‘It wasn’t me’: the impact of social responsibility and social dominance attitudes on AI programmers’ moral imagination (intention to correct bias)

Abstract A plethora of research has shed light on AI’s perpetuation of biases, and the primary focus has been on technological fixes or biased data. However, there is deafening silence regarding the key role of programmers in mitigating bias in AI. A significant gap exists in the understanding of how a programmer’s personal characteristics may influence their professional design choices. This study addresses this gap by exploring the link between programmers’ sense of social responsibility and their moral imagination in AI, i.e., intentions to correct bias in AI, particularly against marginalized populations. Furthermore, it is unexplored how a programmer’s preference for hierarchy between groups, social dominance orientation-egalitarianism (SDO-E), influences this relationship. We conducted a between-subject online experiment with 263 programmers based in the United States. They were randomly assigned to conditions that mimic narratives about agency reflected in technology determinism (low responsibility) and technology instrumentalism (high responsibility). The findings reveal that high social responsibility significantly boosts programmers’ moral imagination concerning their intentions to correct bias in AI, and it is especially effective for high SDO-E programmers. In contrast, low SDO-E programmers exhibit consistently high levels of moral imagination in AI, regardless of the condition, as they are highly empathetic, allowing the perspective-taking needed for moral imagination, and are naturally motivated to equalize groups. This study underscores the need to cultivate social responsibility among programmers to enhance fairness and ethics in the development of artificial intelligence. The findings have important theoretical and practical implications for AI ethics, algorithmic fairness, etc.