Factors Behind AI System Non-Development and Abandonment

A recent study published on arXiv examines the reasons behind the failure to build or the abandonment of AI systems prior to their deployment. This research performs a scoping review of existing academic literature, resources from civil society, and grey literature, which encompasses journalism and industry reports. Utilizing thematic analysis, it identifies six categories of influencing factors: ethical issues, feedback from stakeholders, challenges in the development lifecycle, organizational dynamics, limitations in resources, and legal complications. The findings emphasize that early-stage decisions significantly influence which systems make it to release, highlighting a critical area for intervention. The paper seeks to shed light on pre-deployment decision-making, a facet often neglected in responsible AI research that usually concentrates on systems that have been deployed.