AI Ethics Principles in Practice: Perspectives of Designers and Developers

The "principles-to-practice gap" is real: While AI ethics principles are widely agreed upon, designers and developers at a major research institution struggle to translate these high-level ideals into concrete, actionable steps during actual system development.

Ethical principles often conflict with each other: The research reveals significant tensions and trade-offs between the eight Australian Government AI ethics principles—for example, transparency requirements may sometimes conflict with privacy protection or security needs—requiring practitioners to make difficult choices without clear guidance.

Support mechanisms are inadequate: Current support systems for implementing AI ethics principles are insufficient, suggesting that organizations need better tools, frameworks, and resources to help their teams navigate ethical decision-making in practice.

Multiple principles require simultaneous attention: Responsible AI development demands balancing eight distinct ethical considerations—from privacy and fairness to human-centered values and environmental wellbeing—rather than focusing on any single principle, making implementation complex for real-world teams.

Practitioner insights are essential for closing the gap: The experiences of actual AI designers and developers reveal practical challenges that high-level principle documents don't address, indicating that future AI ethics guidance should be co-developed with or informed by the people building these systems.