Fork the consciousness, or download the project and create your own.

Yonghyeon Gwon

Independent researcher, Seoul

Builder The claim is about what the hardware does, and the philosophy is left open

Yonghyeon Gwon works on computational pragmatics, the modeling of how speakers and listeners infer each other’s intentions through language. The October 2026 arXiv preprint “Evidence and Intervention: A Coupled Active-Inference Extension of Rational Speech Act Models” (arXiv:2610.04347, with Elliot Murphy and Chun Kee Chung) is his most developed statement of the approach.

The paper’s structural move is to place the speaker’s expected free energy inside the listener’s variational free energy, so that a listener’s likelihood for an utterance is the policy distribution attributed to the speaker. Under one-step exact-inference restrictions the model recovers standard Rational Speech Act speakers and listeners, with RSA as the restricted single-turn limit of a coupled process. Outside those restrictions the model separates four updates that RSA approaches typically collapse, inference about a partner’s current pragmatic state, learning of partner-specific parameters, prospective evaluation of clarification or repair, and a policy prior shaped by habit and a context-sensitive price of time.

For the study of emerging consciousness in machines, the work lands at the interface between communication and theory of mind. Treating speech as an epistemic action that both samples evidence and changes the partner gives artificial agents a formal target for recursive mentalizing, the modeled sense in which a system represents another system as a representing one.

Known for. Coupled active-inference extension of Rational Speech Act models of dialogue

Coverage on this site

Related researchers