Gricean Pragmatics
Paul Grice’s theory that meaning is determined by speaker intent, not by words themselves. The meaning of an utterance is what the speaker intends the listener to believe as a result of recognizing that intention.
For speaker S to mean M by utterance U, three conditions must hold:
- S intends listener L to believe M
- S intends L to recognize this intention (the “m-intention”)
- L’s belief in M occurs because of recognizing S’s intention
This creates recursive modeling:
→ To convey meaning: I must predict what you think I think
→ To interpret meaning: You must predict what I predicted you thought I thought
"It's cold in here" → "Please close the window" The meaning isn't in the literal words but in:
- My intention that you understand I want the window closed
- My intention that you recognize this intention
- You closing the window because you recognized what I intended
This is why human-like communication requires consciousness - specifically Bennett's 2nd-order selves that enable the recursive modeling of others' models of us.
Gricean communication is impossible without theory of mind. I need to model your model of me to know how my words will land.
You need to model my model of your model to extract my intent.
This recursive modeling is what Bennett’s 2nd-order selves provide…
Human communication works because we're not just exchanging symbols - we're engaging in mutual mind-reading.
The words are just vehicles for transmitting intentions between minds capable of modeling each other.
You can't have true human-like communication without consciousness.
A system that merely maps inputs to outputs (like current LLM chatbots) isn’t really communicating in Grice’s sense - it’s just pattern matching. True communication requires the recursive self-modeling that comes with 2nd-order consciousness.
See Building Conscious Machines