User-user-machine interfaces

Watching Engelbart’s 1968 demo, picked up on the two-person collaboration scenes (Bill on the intercom, bug-fight, shared screen). Reading more on his framework, the pattern becomes load-bearing.

We shouldn’t only assume that we need to create user-machine interface, but user-user-machine interfaces too.

Engelbart’s 1962 framework calls the augmentation system H-LAM/T: Human using Language, Artifacts, Methodology, in which he is Trained. Four components, not one. The artifact (the tool, the UI) is one quarter of the system. Designing only the artifact and ignoring language, methodology, and training is what most of UX has done for forty years. It is also a fraction of the augmentation Engelbart had in mind.

He named the proper deployment context a Networked Improvement Community (NIC): a group consciously using shared tools to improve their shared practice. The unit of analysis isn’t the individual user. It’s the community co-evolving with its tools.

The implication for our current LLM moment: stop designing chat as a one-human, one-agent surface. The real win is user plus user plus machine: shared workspaces with the LLM as a present-but-not-central participant, where two humans get sharper because the third presence can fetch evidence in real time.

For the Guildford talk: this is the missing frame. Conversation design has been “user-machine dialogue” since the 1960s. Engelbart was already past that in 1962.