A general-purpose LLM assessing team emails' emotional tone was biased toward interpreting messages as anxiety-related only¶
Claim · All claims
Evidence · 1 study · designr2·q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly
Subclaims¶
Reading the evidence codes
q is the design tier, i the impact, n the sample, and the kind with its r
the rigour judged within that kind: what each code means.
This claim: no standardized effect is recorded in observations/ for it yet, so no benchmark reading is possible.
q2 i? When experimenting with categorizing teams' email emotional tone, the general LLM almost always interpreted messages as exhibiting only anxiety-related emotions, likely due to the prevalence of incident words in the dataset. → V. Švábenský 2026
Evidence¶
V. Švábenský 2026¶
V. Švábenský, J. Vykopal, S. Leelaluk, P. Čeleda, F. Okubo, A. Shimada. (2026). Assessment in Team Problem-Solving Exercises in Computing Education. Proceedings of the 56th IEEE Frontiers in Education Conference (FIE '26). https://arxiv.org/abs/2607.19209 doi:10.48550/arxiv.2607.19209
q2 · i? · design · r2
Exploratory experiment (RQ2) in which the authors asked the LLM to categorize the emotional tone of teams' TTX emails; the authors attribute the bias to incident vocabulary and conclude valid results require framing assessment within the TTX context or a domain-specific model such as CyLLM.
"However, the general LLM was biased to almost always interpret messages as exhibiting only anxiety-related emotions. Most likely, this stemmed from the prevalence of incident words (e.g., "cyber attack" or "data loss") in the dataset."
Discussion¶
Related Claims¶
- GPT-4o's rubric-guided grading of team communication disagreed with instructors at near-chance levels (55% RMSE), while GPT-5.2 improved but remained insufficiently aligned (35%) — related
- Perceptions of feedback characteristics vary between researcher and teacher coders, with agreement shaped by professional background — related
- Fully inductive LLM codebook development risks importing unexamined sensitizing concepts, such as folk theories and scientific misconceptions, from the model's training data — a broader claim this one bears on
- Some LLMs exhibit a selection bias against selecting option D on multiple-choice pedagogy questions — related
- Priming human shadows as math lovers or haters shifts LLM emotional tone about mathematics — related
Page metadata
| Field | Value |
|---|---|
type |
claim |
id |
llm-emotional-tone-bias-incident-words |
title |
A general-purpose LLM assessing team emails' emotional tone was biased toward interpreting messages as anxiety-related only |
description |
A general-purpose LLM assessing team emails' emotional tone was biased toward interpreting messages as anxiety-related only |
status |
review |
evidence_strength |
weak |
newest_evidence |
2026 |
recheck_by |
2029 |
generated |
by process:wiki-ingest on 2026-10-10 |
trust |
unverified — no one has confirmed this page's content |
sources |
1 — v-švábenský-2026 |
Frontmatter is the page's machine-readable record — see the Schema & Guide.