AI and Gender Equality: One Step Forward or Two Steps Back?

Abstract

This research focuses on potential gender biases in AI algorithms and their implications for gender equality. The recent sudden and unexpected growth in the artificial intelligence (AI) user base has attracted significant attention. It has raised public concerns about AI's impact on jobs becoming redundant in the future, and about the accuracy of AI’s content and its power to influence and manipulate users. However, the AI-driven transformation began earlier and yielded several ongoing changes in organizations – for example, enhancing service quality, reducing costs in routine tasks, and improving access to existing information. Nevertheless, from the gender perspective (one of the UN’s sustainable development goals), there is a risk that this AI-driven transformation could negatively impact the complex journey toward equality between men and women. This may occur either from women’s reduced involvement in the technology development processes (which can lead to developer biases being reflected in the technologies) or because the historical data used to inform and develop AI algorithms may reinforce existing gender biases. We address the challenge proposed by Manasi et al. (2023) to test AI algorithms in this context. Using an experimental approach, our research examines how three popular AI tools (ChatGPT, CoPilot, and Gemini) behave in relation to sensitive gender-related topics and scenarios. This study aims to clarify the associated risks and assist stakeholders, particularly policymakers, in defining guidelines to prevent the replication of gender-biased content. Nonetheless, we believe that AI can serve as a resource to advance the field of gender equality. Given the recent massification of AI use and the recent challenges to progress on gender equality (including the asymmetric impact of COVID-19 and recent armed conflicts), this research is timely and carries direct implications for organizations’ managers, policymakers, and the research community. We position it as exploratory research and propose avenues for future research, and we highlight the need to monitor the evolution of AI by regularly conducting similar analyses.