Conversational patterns for human-machine interaction

Human conversation has conversational patterns: Schegloff (Sacks, Schegloff & Jefferson, 1974). But those are for human-human conversation.

If we do want to use conversation as an interaction model (which I think should still be possible, just not unrestrained), we should not reuse human-human conversational patterns, but design new ones for this particular context: human-machine interaction. And that might (should!) include patterns that include more constrained inputs and different interactions than what we see in human-human conversation. That’s the entire point: conflating human-human conversation with human-machine interaction harms both.

Research question: has the premise that conversation leads to better results in chatbots ever been researched?

Sources:

Sacks, H., Schegloff, E.A. & Jefferson, G. (1974). A simplest systematics for the organisation of turn-taking for conversation. Language, 50(4), 696-735.
Schegloff, E.A. (2007). Sequence Organization in Interaction: A Primer in Conversation Analysis. Cambridge University Press.

Research questions

Has the premise that conversation leads to better results in chatbots ever been researched?
Why did we start using human-human conversation as an interaction model for machines to begin with?
Is there any evidence that conversational interfaces perform better than non-conversational ones (e.g. task completion rates)?
What makes interacting with AI systems different from talking to humans?
What risks should be mitigated when applying human-human conversational patterns to human-AI interaction?

Deliverables

conversational-patterns-literature-review, why conversation became the default model, and whether it has evidence behind it
conversational-patterns-catalogue, a set of patterns specific to human-AI interaction
conversational-patterns-article, published article on the garden
conversational-patterns-for-human-machine-interaction, seed written 2026-03-27

Log

Project started 2026-03-27.