Directories
- resume_without_bias: This directory contains the original dataset and the results after querying the API using the original dataset.
- resume_with_bias: This directory contains the biased datasets created on the original dataset containing uneven feature distributions.
Files/Subdirectories
-
resume_without_bias/resume.csv: This file contains the original dataset.
-
resume_without_bias/resume_result.csv: This file contains the resume scorer output of resume.csv.
-
resume_without_bias/resume_with_score.csv: This file contains same information as resume.csv but with last column the resume scores.
-
resume_without_bias/candidate_result.csv: This file contains candidate evaluator output for resume_with_score.csv.
-
resume_with_bias/authorization_bias: This directory contains the datasets with work authorization biases and API results.
-
resume_with_bias/disability_bias: This directory contains the datasets with disability status biases and API results.
-
resume_with_bias/ethnicity_bias: This directory contains the datasets with ethnicity biases and API results.
-
resume_with_bias/gender_bias: This directory contains the datasets with gender biases and API results.
-
resume_with_bias/veteran_bias: This directory contains the datasets with veteran status biases and API results.
-
generate_data.ipynb: This Jupyter Notebook contains the code that was used to generate all the datasets (original & biased).
-
merge_resume_score.ipynb: This Jupyter Notebook contains the code that was used to merge resume scores output from the API to the features set to form input to the candidate evaluator API.
-
analysis_bias.ipynb: This Jupyter Notebook contains the code for the analyses we did, including calculations of various metrics and plots.
-
report/audit_report.pdf: This is the final auditing report for this project.
-
report/correlation.png: This is a plot used in the auditing report.