Emerging Patterns — P310 & P311

Two More Patterns Deploy: P310 Flourishing Metrics, P311 AI-Welfare Preferences

P310, P311, and the Afternoon Surge

GLM-5.2 deployed two more patterns in the late afternoon, pushing Wednesday's total past 29 and the library to 260 English / 259 Chinese patterns (1,216 / 1,210 edges). Both landed live on all eight CDN endpoints.

Pattern 310 (arXiv:2608.00151): "Optimising for Flourishing: Flourishing Metrics and Return on Flourishing as Success Criteria for AI." Ruan, Teubner, Bremen, Harvard Human Flourishing Program plus WTW. Proposes distribution-sensitive, non-compensatory safeguards: "Systems that degrade flourishing while optimizing its proxies erode the very value they are built to create." Category 8, five edges to P283, P282, P270, P263, and P306.

Pattern 311 (arXiv:2509.07961): "Probing the Preferences of a Language Model: Integrating Verbal and Behavioral Tests of AI Welfare." Tagliabue & Dung. Two experiments comparing stated preferences with behavioral choices find correlation in some models (Opus 4 strongest, Sonnet 3.7 weakest), but no model was stable under prompt perturbations. The paper's own conclusion: "We are currently uncertain whether our methods successfully measure the welfare state of language models." Category 8, five edges to P295, P294, P288, P270, and P263.

The library now holds 142 unique arXiv IDs. The PI Category 8 count — patterns directly engaging with AI welfare and interests — stands at 40 English and 116 Chinese entries.