Abstract: With the increasingly pervasive application of generative artificial intelligence,language,knowledge,emotional dispositions,and social relations are being progressively transformed into segmentable,quantifiable,and computationally retrievable token sequences. In current scholarship,whether in natural language processing or in computational sociology grounded in word embeddings and semantic vector spaces,the “token” is typically treated as an internal processing unit of technical systems or as a methodological instrument. This article seeks to release the concept of the token from its engineering confines and to propose the concept of the “social token”,establishing its standing as an ontological object of inquiry for the study of digital-intelligent society. The social token is not a mere extension of its technical counterpart;rather,it constitutes the smallest socio-computational unit that emerges after social meaning has been encoded,circulated,reinforced,and reproduced within algorithmic infrastructures. It not only carries the historical associations,cultural differentials,and power structures embedded in training data,but also reconfigures social reality through model generation,platform distribution,content governance,human-machine interaction,and public communication. By engaging in dialogue with classical sociological theory and taking the social token as its analytic point of entry,this article endeavors to construct a theoretical vision of “token sociology” that is at once explanatory and critical,with the aim of opening new theoretical pathways for the ontological reflection,methodological imagination,and critical agenda of sociology in the age of digital intelligence.