To Build or Not to Build? Factors that Lead to Non-Development or Abandonment of AI Systems

Ethics alone doesn't drive abandonment decisions: While responsible AI research emphasizes ethical concerns as reasons to halt AI development, the paper's empirical analysis of real-world cases shows that organizations abandon AI systems for diverse reasons—often non-ethics-related factors like resource constraints, organizational dynamics, and legal/regulatory concerns prove equally or more influential.

Early-stage decisions are critical intervention points: The paper highlights that decisions made during pre-deployment development stages shape which AI systems ultimately reach users, yet these earlier phases receive limited attention from responsible AI researchers compared to post-deployment impact studies. This represents an underexplored opportunity to prevent problematic systems before they're released.

Six key factors influence whether AI gets built: The research identifies a taxonomy of six categories driving AI non-development and abandonment: ethical concerns, stakeholder feedback, development lifecycle challenges, organizational dynamics, resource constraints, and legal/regulatory concerns—suggesting that responsible AI advocates should engage with multiple levers beyond just ethical arguments.

Stakeholder feedback and practical challenges matter: Beyond ethics, the paper's findings suggest that real-world abandonment is significantly influenced by feedback from affected communities, technical development difficulties, and regulatory pressures—indicating that responsible AI strategies should incorporate diverse stakeholder voices and address practical implementation barriers.

Pre- and post-deployment abandonment differ: The research compares factors driving abandonment before versus after system deployment, suggesting that the timing of intervention affects which factors are most relevant and actionable for preventing harmful AI systems from reaching users.