Algorithmic Accountability as a Virtue or a Mechanism? The Ethical Divide Among AI Developers

Abstract

Algorithmic accountability is gaining prominence, driven by the ethical challenges of increasingly advanced information systems (IS) based on artificial intelligence (AI). Legal and practical initiatives often lack a clear accountability definition, leaving AI developers to develop their own understanding. In our qualitative study, we interviewed 17 AI developers to explore how their ethical orientations affect their understanding of algorithmic accountability and its professional and personal effects. Our findings indicate that consequentialist-oriented AI developers typically understand accountability as a mechanism for quality assurance in AI development, leading to operational impacts. Conversely, deontological-oriented AI developers tend to understand algorithmic accountability as a virtue they and their developed AI systems must live up to, often with significant ethical implications. Our study contributes to IS research by clarifying how ethical orientations shape algorithmic accountability understandings as either a mechanism or a virtue, which is crucial for effective management and communication in AI development projects.