ATLANTA — The goal of most new technology is to make life more efficient and safe, but a group of Georgia Tech researchers believe self-driving cars could be dangerous for minority groups.
Georgia Tech released research that suggested pedestrians with darker skin may be more likely to get hit by self-driving cars than those with lighter skin.
“Companies don’t want the public to know about any issues of inaccuracy, so consumers need to learn to ask a lot of questions,” Jamie Morgenstern, School of Computer Science assistant professor, and the study’s lead author, said in a news release.
Researchers said they tested machine learning object detection models to see how well they could see people with different skin tones. Models were nearly 5 percent less likely to detect darker-skinned pedestrians, the study revealed.
Georgia Tech researchers claimed the imbalance remained, regardless of how they accounted for variables - such has the time of day, the pixel size of the person, or partially blocked views of pedestrians.
However, they also noted the biases with the training dataset, which they said had more examples of people with lighter skin tones.
Despite the bias, the group who studied the models said they made adjustments and was able to correct the lack of fairness by reweighing the model to better analyze smaller groups.
It is important to note, however, the researchers in this study didn't investigate ML models and training data actually used by the self-driving car industry because they said those are not publicly available.
Read the full study here.
MORE NEWS |