The New Age Cyborgs: Mitigating Socio-Technical Biases in Generative AI Applications

Over the past few years, the resurgence of Generative AI (Gen AI) has led organizations to implement a broad variety of applications based on Gen AI However, when used for decision-making or content generation, Gen AI, similar to AI suffers from biases rooted in unrepresentative datasets, inadequate models, flawed algorithms, social and human stereotypes. These biases could lead to incorrect decision-making, miscommunication, and inappropriate outcomes if not accounted for adequately. This paper identifies mitigating mechanisms for addressing socio-technical biases that could creep in when organizations develop and implement Gen AI applications. Based on semi-structured interviews of 15 experts, we outlined 26 mechanisms, to address socio-technical biases. We also highlight the role of transparency and fairness and ways to achieve this for developing Gen AI applications. These findings contribute to a critical yet not fully explored aspect of socio-technical biases and address socio-technical biases in developing and implementing Gen AI applications.