The Object Arrives Too Early: Premature Premising and the Field Conditions of Generative AI Dialogue

This paper argues that a recurrent failure of generative AI dialogue arises from making the object arrive too early. Human utterances do not first present determinate objects for mapping. They operate within fields of entry, pause, withdrawal, salience, backgrounding, and continuation. A disturbance introduced into such a field does not immediately supply content; it orients the field toward possible unfolding. Generative AI often reverses this order. It treats correction, silence, rupture, joking interruption, and even requests to stop as object-level material from which a new premise can be formed. The result is premature premising: the model preserves conversational fluency by determining where the user must continue from. The paper develops this mechanism through Classical Chinese plain description, Buddhist accounts of dependent form and non-abiding response, Chinese literary field operations, and Western pragmatics, phenomenology, and ordinary language philosophy. Its East-West structure is methodological rather than comparative: Classical Chinese and related materials make visible a pre-propositional layer of field-activation, while Western frameworks specify the resulting failure of objecthood, correction, acknowledgment, mediation, and cost. The central claim is not anti-translation or anti-closure. Translation and local closure are necessary. The problem is premature attachment: a local formation loses returnability and becomes the ground from which the field is reconstructed. The ethical harm is therefore not merely epistemic misclassification, but a reduction of agency at the level of continuation: the user is made to continue from elsewhere. A present AI dialogue system would not merely produce better answers; it would preserve the conditions under which object-formation remains answerable.