PsyDefConv

PsyDefConv is the first conversational dataset annotating Defense Mechanism Rating Scale (DMRS) defense levels at the utterance level, built on top of ESConv.

📄 Paper: You Never Know a Person, You Only Know Their Defenses: Detecting Levels of Psychological Defense Mechanisms in Supportive Conversations (ACL 2026)

Overview

• 200 dialogues / 2,336 seeker utterances labeled with defense levels
• 9 labels: 7 DMRS levels + No Defense (0) + Needs More Information (8)
• Double-blind annotation, Cohen's κ = 0.639 (substantial)
• 80/20 stratified split (train / test)

Label Scheme

Label	Name
0	No Defense
1	Action
2	Major Image-Distorting
3	Disavowal
4	Minor Image-Distorting
5	Neurotic
6	Obsessional
7	High-Adaptive
8	Needs More Information

See the paper for full definitions and examples.

Citation

@inproceedings{na-etal-2026-psydefconv,
  title = "You Never Know a Person, You Only Know Their Defenses:
           Detecting Levels of Psychological Defense Mechanisms
           in Supportive Conversations",
  author = "Na, Hongbin and Wang, Zimu and Chen, Zhaoming and
            Zhou, Peilin and Hua, Yining and Zhou, Grace Ziqi and
            Zhang, Haiyang and Shen, Tao and Wang, Wei and
            Torous, John and Ji, Shaoxiong and Chen, Ling",
  booktitle = "Findings of the Association for Computational
               Linguistics: ACL 2026",
  month = jul,
  year = "2026",
  address = "San Diego, USA",
  publisher = "Association for Computational Linguistics",
}

Please also cite ESConv (Liu et al., 2021).