Situation Testing-Based Discrimination Discovery: A Causal Inference Approach

Authors: Lu Zhang, Yongkai Wu, Xintao Wu

IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Through empirical assessments on a real dataset, our approach shows good efficacy both in accuracy and efficiency. To evaluate the proposed discrimination discovery algorithm, we have conducted experiments by using the Dutch Census of 2001
Researcher Affiliation Academia Lu Zhang, Yongkai Wu, and Xintao Wu University of Arkansas {lz006,yw009,xintaowu}@uark.edu
Pseudocode Yes Algorithm 1: Discrimination Discovery (CBN-DD)
Open Source Code No The paper does not provide any explicit statement about releasing open-source code for its methodology, nor does it provide a link to a code repository.
Open Datasets Yes To evaluate the proposed discrimination discovery algorithm, we have conducted experiments by using the Dutch Census of 2001 [Netherlands, 2001]. The reference [Netherlands, 2001] points to: Statistics Netherlands. Volkstelling. https://sites.google.com/site/faisalkamiran/, 2001.
Dataset Splits No The paper describes how test data was constructed ('We randomly select 200 tuples as the targets for discrimination testing.', 'We manually modify the dataset to obtain a data with ground truth...'), but it does not specify any training or validation dataset splits (e.g., 80/10/10 split or k-fold cross-validation for training).
Hardware Specification No The paper mentions 'The average CPU time for one target tuple is 0.3s for CBNDD and 20.3s for KNN-DD' but does not specify any particular CPU model, memory, or other hardware details.
Software Dependencies No The CBN is constructed by TETRAD [Glymour and others, 2004], an open-source platform for causal modeling. We employ the original PC algorithm [Spirtes et al., 2000]. The paper mentions software tools used (TETRAD, PC algorithm) but does not provide specific version numbers for them.
Experiment Setup Yes The threshold is set as 0.05. We employ the original PC algorithm [Spirtes et al., 2000] and the significance level = 0.05 for the structure learning. The paper also discusses varying the parameter K (e.g., K=10, 50, 90).