About Bias Insights

Bias Insights is a small Statistical Analysis Project for bias detection and demo:

In this project I will be using the Chicago Face Database as my main dataset. After observing the dataset, I have formulated the following hypothesis:

"The Chicago Face Database, while comprehensive, appears to have an underrepresentation of individuals aged 60 and above. This observed age bias could potentially impact the performance of machine learning models trained on this dataset, particularly in tasks involving the generation or recognition of faces belonging to this age group. Consequently, it is hypothesized that a model trained on this dataset may exhibit decreased accuracy and reliability when applied to faces of individuals older than 60 years old. This project aims to statistically validate this hypothesis and explore potential mitigation strategies to address this bias, thereby enhancing the model’s performance across all age groups."

To demonstrate whether my hypothesis is right or not I use a statistical analysis thecnique called: chi-square test

A chi-square test is used to compare the observed frequencies with the expected frequencies. The Chi-Square test will give you a p-value that you can use to determine whether the difference between the observed and expected frequencies is statistically significant.