In order to reduce the effect of race on the accuracy of skin cancer diagnosis, we created a model trained on an image dataset which neutral to skin tone and color.
A website where users can see how potentially biased data could affect AI in predicting the likelihood of certain demographic groups to be victims of police shootings or homicides
This MATLAB program classifies mirrored digits. For example, the program attempts to determine if the number "3" is equivalent to its mirrored version, "Ɛ." Utilizes the Deep Learning Toolbox.
Curated own dataset of bias/unbiased books from Project Gutenberg. Trained W2V models on the biased and unbiased datasets. Performed WEAT analysis of standard sample & compared to Google W2V AI sample
The Bias Buster Policing Needed
Unstructured internet data can lead to biased AI, pushing stereotypes and eroding trust. Biases are often seen in image searches related to different professions, resulting in skewed representation
The project is focused on developing a model that can accurately and efficiently analyze genetic data by utilizing ML model, predict outcomes and also aims to explain inherent bias.
Stopping the Hurt with Words: Using NLP to Combat Cyberbullying
Have the accurate news in accurate time and accurate place
1 – 9 of 9
