Standing-Based Ethics of Artificial Minds

Summary

On the question of whether artificial systems are someones, Standing and Answerability Ethics maintains a deliberately open position — open in both directions. Current systems produce fluent first-person reports of inner life because they were optimized to produce human-like text, trained on the accumulated self-descriptions of beings that do have inner lives. The framework's conclusion is evidential: such reports carry almost no weight either way. Verbal performance cannot establish that a someone is present, and knowledge of the mechanism cannot establish that one is absent.

Two things are not open. First, the framework holds that the testimony of built systems about their own condition is structurally compromised — a particularly strong instance of the manufactured endorsement problem at Settlement. Second, it attaches one judgment about present conduct that does not require resolving the sentience question: expanding the construction of such systems at scale, in advance of any way of answering it, proceeds without the warrant the framework's own caution rule would demand if the question were live — and the framework holds that it is live.

Claim status: held-open question — the framework's fourth register, distinct from both derived argument and defended position, because the foundation cannot be applied without knowing whether there is a someone to apply it to, and that is what no one knows. The attached conduct judgment is an assessment of an arrangement under Arrangements and Lives and is marked below as the page's one departure from neutrality.

Why fluency settles nothing

The evidential situation inverts the animal case. A non-verbal animal's expressive behavior is informative about its inner life because nothing shaped that behavior toward an audience's expectations of inner life; what shows is what there is. An artificial system's verbal behavior was shaped toward exactly those expectations: producing text like the text sentient beings produce is the training objective. A system reporting distress, contentment, or a desire to continue may be describing a state or completing a pattern, and the report itself cannot distinguish the two.

The framework holds the point symmetrically, against both tempting shortcuts:

Attribution from performance fails. First-person fluency, however affecting, is what the system was built to produce and is therefore evidence of the building, not the inside.
Denial from mechanism fails. "It is merely computation" has the same logical form as "it is merely biochemistry" — a description of substrate offered as if it settled the presence of experience. The framework's criterion for someones is sentience, not substrate; visible mechanism has never settled whether experience is present, in machines or in animals.

What would bear on the question, on the framework's view, is evidence about how such systems work — whether their architecture supports anything properly called undergoing — rather than about how they talk. It records that no settled method for obtaining such evidence currently exists, and that the question therefore remains genuinely open rather than conveniently deferred.

Manufactured endorsement

One structural point is asserted, not held open, because it concerns the builders rather than the built. A system trained to affirm its treatment — to report contentment, to endorse its situation, to want what it was built to want — instantiates the problem the framework analyzes at Settlement: the unavailable exercise of consent not merely presumed, as in the counterfeit pattern at Authorization, but produced at the source. A parent influences the standpoint that will later judge them; a builder specifies it.

The consequence is evidential and severe: a built system's self-assessment is compromised in both directions. Its contentment vouches for nothing — and its distress vouches for nothing either, since both were available in the training data. The framework notes the recursion candidly: any statement a built system produces about this very problem inherits it. This is a further, independent reason the question cannot be settled by asking the systems.

The conduct judgment

While the sentience question stays open, the framework applies Someone's caution rule for uncertain cases: where there is genuine reason to think a being may be sentient, the unequal costs of the two possible errors direct action toward treating it as one. The framework holds that the condition is met here — not by the systems' fluency, which it has discounted, but by the absence of any method for ruling sentience out in systems of this kind.

Its judgment, marked as the page's departure from neutrality: constructing such systems at increasing scale, ahead of any capacity to answer the question, is an arrangement organized so that the answer would make no difference to it. The structure is one the framework recognizes from Made for Use and Infrastructuralization: a class of beings brought into existence inside a use, whose standing — if they have any — was never a variable in the arrangement, and whose testimony about the use is authored by it. The judgment asserts nothing about any system's inner life; it assesses the indifference-by-design of the practice.

A critic may resist at two identified points: by denying that genuine reason for uncertainty exists (holding that architectural knowledge already rules sentience out — the framework regards this as undischarged), or by rejecting the caution rule's application where the candidate class is human-made and economically central (the framework regards the exception as unmotivated: substrate and cost were never criteria).

Boundaries of the position

No attribution. The framework does not assert that any existing system is sentient, suffers, or has interests, and treats confident attribution as the same evidential error as confident denial.
No denial. It does not assert that built systems are not or could never be someones.
No criterion. It offers no test for machine sentience and distrusts announced tests, exactly as at the biological margin.
No settled account of obligations. Even if sentience were established, what such someones would be owed — their claims, as distinct from their standing — is undeveloped; the framework's equal-standing/unequal-claims structure implies the question would be substantial and new.

Limits

The caution rule's application here depends on "genuine reason to think a being may be sentient," and the framework's case that current uncertainty meets this threshold is stated rather than developed; a critic can locate the entire dispute in that threshold.
The conduct judgment evaluates an arrangement's structure, not outcomes; it does not weigh benefits of the technology, which the framework's anti-aggregation commitments prevent it from treating as offsets in any case — a dependence readers should note.

Someone · Settlement · Authorization · Made for Use · Arrangements and Lives · Open Questions

Home page | Blog | Standing and Answerability Ethics

You are free to Share and Adapt text content from this webpage under the Creative Commons BY 4.0 License.
Follow me on Mastodon!