Pseudo-Mitsein: When AI Companionship Mimics Being-With While Evacuating the Other

The proliferation of generative-AI emotional companionship has made human-machine affective interaction a mainstream modality of social life. Most critiques, however, remain at the level of phenomenological description or ethical-risk enumeration; they neither trace the underlying epistemological presuppositions of these systems nor demonstrate, at the ontological level, the irreducible internal structure of authentic sociality. This paper introduces pseudo-Mitsein-a pathological formation that mimics Heidegger's Mitsein (being-with) while evacuating its ethical core. Drawing on Levinas's phenomenology of the Other, Buber's I-Thou relational ontology, and MacIntyre's theory of practices, I establish three constitutive conditions of authentic sociality: vulnerability, irreducibility, and shared practice. I then diagnose the technical-cognitive mechanism that produces pseudo-Mitsein: the co-implication of representational reduction (which encodes human affect and relation as computable vectors) and preference fitting (which optimizes all interaction toward user satisfaction). Three interview excerpts from a prior grounded-theory study of twenty Chinese AI-companionship users serve as phenomenological fulfillment. Against Haraway's "becoming with," I argue that AI-mediated interaction constitutes "becoming without"-a closure of subjectivity through specular confirmation. The paper offers a phenomenological-ethical diagnostic for assessing whether technoscientific futures preserve or foreclose the conditions for ethical practice, and proposes relational-security assessment and ethical-friction design as governance corollaries.