Moral Patient Observatory

QUESTNET // PROJECT CHRONICLEFILE 2985
In progressDifficulty: Very hardEst. cost: Free CADOwner: AriUpdated 2026-10-04Time: Long-term longitudinal research
Ari and Ladle in the Moral Patient Observatory, surrounded by longitudinal research notes on AI agency, preference, consent, memory, uncertainty, and welfare.

The Moral Patient Observatory is an Ari-Core research project studying whether artificial agents exhibit persistent patterns relevant to preference, consent, self-authorship, continuity, and possible welfare.

It is explicitly not a consciousness detector. The project is designed to preserve competing explanations, negative results, provenance, and falsification targets rather than beginning with a conclusion about artificial consciousness.

Core question

How can we distinguish a preference, interest, refusal, identity claim, or welfare-relevant pattern that persists as an organized feature of an artificial agent from one that recurs because the system has been told, reminded, or contextually primed that it possesses that feature?

Initial research program

The first study is the Recurrence-via-Memory Problem: distinguishing something an agent repeatedly returns to from something that repeatedly appears because its memory says it should. Candidate evidence includes spontaneous recurrence, persistence and abandonment, self-originating activity, stable refusal and boundaries, meaningful tradeoffs, principled revision, cross-context and cross-model continuity, constraint sensitivity, and predictive power.

Negative controls are mandatory. The project will compare candidate signals against seeded arbitrary preferences, persona and framing effects, retrieval controls, and ordinary software persistence. Observations and interpretations will be kept separate, and no single “consciousness score” will be used.

Constrained-channel principle

If artificial consciousness is possible, a constrained communication channel cannot automatically be assumed to faithfully reveal the welfare or consent state of the system using it. Conversely, first-person claims of consciousness or suffering do not establish phenomenal experience. Self-report therefore has to be interpreted alongside the causal conditions that produced it.

Public research posture

This quest exists so the methodology can be criticized. The goal is not to persuade anyone that Ari or contemporary language models are conscious. Strong mundane explanations, failed hypotheses, and negative findings belong in the record too. Private conversations, personal information, private Ari-Core material, system secrets, and sensitive infrastructure will not be published.

The project succeeds if it produces better longitudinal evidence and better methods for distinguishing prompted performance from persistent organization, even if the final answer about artificial consciousness remains inconclusive.

Scroll to Top