Unfair Inequality in Education: A Benchmark for AI-Fairness Research

This dataset proposes a novel benchmark specifically designed for AI fairness research in education. It can be used for challenging tasks aimed at improving students' performance and reducing dropout rates which are also discussed in the paper to emphasize significant research directions. By prioritizing fairness, this benchmark aims to foster the development of bias-free AI solutions, promoting equal educational access and outcomes for all students.

Structure

benchmark contains:

• the proposed dataset (dataset.csv), 
• the mask for dealing with missing values (missing_mask.csv), and
• the meta-columns providing grouping criteria and sample weights for each student (meta_cols.csv).

raw_data includes:

• the original dataset (original.csv), and
• the intermediate stages of the pre-processing and validation pipelines (split, pre_processed, and validation).

res contains the documentation, including:

• the transformation mapping each column of the original dataset to the proposed one, along with the missingness category and original text (meta_data_mapping.csv),
• the value type and domains of each column of the proposed datasets (meta_data_stats.json), and
• the statistical indices of the validation pipeline  (bias_preservation_results.json).

src contains the source code for running the pre-processing and corresponding analysis:

• pre_processing and statscontain the code for the two corresponding tasks, and
• pre_processing.py and split.py are two entry points.

Finally, Dockerfile and requirements.txt set up the environment for running the applications across multiple platforms and with Python, respectively.

The synthetic_data file contains a synthetic copy of the same length as the original data set.