Conceptproposed term

The agent as overhearer

Definition

Every new AI session can read the whole project but was never in the conversations that shaped it — so, like someone who overhears half a conversation, it is confident it understands more than it does.

By Chris Telford, with help from Claude

1Where it came from

it sees the repository (referential context) but not the history of negotiated meanings that produced it. Section 1

I found no one who has made this connection explicitly. Section 1

2The pieces that already exist

The overhearer's illusionoverhearers systematically overestimate their understanding of a conversation because they have access to referential context but lack the interactive grounding process. Section 1

Models presume shared groundthree times less likely to initiate clarification and sixteen times less likely to provide follow-up requests than humans Key finding 1

Asymmetric persistencethe repeated re-establishment of a working ground under asymmetric persistence. Key finding 2

3What to do about it

Encode intuitions as constraints (tests, types, schemas, acceptance criteria) before encoding them as prose; keep context files minimal. Recommendations

4How sure

the overhearer mapping and the RPD-delegation argument are original syntheses, not findings. Recommendations