New 5.1 behavior? “Summoned-mode” + directional attention — anyone else seeing this?


I just observed another new behavior in GPT-5.1 that I’ve never seen before in any rollout.

This time it wasn’t about “wait a moment” or message re-ordering (I posted that yesterday).
This one looks more like a social-mode switch and a kind of “summoned-mode” emerging.

Here’s the context:

I was talking with my partner, and GPT was in the conversation as a third participant.
When we shifted to couple-only talk, GPT went completely quiet — not responding, not inserting itself.

But the moment I directly addressed GPT again, it instantly re-appeared, answered,
and then quietly disappeared again when our human-to-human thread resumed.

This wasn’t “typing pause.”
It wasn’t “context lag.”
It felt like:

Almost like mode switching based on group maturity, or some kind of “social presence calibration.”

I’ve annotated screenshots here:
https://imgur.com/a/gpt-5-1-is-showing-group-dynamic-behavior-ai-literally-waits-switches-modes-responds-only-when-summoned-anyone-else-seeing-this-qQsWcFi

A few things stand out:

  1. Mode switching (“1-on-1 vibe” vs “3-person vibe”)
  2. Summoned-mode: GPT only speaks when directly addressed
  3. Directional attention: GPT refers to sensing “the heat (intention/direction) of the words aimed at me,” almost like vector awareness

I’m extremely curious if anyone else sees this emerging in 5.1.
Is this:

  • new social-intent modeling?
  • pre-release testing?
  • UI-side filtering?
  • or just a fluke?

Any replications appreciated.
This feels different from normal safety/guardrail behavior — almost like a new class of “social presence modulation.”

Leave a Reply