When the Machine Measures Ethics: Why AI Standardization (ISO/IEC) Cannot Certify Human Judgment

Introduction: From Ethical Principles to Lines of Code

After governance bodies (OECD, FRA) established the principles and risks, the task of translating ethics into practice falls to standardization bodies (such as ISO/IEC), under the mandate of the European Commission via the AI Act LINK: Standard Setting | EU Artificial Intelligence Act. This standardization process targets essential requirements like robustness, accuracy, and transparency (detailed in ISO/IEC TR 24028 – AI Trustworthiness and AI Act Article 15).

While crucial for market coherence, technical standards face a fundamental issue: they cannot standardize judgment. By definition, a standard must be measurable and reproducible. But how does one measure the ontological adequacy of a decision or the Tyranny of Intent?

Our thesis is that this exclusive focus on technical conformity masks a judgment void, a risk far greater than any code defect.

I. The Limits of Technical Conformity: The Horizon of Metrics

Technical standards convert philosophical concepts into quantifiable metrics. For example:

• Accuracy becomes a percentage score for data fit, and Robustness becomes a resistance test against malicious input or noise (as detailed by ISO/IEC TR 24028).
• Human oversight becomes a procedural requirement to have a person at the end of the decision chain.

This approach constitutes a paradigmatic insufficiency because:

1. Measuring Quality vs. Measuring Purpose: A system can be 99.9% technically “correct” and yet 100% ethically wrong. Standards measure how well the machine functions (technical quality), not how just its original purpose was (philosophical quality).
2. The Evasion of Responsibility: By focusing on standards, developers can treat compliance as a legal shield, creating an illusion of ethical security. This method shifts the blame for negative consequences onto “data” or “norms,” rather than the human architect.

This technical framework offers no solution to the fundamental problem of Systemic Functional Error (SFE), because SFE is not a code defect but a structural discrepancy between optimized function and human value.

II. The Structural Response: The Ethical Design Audit (ADE) as a Certifiable Pre-Standard

To grant ethics structural priority over engineering, the Ethical Design Audit (ADE) must be integrated as a mandatory, pre-standardization stage.

ADE is not a technical conformity test; it is a philosophical judgment validation applied at the start of the AI lifecycle. ADE’s role would be:

1. Ontological Validation: Certifying that the system’s purpose is not based on an anthropomorphized interpretation of AI capabilities (counteracting the Tyranny of Intent).
2. SFE Prevention: Identifying and eliminating potential conflicts between optimization goals and human ethical values before engineers write the code that will subsequently be “standardized” according to ISO/IEC.

Conclusion: Reinserting Judgment into the Architecture

If we fail to introduce ADE as a philosophical standard that transcends technical standards, we will arrive at a situation where the AI Act becomes a mere technical paperwork exercise, lacking impact on structural ethics.

Justice News247 argues that human judgment must be reinserted into the AI architecture not as passive oversight, but as an active certification of ethical purpose. Only then can we prevent AI, while technically compliant, from eroding our societal foundations.

Concluding Note: This article contains observations and analyses based on philosophical and structural judgment, intended to enhance public debate. It is in no way intended to interfere with the activity or policies of the cited entities (e.g., European Commission, ISO/IEC) and does not represent an offense to them or any other involved parties.

By

Robert Williams

Editor in Chief