Social Bias in AI and its Implications

Previous studies have documented many different types of biases that exist in artificial intelligence (AI) and machine learning (ML) systems. We reviewed the literature on AI and ML bias with a focus on social implications and found that bias in AI and ML can potentially have harmful social impacts on individuals and/or groups of people. By affecting people differently according to characteristics such as race, gender, or sexual orientation, AI and ML systems may lead to harm by exacerbating social inequities. We recount examples of issues that have occurred in systems that use technology that might be used at NASA and elsewhere so that similar issues might be identified and mitigated in future systems. We also provide interested parties with a gateway into existing work on social bias in AI and ML systems.