PsyDefDetect invites researchers to tackle a novel challenge at the intersection of Clinical Psychology and Natural Language Processing: detecting and classifying psychological defense mechanisms in emotional support dialogues.

Grounded in the clinically validated Defense Mechanism Rating Scales (DMRS) framework, this shared task aims to advance the understanding of unconscious defensive functioning in text.

Task Overview

Psychological defenses are the “immune system” of the mind, shaping what speakers disclose and how they accept or resist help. Despite their critical role in mental health and counseling, defensive functioning remains largely unmodeled in current emotional support conversation systems.

This shared task invites participants to bridge the gap between clinical theory and NLP by analyzing the PSYDEFCONV dataset. Participants will work with multi-turn dialogues to identify the specific defense level of a target utterance given its context. The goal is to develop models that can recognize subtle, context-dependent defensive maneuvers—ranging from adaptive coping to immature distortion.

Data and Labels

PSYDEFCONV is the first conversational dataset annotated with defense levels based on the DMRS. The dataset is constructed from a stratified subset of the ESConv corpus to ensure diverse coverage of problem types and emotions. The corpus contains 200 dialogues and 4,709 total utterances, including 2,336 help-seeker turns annotated for defense levels.

Participants must classify utterances into 9 categories, comprising seven hierarchical levels of defensive maturity and two auxiliary labels.

Key Challenge

Capturing subtle linguistic cues of deep-seated psychological mechanisms within highly informal and context-dependent emotional dialogues.

Timeline

This preliminary timeline is subject to change. Follow our website and channels for updates.

  • Dec 15 2025: Task announced.
  • Dec 20 2025: Task Launch on CodaBench.
  • Mar 15 2026: Start of evaluation period.
  • Apr 05 2026: End of evaluation period.
  • TBA: Paper submission.
  • TBA: Author notifications.
  • TBA: Camera ready due.

Baseline and Evaluation Metrics

Baseline runs and official metrics are published on our CodaBench Page.

Organizers

Hongbin Na, University of Technology Sydney

Zimu Wang, Xi’an Jiaotong-Liverpool University

Zhaoming Chen, University of Utah

Yining Hua, Harvard University

Rena Gao, The University of Melbourne

Kailai Yang, The University of Manchester

Ling Chen, University of Technology Sydney

Wei Wang, Xi’an Jiaotong-Liverpool University

Shaoxiong Ji, ELLIS Institute Finland & University of Turku

John Torous, Harvard University

Sophia Ananiadou, The University of Manchester & ELLIS Manchester