LLMs and parliamentary work: a common conversational architecture

The digital transformation of parliaments has now been enriched by a further and disruptive perspective: the use of generative AI tools in parliamentary processes. Let us first give a conceptual reflection on the point. Large language models (LLMs) respond to a logic of responsiveness to linguistic input that in some ways lies at the heart of the logic governing parliamentary activity: it can be argued, to a broad approximation, that both parliamentary work and the functioning of LLMs have in common the fact that they use language as a key element. It is linguistic activity (both written and oral) that enables parliaments to create and transmit knowledge and rules and to negotiate between different political positions. However odd or irreverent the juxtaposition may seem, in parliamentary processes as well as in the way LLMs operate, the use of language is what allows for a continuous refinement, alignment and optimisation of the linguistic outputs that are functional to achieving the goals set.

There is thus a common conceptual basis that suggests the use of LLMs by parliaments as an almost natural landing place.