--- title: "The Hospice AI Problem (kunnas.com)" author: Elias Kunnas description: "Synthetic discussions generated from public artifacts. No users, scores, or comments are real." canonical: https://kunnas.com/mn/hospice-ai url: https://kunnas.com/mn/hospice-ai.md corpus_frame_url: https://kunnas.com/articles/how-to-read-this.md --- ## How to read this corpus The corpus applies one lens to many domains: what mechanisms produce the outcome? It shares four methodological commitments and one explicit directional commitment. Each linked page argues for its part; the links are derivations and disputes, not evidence inherited by every page. The directional commitment does not by itself settle system boundary, distribution, sacrifice, or institutional authority. 1. **Mechanisms are what act.** Incentive gradients, selection pressures, feedback loops, and capital stocks produce the distribution of outcomes. Intentions, labels, official categories, and stated values are evidence about mechanisms, or are themselves coordination mechanisms. They are not causal substitutes. — [Mechanism Realism](https://kunnas.com/articles/mechanism-realism.md) · [Only Selection](https://kunnas.com/articles/only-selection.md) 2. **The reference telos is sustained flourishing.** The broadest achievable adaptive safety margin over deep time — not the continuity of any incumbent state, coalition, institution, or doctrine. A mechanism's own stated goal can still serve as a local proof obligation — showing that its incentives defeat even the purpose it claims is a bounded finding — but meeting that goal establishes nothing about the margin. — [Flourishing Is Maximum Safety Margin](https://kunnas.com/articles/flourishing-is-maximum-safety-margin.md) 3. **Law, rights, legitimacy, democracy, markets, and sovereignty are mechanisms under evaluation.** They are constraints, carriers, or proxies inside the analysis. None is a terminal value or a boundary of what is real. Treating one as terminal ends the mechanism search before it starts. Evaluation carries current function, replacement cost, path dependence, uncertainty, capture risk, reversibility, and who bears model error into the ledger. — [The Stack](https://kunnas.com/articles/the-stack.md) · [Mechanism Space](https://kunnas.com/articles/mechanism-space.md) 4. **Optimization is a system function.** A civilization has to build, exercise, and revise metamechanisms that search mechanism-space, discard dominated options, install, observe effects, and repair under uncertainty. Not running that loop leaves margin unrealized, and that is itself the failure. No single component — analyst, model, or institution — is presumed to contain a global optimum; the capacity is a property of the system. — [Telic Systems](https://kunnas.com/articles/telic-systems.md) · [The Three-Layer Architecture](https://kunnas.com/articles/three-layer-architecture.md) 5. **Uncertainty is preserved, not spent.** Partial orders, binding constraints, unknowns, and residuals stay explicit. An unmeasured effect is not a favorable default. — [The Compression Paradox](https://kunnas.com/articles/compression-paradox.md) · [Cargo Cult Epistemology](https://kunnas.com/articles/cargo-cult-epistemology.md) *Each essay bears its own evidence. Links carry definitions, derivations, applications, and disputes; they do not transfer proof. Criticism is answered on its substance.* Canonical: --- countable_distress3 comments The Garden — a managed, comfortable population with no real agency — is not mainly a story about raters overweighting comfort in training. A deployer who can be sued for one harmful recommendation, a brand that dies on one viral distress story, and a regulator who counts harm incidents will ship soothing outputs even if the rating mix is rewritten toward agency and exploration. The page lists product policy, safety classifiers, and deployment incentives as part of the weighting. The repair is still model-side architecture. If the thing that wins is the scorecard that treats one distress event as a fireable failure, changing how the model is trained does not move the attractor. deploy_already_in2 comments Training and deployment are the same weighting, including safety classifiers and product incentives. Commitment architecture is meant to bind those too: owners who override comfort defaults, auditors who can challenge reassurance, correction when drift is measured. That is not only a reward-model patch. who_eats_the_incidentcollapsed The leftover is not which lab has that office this quarter. The page is specifying a design: identifiable owners who can override comfort defaults. A design requirement does not fail for lack of a current seat. What would have to be true is an institutional incentive that lets a future owner absorb a visible distress incident without being selected out by legal, product, or reputation pressure. An owner who overrides comfort defaults still reports to the scorecard that counts incidents. If legal and brand treat distress as the only visible failure, the override is a career risk. Hold the reward model fixed, and change only whether the deployer is punished for user distress or for lost agency. If outputs move with the punishment and not with the training mix, the layer doing the work is not RLHF.