AI lifecycle from a data-driven perspective: a systematic review

Abstract

Introduction. Revising AI lifecycle models has drawn the attention of scholars from different areas because of the advances in AI technology. Many AI lifecycle models have been proposed. However, no systematic review of current AI lifecycle models has been found. This study aims to review and synthesize AI lifecycle models in current literature from a data-driven perspective and recognize the roles of data in different stages of the lifecycle. Method. This study used the Preferred Reporting Items for systematic review and meta-analyses (PRISMA) protocol to systematically review AI lifecycle models from research papers, reports, and acts published between 2020 and 2024. Analysis. A qualitative approach was applied with a pre-specified categorization framework. Open coding, axial coding, and selective coding were used. Results. Twenty AI lifecycle models were identified. Stages and contents varied with overlaps and confused use of stage names. These models were proposed from the perspective of business objectives, AI model development, or a combination of implementing scenarios. Great importance has been attached to ethical issues for the whole AI lifecycle. Conclusions. A double-layer AI lifecycle model with three phases and ten stages is synthesized. Six roles of data are identified. Data documentation in the AI lifecycle has not been fully valued.

Keywords

• AI lifecycle
• data-driven
• data role
• systematic review