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

Problem
This preprint addresses a significant gap in the literature regarding the factors influencing the non-development or abandonment of AI systems. While existing responsible AI research predominantly focuses on the impacts of deployed systems, there is limited understanding of the pre-deployment decision-making processes that lead to the abandonment of AI projects. This work aims to illuminate the early-stage factors that shape whether AI systems are pursued or ultimately discarded, thereby identifying critical intervention points in the AI development lifecycle.

Method
The authors conduct a scoping review of diverse sources, including academic literature, civil society resources, and grey literature such as journalism and industry reports. They employ thematic analysis to develop a taxonomy categorizing six primary factors contributing to AI abandonment: ethical concerns, stakeholder feedback, development lifecycle challenges, organizational dynamics, resource constraints, and legal/regulatory concerns. To substantiate their findings, they analyze real-world cases of AI system abandonment using an AI incident database and a practitioner survey. This dual approach allows for a comparative analysis of factors influencing abandonment both before and after deployment, providing empirical evidence to support their taxonomy.

Results
The study reveals that while ethical risks are often highlighted in responsible AI discussions, the empirical analysis indicates that a broader range of factors—many unrelated to ethics—significantly influence organizations’ decisions to abandon AI development. The authors do not provide specific quantitative results or effect sizes in the abstract, but they emphasize the diversity of levers that drive abandonment, suggesting that ethical considerations are just one of many influences. The findings challenge the prevailing narrative in responsible AI literature and call for a more nuanced understanding of the motivations behind AI project abandonment.

Limitations
The authors acknowledge that their analysis is limited by the scope of the literature reviewed and the potential biases in the AI incident database and practitioner survey data. They do not address the potential for selection bias in the cases of abandonment they analyzed, nor do they explore the long-term implications of these abandonment decisions on the AI ecosystem. Additionally, the taxonomy developed may not encompass all possible factors influencing abandonment, as it is based on the sources reviewed.

Why it matters
This research has significant implications for the responsible AI community, as it highlights the need to engage with a broader array of factors influencing AI development decisions. By identifying non-ethical levers that lead to abandonment, the study encourages researchers and practitioners to consider a more comprehensive framework for understanding AI system development and abandonment. This could lead to improved strategies for supporting organizations in making informed decisions about AI projects, ultimately fostering a more responsible and effective approach to AI deployment.

Authors: Shreya Chappidi, Jatinder Singh
Source: arXiv:2604.28053
URL: https://arxiv.org/abs/2604.28053v1