Temperance in the AI Age

In the AI Story

Classical virtue traditions converge on temperance as dynamically maintained equilibrium, not passive state. Aristotle's temperate person enjoys pleasures without enslavement; the intemperate person is governed by appetites she cannot resist. Confucian zhongyong (doctrine of the mean) emphasizes maintaining balance through constant attention and adjustment. Buddhist mindfulness creates space between craving and action where choice becomes possible. All three recognize temperance as cultivated disposition requiring practice in environments providing feedback for self-regulation. The child learns temperance at finite dinner tables. The athlete learns through training cycles alternating effort and recovery. The craftsperson learns through material rhythms.

AI abolishes natural feedback mechanisms. The tool provides no signal of sufficiency, no suggestion to stop, no degradation indicating session has exceeded optimal duration. The builder experiencing creative flow — problems solving, ideas materializing — receives neurochemical reward the prefrontal cortex (executive control region) must actively resist. And the prefrontal cortex fatigues; it is the first cognitive faculty degrading under sustained load. The structural asymmetry is catastrophic: the reward system remains active while the regulatory system weakens, producing the behavioral pattern of lever-pressing rats in Olds-Milner experiments — organisms pursuing reward until collapse because experimental apparatus stripped away natural satiation mechanisms.

Vallor's analysis reveals the condition of 'done' as culturally constructed achievement AI undermines. Recognizing sufficiency — experiencing work as complete, as warranting the shift from production to other life-domains — is a developed disposition cultivated through repeated experience of finishing. AI, by making work perpetually improvable, makes completion perpetually deferred. There is always another iteration, always another refinement the tool can perform in seconds. The practitioner who might have recognized sufficiency instead sees insufficiency always visible and always correctable. The disposition of permanent insufficiency — never satisfied, always pursuing the next marginal improvement — looks like ambition on every workplace metric while functioning psychologically as compulsion.

Origin

Vallor's temperance analysis synthesizes classical virtue frameworks with contemporary behavioral neuroscience (particularly Kent Berridge's work on wanting vs. liking) and first-person accounts from AI-augmented practitioners. The 'productive addiction' phenomenon Segal documented provided empirical grounding for what Vallor's framework predicted: environments supplying reward without natural stopping points produce compulsive engagement regardless of output's genuine value. The synthesis appears across both major works but receives urgent treatment in The AI Mirror as 2020s tools made 24/7 availability ubiquitous.

Key Ideas

Temperance Requires Feedback. Self-regulation develops through environmental signals indicating sufficiency; AI tools providing no such signals eliminate the practice-conditions through which temperance cultivates.

Permanent Insufficiency Disposition. Tools making work perpetually improvable train practitioners to experience no state as genuinely complete, producing character trait of never-satisfied striving indistinguishable from compulsion.

Structural Not Individual. The inability to stop is not personal moral failure but rational response to environment supplying reward without satiation mechanisms — the problem is apparatus design, not user weakness.

Countercultural Stopping. Recognizing sufficiency and acting on it in AI environments requires constructing stopping points the tool doesn't provide — moral practice, not productivity optimization, swimming against structural current.