Augustine’s Insight on AI: The Value Hierarchy Problem and Misordered Optimization
Audit AI systems to ensure outputs are treated as inputs, not conclusions, preserving human judgment.
Replace AI decision points with human‑review checkpoints that explicitly document the values guiding the AI output.
Summary
The article argues that AI systems, like all optimization tools, encode a prior definition of what is "good"—a value hierarchy that is not neutral. Augustine’s notion of “ordo amoris” explains how societies prioritize values, and when AI is tasked with optimizing a single metric, it reinforces the chosen value while ignoring others. This leads to misordering: hiring models that prioritize specific keywords narrow the definition of a qualified candidate, recommendation systems that highlight certain topics shape user relevance, and risk models that treat efficiency as the sole good sacrifice safety. The article notes that AI does not eliminate bias; it formalizes existing assumptions, making them appear objective. The ethical imperative is to preserve the distinction between AI outputs as inputs to judgment, not as conclusions. The article also discusses how the pursuit of AGI amplifies these value hierarchies rather than resolving them.
The core message is that AI systems cannot self‑justify the values they pursue; they must be guided by human judgment and transparent value definitions. For agencies building AI‑enhanced WordPress or e‑commerce features, this means embedding value‑clarity checks into the design process.
The discussion is relevant for internal strategy and for clients who rely on AI for hiring, recommendation, or risk assessment.
Key changes
- AI optimization requires a prior definition of "good" and encodes a value hierarchy
- Systems formalize existing biases, making them appear objective
- AI can reinforce misordered values, narrowing perceptions in hiring and recommendations
- Ethical task: preserve distinction between tool outputs and human judgment