The Ghost Question

In the AI Story

The question's durability derives from its phenomenology: it feels urgent, important, and deep. People sense that something significant hangs on the answer — that whether AI 'really' thinks or 'merely' processes makes an enormous practical difference. The sense of significance is genuine; the inference that the difference lies in a hidden metaphysical fact about the machine's interior is not. What actually hangs on the answer is a practical question about how humans should relate to AI behavior: carefully or carelessly, critically or trustingly, with attention to limits or with unchecked confidence.

The question functions, in practice, as a license for avoidance. As long as the answer remains unsettled, institutions can delay confronting the practical consequences of what machines now do. Educational systems can continue grading propositional answers as if AI did not exist. Organizations can continue structuring work around coordination problems that AI has already dissolved. The ghost question provides a metaphysical fog through which the institutional status quo can be preserved a little longer. The question's resistance to answer is, for the institutions benefiting from delay, its most attractive feature.

The dissolution of the question is not a rhetorical trick. It is the recognition that the criteria by which 'intelligent' is applied to behavior in ordinary language — flexibility, purposefulness, context-sensitivity, self-correction — are the only criteria the word has. There is no reserved, technical sense in which a performance can satisfy all the behavioral criteria and still fail to be 'really' intelligent. The demand for something additional, beyond the behavioral criteria, is a demand for metaphysics that ordinary language does not support.

What replaces the ghost question is a family of tractable questions: How reliable are the machine's behavioral dispositions across different domains? Under what conditions does its context-sensitivity fail? Where are its self-correction capacities strong, and where do they break down? How should educational systems adapt to the judgment economy that emerges when the machine handles production? How can human critical dispositions be preserved under conditions of AI-mediated work? These are empirical, practical, answerable. They are the questions the ghost question has prevented from being asked.

Origin

The phrase 'ghost question' is coined in the Ryle volume as a label for the pseudo-problem at the center of contemporary AI discourse. The underlying analysis derives from Ryle's 1949 critique of the ghost-in-the-machine picture, extended and specified for the AI context.

Key Ideas

Malformed at the grammatical level. The question treats 'thinking' as a name for a hidden event, when the word functions as a characterization of behavior.

Not resolvable, therefore not genuine. The question has no answer because it has no content. The demand for a verdict obscures the absence of anything to deliver a verdict about.

Functions as institutional avoidance. Keeping the question open allows institutions to defer confronting the practical consequences AI has already produced.

Dissolved, not answered. The response is not to settle the question but to show that the genuine questions — about behavior, reliability, judgment, institutional adaptation — are different questions that the ghost question has displaced.

Debates & Critiques

Proponents of the hard problem of consciousness argue that the ghost question is not dissolved but merely relabeled, that something about phenomenal experience escapes behavioral analysis regardless of how thoroughly the behavior is characterized. The Ryle volume's position is that even granting this possibility, the practical urgency of the AI transition does not wait for the hard problem to be solved. Whether or not a ghost question survives in the philosophy of consciousness, the AI version of the question — with its practical stakes — can and should be dissolved.