[DATA SCIENCE TRACK] PRESENTATION: Risk mitigation through model transparency in bias: improving model safety through decreasing bias

Bias in AI systems can lead to harmful outcomes. We investigate methods to increase model transparency and explainability in order to detect, understand, and mitigate risks from bias. Techniques like saliency maps, attention mechanisms, and adversarial testing can shed light on model behavior. Improving model transparency and reducing bias is key to developing safer, more trustworthy AI.