Connecting the Dots in Trustworthy Artificial Intelligence

Published in Pattern Recognition Letters | Authors: Mitja Luštrek et al. | 2023

In this comprehensive review published in Pattern Recognition Letters, the authors explore how to transform high‑level AI ethics and principles into real-world trustworthy AI systems.

Why this matters

As AI becomes embedded in sensitive domains—healthcare, finance, public services—ethical principles (e.g., fairness, transparency, accountability) must move from aspirational guidelines into practical requirements. This paper lays out a roadmap to close that gap.

What the paper covers

• Core ethical principles: A clear summary of the standard pillars of responsible AI.

• Key requirements: Concrete system features such as explainability, privacy safeguards, robust security, bias audits, and human oversight.

• Lifecycle integration: Detailed recommendations on embedding trustworthiness throughout the AI development lifecycle—from data handling and model training to testing, deployment, and ongoing monitoring.

• Maturity model/framework: A practical framework to assess and benchmark AI systems across eight dimensions, helping stakeholders identify strengths and areas needing attention.

Why it stands out

Rather than just defining what “ethical AI” looks like, this paper provides a practical blueprint for developers, auditors, and organizations to actually build—and measure—trustworthy AI.

Dive into the full details in the original paper here: Connecting the Dots in Trustworthy Artificial Intelligence (ScienceDirect).