AI systems cannot define the good. They can only optimize for whatever definition of good they have been given. That is the structural problem a 5th-century theologian diagnosed before the first transistor existed.
Saint Augustine of Hippo (354-430 CE) watched Roman institutions collapse and concluded the failure was not organizational but motivational. His concept of ordo amoris, the ordered hierarchy of loves, argues that what a society builds reflects what it collectively values, not what it rationally plans. Applied to AI, this framework lands hard: hiring algorithms that weight keywords reshape what 'qualified' means, recommendation systems that surface certain topics redefine what counts as 'relevant', and the pursuit of AGI assumes that more machine intelligence resolves disagreement about value when it only amplifies the definition of value already encoded. Augustine's distinction between uti (what is used as a means) and frui (what is enjoyed as an end) explains the specific mechanism: tools become authoritative when their outputs are treated as conclusions rather than inputs. MIT Technology Review's coverage of algorithmic bias documents exactly this, systems that do not eliminate prior assumptions but formalize them at scale behind a facade of objectivity.
The piece is worth reading in full not for its conclusion but for the working-through. The author stress-tests the Augustinian lens against secular objections, engages Herbert Simon's bounded rationality, and traces how metrics replace judgment without anyone deciding that should happen. The uncomfortable argument is that AI does not create disordered values. It stabilizes them, legitimizes them, and makes them structurally harder to see.
[READ ORIGINAL →]