,000 elections with 100 candidates and 100 voters each, and computed the winning committees of size 10 (breaking ties uniformly at random; for NP-hard rules we used the ILP formulations of Skowron et al. [2016], adapted to the egalitarian setting, if needed (we omitted utilitarian Pessimist as it was particular difficult to compute). Then, for each scenario, we partitioned our [ 3, 3] [ 3, 3] square into 120 120 equal-sized cells, and computed how many winners end up in each cell. We present the obtained histograms in Figure 1 (the darker a given point is, the more winners landed there; see the work of Elkind et al. [2017a] for details on drawing the histograms). |
| Researcher Affiliation | Academia | 1 UNSW Sydney and Data61 (CSIRO), Australia 2 AGH University of Science and Technology, Krakow, Poland 3 University of California, Irvine, USA 4 University of Auckland, Auckland, New Zealand 5 Ben-Gurion University, Be er Sheva, Israel |
| Pseudocode | No | The paper does not include pseudocode or clearly labeled algorithm blocks. |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described. |
| Open Datasets | No | The paper describes how candidate and voter points are generated for simulations (e.g., "uniform disc setting", "2-Gaussians setting"), but does not refer to a publicly available dataset or provide access information for the generated data. |
| Dataset Splits | No | The paper describes generating 10,000 elections for experiments but does not specify train, validation, or test dataset splits. The data is generated for each experiment rather than being split from a fixed dataset. |
| Hardware Specification | No | The paper does not provide any specific hardware details (e.g., CPU, GPU models, or cloud instance types) used for running the experiments. |
| Software Dependencies | No | The paper mentions using "ILP formulations of Skowron et al. [2016]" but does not specify any software names with version numbers for reproducibility (e.g., a specific ILP solver and its version). |
| Experiment Setup | Yes | For each of the settings and for each of our rules (both utilitarian and egalitarian), we drew 10,000 elections with 100 candidates and 100 voters each, and computed the winning committees of size 10 (breaking ties uniformly at random; for NP-hard rules we used the ILP formulations of Skowron et al. [2016], adapted to the egalitarian setting, if needed (we omitted utilitarian Pessimist as it was particular difficult to compute). Then, for each scenario, we partitioned our [ 3, 3] [ 3, 3] square into 120 120 equal-sized cells, and computed how many winners end up in each cell. |