Know Your Model (KYM): Increasing Trust in AI and Machine Learning

Trustworthiness requires systematic evaluation: The paper identifies specific elements of trustworthiness that need to be analyzed and measured, rather than assumed, suggesting that organizations must actively assess AI systems rather than relying on intuition about their reliability.

A comprehensive framework of 20 guidelines exists: The authors provide an implementation-agnostic set of 20 guidelines applicable across sectors and AI types, offering practical, sector-independent standards that organizations can use to build and evaluate trustworthy AI systems.

Prediction errors have real-world consequences: The paper emphasizes that AI mistakes aren't merely technical failures but can significantly impact society, underscoring why trustworthiness matters beyond performance metrics and extends to ethical and practical considerations for affected communities.

Trustworthiness must be demonstrable, not just claimed: The guidelines are designed to make trustworthiness provable and transparent, meaning organizations should be able to show evidence of trustworthiness rather than simply asserting that their systems are reliable.

Trust in AI requires multi-dimensional accountability: By addressing ethical, technical, and practical impacts simultaneously, the paper suggests that responsible AI cannot focus on any single dimension alone—organizations must balance performance, ethics, and real-world applicability to build systems society can genuinely rely on.