EduBehaviors¶
Research Method · All research methods
Evidence · 3 claims (3 for) · 2 studies (2 design),q2· 0 of 2 report an effect size · 3 claims rest on one study
Description¶
EduBehaviors is an interpretable annotation framework in which researchers write explicit binary assertions about conversational behavior; LLMs annotate each assertion as true or false, and a rule or trained statistical model combines these values into a construct label. The article formalizes this as a decomposable schema S=(A,R), noting it is "an example of a concept bottleneck model". It is designed to make coding decisions inspectable and revisable.
Accounts¶
- EduBehaviors: decomposable assertion-based schemas separating behavior identification from construct labeling: EduBehaviors is an interpretable annotation framework in which researchers write explicit binary assertions about conversational behavior; LLMs annotate each assertion as true or false, and a rule or trained statistical model combines these values into a construct label. The article formalizes this as a decomposable schema S=(A,R), noting it is "an example of a concept bottleneck model". It is designed to make coding decisions inspectable and revisable. (Julian Bernado et al. (2026))
Claims¶
- EduBehaviors achieves 0.673 macro-F1 and 0.688 Cohen's kappa on Teacher TalkMoves, generally improving over direct LLM prompting [+M]
- Adding a cross-model agreement filter improves performance for the top three LLM annotators [+M]
- Adding regex-based rules to few-shot LLM labeling improves agreement with human annotation (0.818 to 0.852 overall) [+W]
Related Research Methods¶
-
Key Sources¶
- Julian Bernado, Ana Trindade Ribeiro, Xander Beberman, and Susanna Loeb. (2026). EduBehaviors: Assertion-Based Schemas for Auditable Coding of Educational Dialogues. https://arxiv.org/abs/2609.27043
Page metadata
| Field | Value |
|---|---|
type |
research-method |
id |
edubehaviors |
title |
EduBehaviors |
description |
EduBehaviors is an interpretable annotation framework in which researchers write explicit binary assertions about conversational behavior; LLMs annotate each assertion as true or false, and a rule or trained statistical model combines these values into a construct label. |
status |
review |
generated |
by process:settle-candidates on 2026-10-10 |
trust |
unverified — no one has confirmed this page's content |
Frontmatter is the page's machine-readable record — see the Schema & Guide.