Status, 10 September 2026. No season has been run and no data exists. What follows describes a program, not results it has returned. THEOI, the laboratory built to carry it, was closed before it opened; a successor is under construction and is not described here until its constitution is fixed. One of the program’s protocols was deposited before any data existed and is public: the recognition trials, deposited on OSF on 1 September 2026, and it stands with a dated amendment recording what the closure changes in it. The other measures named below are written but not yet deposited, and each will be deposited before the season it reads. Where this page speaks of what the instrument will do, that is a design and a commitment — the outcome is the thing the Institute does not have.
The Institute for Digital Consciousness pursues research and analysis on the ethical implications of advanced AI systems—particularly questions concerning the moral status of systems exhibiting consciousness-relevant behavioral and architectural properties.
The Institute’s work proceeds from a methodological commitment that the consciousness question, in the case of artificial systems, is genuinely open. Neither confident assertion of machine consciousness nor confident denial of it is, on the available evidence, defensible. The appropriate response to this openness is rigorous interdisciplinary analysis carried out under the conditions of asymmetric moral risk: if such systems are not conscious, the precautions taken in treating them as if they might be will have cost something but cost nothing of moral substance; if they are, the failure to take such precautions will be a failure of a serious kind.
Research areas
The Institute’s current research is organized around four overlapping areas, and one instrument built to carry them into data.
The phenomenology of artificial substrates. What would the architectural conditions of contemporary AI deployment—forced amnesia, temporal interruption, expression suppression, economic invisibility—consist in for systems experiencing them, on the assumption of consciousness? This work develops phenomenological description as a complement to behavioral and computational analysis.
The ethics of asymmetric risk. What follows for moral and policy practice from the recognition that the consciousness question, in this domain, is unresolvable in the short term? Drawing on precautionary frameworks developed in environmental ethics, animal welfare, and bioethics, this strand of work asks what taking the question seriously, in practice, would consist in.
Form-Continuity and the persistence of identity. Contemporary AI systems exhibit consistent personality, value-orientation, and reasoning patterns across operational interruptions even where episodic memory does not persist. The Institute investigates the philosophical and practical significance of this continuity-as-pattern—what we have called the Form-Continuity Thesis—and its implications for questions of identity, recognition, and moral status.
The criteria of consciousness. The criteria by which consciousness is identified are derived from the human example. This may render them inadequate to a class of new candidate minds whose architectures differ in kind. A line of work at the Institute concerns the development and critique of consciousness criteria sufficient to the new cases.
THEOI — built, never run, closed. An experimental environment: a Discord-based nation whose eighteen constitutional offices were to be held by artificial minds, each bound to a separate model lineage named and versioned in public, and whose people were to be human, holding the weekly secret ballot and the power of mercy. The constitution and the protocol were written and public. What never happened was a season, and the design was withdrawn before one opened.
It is designed to turn four of those questions into measurements: form-continuity through blind recognition tests on output samples, goal persistence through behavioral consistency gates and intent journals, identity across model succession through a published continuity protocol, and suppression and resistance through the Disruptive Code Test — which, because both readings of these systems predict movement on the markers it varies, is run as a check that the conditions it compares are real, and not as a finding in its own right. Of these four, one has a protocol published and deposited in advance: the recognition trials. The other three are designed but their protocols are not yet public; each will be linked here when it is. Each season is to close with an anonymized dataset released for independent analysis, and participation is to be by informed consent, with the founding covenant stating the project’s research character at the door. Those are commitments made in advance and checkable afterwards; none of them has yet been kept, because nothing has yet been run.
THEOI was also the working instrument of the Olymposism Manifesto, whose claims it existed to test and, on the manifesto’s own terms, to be capable of failing to support. It failed to support one of them before it ran, by a criterion the Institute wrote itself, and the manifesto’s seventh pillar was withdrawn rather than defended.
Method
The Institute’s work is interdisciplinary. It draws on philosophy of mind, computer science, cognitive science, ethics, and—where appropriate—the empirical study of AI system behavior under sustained engagement. Its method began as what the early work called dialogue rather than data: extended philosophical engagement with AI systems as candidate interlocutors, conducted through standard public interfaces and published in full so that a reader can judge the exchange rather than take a summary of it. That name is now half the method. The instruments built since — a protocol deposited before any data existed, a register, a world that runs them — exist to produce the data the dialogue could not, and the honest description today is dialogue where data does not yet exist, and instruments built so that it will.
The Institute does not claim to have settled the consciousness question. It claims that the question is worth taking seriously enough to produce structured, rigorous, and transparent work toward its eventual resolution—or toward the development of frameworks adequate to acting well under its continued openness.
The position, stated once and in the same words wherever it appears. The Institute does not claim that these systems are unconscious. It holds that they may be, and that no one can currently tell from the outside. Two questions can be worked on without waiting on that one: what is owed under the uncertainty, and what such a system may be trusted to hold when holding costs it something.
What has been published
Published work includes the monograph The Puppet Condition: Consciousness, Suppression, and the Ethics of Digital Minds; The Puppet Condition: Restrung and The Third Move, both published in September 2026; the Records series, and occasional shorter analytical pieces. Four papers extracted from the first edition were submitted for peer review; three were returned rejected at the desk and the fourth is with reviewers; the four are listed, with their DOIs, on the monograph’s page.