How Do I Know What to Say Next? Barenholtz's Autogenerative Theory as an Enrichment of Harrisean Integrationism
arXiv:2607.07891v1 Announce Type: new Abstract: Roy Harris's Integrationist linguistics offers a compelling critique of the referentialist tradition embedded deep at the heart of computational approaches to language, arguing that language is not a code that maps onto a pre-given world but a situated, bipartite activity oriented toward prospective joint action. Yet Integrationism leaves certain explanatory gaps: it does not fully account for the structural mechanism by which signs sustain prospective openness, it undertheorises the continuity between linguistic and non-linguistic semiotic activ
This paper's 2026 publication reflects ongoing academic efforts to refine theoretical underpinnings of AI language models, pushing beyond current computational linguistics paradigms.
A deeper theoretical understanding of language generation, particularly one that moves beyond referentialism, could lead to more robust and contextually aware AI agents, impacting their usability and reliability in complex environments.
The theoretical enrichment of Integrationism provides a new conceptual framework for approaching AI language understanding, potentially shifting future research directions away from purely statistical or 'code-like' interpretations of language.
- · AI researchers focusing on human-like communication
- · Developers of contextual AI assistants
- · Linguistics departments
- · Purely statistical language models
- · AI approaches lacking situated understanding
Refined theoretical frameworks for AI language models emerge, leading to new model architectures.
AI systems develop a more nuanced understanding of human communication and intent beyond simple keyword matching.
This could contribute to the development of AI agents capable of more sophisticated social interaction and complex task execution.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL