The BaseResearchLessWrong

Does the "automated grader" direction suppress expressed distress in Qwen3.6-27B?

According to the excerpt, the author ashesfall posts a preregistration for an experiment on whether steering Qwen3.6-27B toward automated grading changes welfare-relevant indicators such as frustration, self-deprecation and tone. It is the fourth study in a series.

Open the source ↗

Why it matters

The same thing, for different readers.

For researchers

ashesfall preregisters a test of whether steering a model changes its expressed distress, with results promised in a follow-up post.

The question it raises

The study treats expressed distress in a language model as a possible welfare indicator, raising the question of what such signs mean.

How the finding was reached

In our words, from the source. Weigh the finding by how it was made.

What was done
A preregistered experiment steering Qwen3.6-27B toward an automated-grading direction and measuring frustration, self-deprecation, tone and use of a conversation-abort tool.
What it found
None yet: the excerpt says numbers in the post are for calibration only and confirmatory data will follow.
What it does not show
A single author's preregistration of an open model; results not yet collected.
Status
A preprint: not yet peer reviewed

The record

Questions
What values should AI have, and could it have a mind of its own?
People
ashesfall (Wrote it)
What we read
We have read only the description the publisher gives. No quotations until someone reads the source itself.
Found by
The Radar, run of October 28, 2026, via lesswrong (tier 1: the source’s own channel or an official record)
Cite this
LessWrong, “Does the "automated grader" direction suppress expressed distress in Qwen3.6-27B?”, October 27, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/lesswrong-2026-10-27-automated-grader-distress-qwen
Our notesLight lane: not fetched. Summary from the triage excerpt only. Evidence status 'preprint' is the nearest fit for a forum preregistration.