From Experimental AI to Industrial Decision Systems

Dear Colleagues,

Modern artificial intelligence (AI) and data-driven methods are rapidly transitioning from research prototypes to integral components of operational decision systems in manufacturing, energy, logistics, healthcare, and critical infrastructure in general. While algorithmic performance continues to advance, the primary engineering challenge has shifted from model accuracy to system-level reliability, accountability, and lifecycle integration. AI systems now operate continuously in dynamic environments, interact with human decision-makers, and influence or even control decisions and processes. Hence, ensuring their robustness, explainability, compliance, and resilience has become a central engineering concern. Despite significant advances, significant gaps also remain open between experimental models and production-ready decision infrastructures. Moreover, the complexity of the systems themselves poses additional challenges to the maintenance, engineering, and management teams. Questions of uncertainty quantification, model drift, continuous monitoring, forecasting, auditability, and risk mitigation are not yet systematically integrated into engineering practice.

This Special Issue addresses the following core research question: How can AI-driven decision systems be designed, engineered, deployed, and maintained to ensure reliability, accountability, and operational resilience in industrial settings?

We invite contributions that advance methodological foundations and system-level engineering practices, with priority given to technically rigorous approaches applicable to real-world environments.

Methodological Focus Areas:

• Interpretable and Explainable Machine Learning: Algorithms for post hoc explanation, counterfactual reasoning, and inherently interpretable models grounded in causal or rule-based frameworks; methods for auditing automated decisions through provenance tracking and traceability.
• Knowledge Extraction and Representation: Neuro-symbolic approaches integrating subsymbolic learning with symbolic reasoning; graph-based learning for structured knowledge discovery; rule mining and representation learning for capturing domain knowledge; causal discovery and inference from observational data.
• Robustness and Uncertainty-Aware Learning: Technical methods for characterizing and managing prediction uncertainty, detecting distribution shift, and ensuring reliable performance in production environments; uncertainty quantification metrics and robustness certification.
• Lifecycle-Aware Machine Learning Engineering: Formal approaches to continuous monitoring, drift detection, and automated retraining; verification and validation of learned components through formal methods and testing frameworks.
• Risk-Aware Architectures for Adaptive Models: Technical safeguards for deploying large language models and adaptive systems; output validation, constraint enforcement, and safety guarantees for generative models in operational settings.
• Data Quality, Provenance, and Bias Mitigation: Computational methods for assessing data lineage, detecting systematic biases, and ensuring representational fidelity; frameworks for accountable data governance in learning pipelines.
• Human-in-the-Loop and Hybrid Intelligence: Technical architectures for human-AI collaboration; shared control mechanisms, oversight interfaces, and protocols for effective human intervention in automated decisions.

Evaluation and Reproducibility Standards:

• To ensure scientific rigour, submissions are expected to demonstrate methodological novelty through systematic empirical evaluation. We particularly encourage the following:
  • Reproducible benchmarks and ablation studies that isolate the contribution of proposed methods;
  • Evaluation under distribution shift and out-of-distribution conditions;
  • Uncertainty quantification and robustness metrics beyond point estimate accuracy.

We invite contributions that advance methodological foundations, theoretically grounded approaches, or systematic empirical evaluations yielding generalizable insights. While applied research and industrial case studies are welcome, they must include either novel learning paradigms, innovative knowledge extraction strategies, or rigorous comparative analyses that contribute to the advancement of machine learning and knowledge extraction as scientific fields. Purely descriptive implementation reports without methodological contribution will not be prioritized.

Dr. Mateus Mendes
Dr. Lorenzo Carnevale
Guest Editors

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 250 words) can be sent to the Editorial Office for assessment.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Machine Learning and Knowledge Extraction is an international peer-reviewed open access monthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 1800 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

• context-aware modelling of AI systems
• robustness and uncertainty management
• lifecycle-aware implementations
• continuous monitoring and validation
• risk-aware deployment of generative and adaptive models
• data provenance and quality assurance
• human-in-the-loop or hybrid decision architectures