Biases in AI-supported industry 4.0 research: a systematic review, taxonomy, and mitigation strategies
Industrial engineering research has been reshaped by the integration of artificial intelligence (AI) within the framework of Industry 4.0, characterized by the interplay between cyber-physical systems (CPS), advanced automation, and the Industrial Internet of Things (IIoT). While this integration op...
| Authors: | , , , , , |
|---|---|
| Format: | article |
| Status: | Published version |
| Publication Date: | 2025 |
| Country: | España |
| Institution: | Universidad Pública de Navarra |
| Repository: | Academica-e. Repositorio Institucional de la Universidad Pública de Navarra |
| OAI Identifier: | oai:academica-e.unavarra.es:2454/55885 |
| Online Access: | https://hdl.handle.net/2454/55885 |
| Access Level: | Open access |
| Keyword: | Industrial AI CPS IIoT Digital twin Predictive maintenance Quality control |
| Summary: | Industrial engineering research has been reshaped by the integration of artificial intelligence (AI) within the framework of Industry 4.0, characterized by the interplay between cyber-physical systems (CPS), advanced automation, and the Industrial Internet of Things (IIoT). While this integration opens new opportunities, it also introduces biases that undermine the reliability and robustness of scientific and industrial outcomes. This article presents a systematic literature review (SLR), supported by natural language processing techniques, aimed at identifying and classifying biases in AI-driven research within industrial contexts. Based on this meta-research approach, a taxonomy is proposed that maps biases across the stages of the scientific method as well as the operational layers of intelligent production systems. Statistical analysis confirms that biases are unevenly distributed, with a higher incidence in hypothesis formulation and results dissemination. The study also identifies emergent AI-related biases specific to industrial applications such as predictive maintenance, quality control, and digital twin management. Practical implications include stronger reliability in predictive analytics for manufacturers, improved accuracy in monitoring and rescue operations through transparent AI pipelines, and enhanced reproducibility for researchers across stages. Mitigation strategies are then discussed to safeguard research integrity and support trustworthy, bias-aware decision-making in Industry 4.0. |
|---|