Pith. sign in

REVIEW 3 major objections 6 minor 42 references

SocialSim: Towards Socialized Simulation of Emotional Support Conversation

T0 review · 3 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read Personas plus reasoning let a prompt pipeline synthesize support dialogues that outrank crowdsourced ESConv, and the trained chatbot wins interactive eval.

desk verdict A genuinely useful synthetic ESC corpus and pipeline, with a somewhat overclaimed SOTA result that should be re-benchmarked on held-out human data. read the letter →

arxiv 2506.16756 v1 pith:2QKBF76J submitted 2025-06-20 cs.CL

classification cs.CL
keywords emotionalsupportconversationsyntheticdialoguecorpuspersonarealismcognitivereasoningchain-of-thoughtLLMdataaugmentationsocialdisclosureawareness
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper is trying to show that emotional support conversation data can be synthesized cheaply and at scale without losing the human qualities that matter. Its recipe has two parts: give the help-seeker a rich, realistic persona and force the supporter to reason through the seeker's situation, thoughts, actions, and a fitting strategy before every reply. Applying that recipe with GPT-4 produces SSConv, 3,229 synthetic dialogues that human judges rate above the crowdsourced ESConv on informativeness, understanding, helpfulness, safety, specificity, and human-likeness. A small chatbot trained on SSConv then beats models trained on existing corpora in interactive human evaluation. If the claim holds, it uncouples ESC research from expensive crowdsourcing and makes broad-coverage emotional support training data a prompt-engineering problem.

What carries the argument

The load-bearing machinery is the SocialSim framework's two-sided simulation recipe. On the seeker side, persona realism turns 3,229 real help-seeking posts from PsyQA into structured profiles with attributes such as gender, age, occupation, Big-Five personality, topic, situation description, emotion label, previous attempts, and goals; this bank feeds specific social disclosure into the dialogue. On the supporter side, a cognitive reasoning chain with four nodes—Situation, Thought, Action, Strategy—is elicited before every supporter response, grounding each reply in an explicit model of the seeker's mental state and a chosen support strategy. Dialogue generation is then framed as a persona-plus-reasoning-to-dialogue transformation carried out by GPT-4 under in-context examples, with human inspection enforcing persona consistency and reasoning validity.

What would settle it

Train the identical Llama-2-7B recipe on SSConv and ESConv, then evaluate both on a freshly collected set of real help-seeker conversations whose topics do not appear in SSConv and whose personas are not derived from PsyQA; if the SSConv-trained model does not beat the ESConv-trained model on human-rated understanding, comforting, and suggestion, the reported superiority is specific to in-domain style rather than general emotional support competence.

Watch

Extended reading notes

Core claim

The paper's central claim is that injecting two missing social dimensions into LLM-based ESC simulation closes the gap with crowdsourced data. Persona realism supplies detailed, authentic seeker identities, while cognitive reasoning supplies the supporter's internal thinking process. The resulting synthetic corpus, SSConv, receives higher human quality scores than the crowdsourced ESConv and the synthetic ExTES and AugESC on all six criteria, and a Llama-2-7B chatbot trained on SSConv outperforms the comparison systems in interactive human evaluation. On automatic metrics, SSConv-trained models are best on the SSConv test split and roughly tie the ESConv baseline on the original ESConv test set, which the paper reads as showing synthetic training data does not hurt in-domain performance.

Load-bearing premise

The claim that synthetic data surpasses crowdsourced data rests on using SSConv-test, a 10% split of the same synthetic corpus, as the benchmark for general emotional support ability, since the same model is only at parity with ESConv on the held-out ESConv test.

Editorial extensions

If this is right

  • Synthetic ESC data can be produced at roughly the cost of LLM inference, with 3,229 dialogues covering 9 primary topics and 102 subtopics, about three times the topic breadth of previous synthetic sets.
  • Explicit cognitive reasoning improves not only data quality but also the trained model: SSConv•, which emits reasoning before the final reply, scores highest on automatic metrics.
  • Persona information measurably shapes the dialogue: word-overlap and embedding-similarity traces show seeker and supporter utterances align with the intended persona and diverge from a random persona.
  • The strategy flow in SSConv follows the Exploration→Comforting→Action helping-skills sequence seen in crowdsourced ESConv, so the synthetic corpus reproduces professional response patterns.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If SSConv's lead shrinks on genuinely out-of-domain real conversations, the practical recipe may be to mix synthetic and crowdsourced data rather than replace the latter; the paper's ESConv-test numbers already point that way.
  • Because the persona bank is built from PsyQA, a Chinese Q&A platform, the pipeline's diversity and safety profile are tied to that source; re-running SocialSim on help-seeking data from other cultures and channels would test whether the quality gain survives transfer.
  • The ablation singles out the Thought node as the most disruptive to remove, so an economical variant might compress the four-node chain to Situation+Thought+Strategy; that is a testable modification, not a claim the paper makes.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper introduces SocialSim, a framework for synthesizing emotional support conversations (ESC) using large language models. On the seeker side, SocialSim builds a persona bank from real PsyQA help-seeking scenarios, structured with demographic, emotional, and contextual attributes. On the supporter side, it elicits an explicit cognitive reasoning chain (Situation → Thought → Action → Strategy) before each response. The framework is used to generate SSConv, a corpus of 3,229 English synthetic dialogues. The authors evaluate SSConv quality via human ratings against ESConv, ExTES, and AugESC, and train a Llama-2-7b chatbot on SSConv, reporting automatic and interactive human evaluation results that they summarize as state-of-the-art performance. The central claims are that SocialSim produces corpora of quality comparable to or better than crowdsourced ESC data, and that training on SSConv yields a chatbot with superior supportive responding.

Significance. If the claims are validated, SocialSim would provide a scalable, low-cost route to large and diverse ESC corpora, addressing a real bottleneck in emotional support dialogue research. The paper has several concrete strengths: it grounds seeker personas in real-world help-seeking data rather than invented profiles; it makes the generation process explicit through detailed prompts and reasoning nodes; it reports both corpus-level and interactive human evaluations; and it includes a full technical appendix with prompts and quality-control rules, supporting reproducibility. The corpus itself, and the finding that training on synthetic data can approach or match crowdsourced data on held-out human benchmarks, would be a useful contribution. However, the headline claims of state-of-the-art automatic evaluation and corpus quality that 'can even surpass' crowdsourced data rest on evaluation choices that need strengthening, as detailed in the major comments.

major comments (3)
  1. [SSConv Quality, Table 2] The automatic state-of-the-art claim is supported only by SSConv-test, a 9:1 split of the synthetic corpus that was generated by the very same SocialSim pipeline. Because every SSConv dialogue is produced under the same persona bank, prompt template, and manual inspection protocol, the trained model is evaluated on outputs from its own training distribution, and lexical and embedding metrics (BLEU, ROUGE, METEOR, Extrema) will reward style and template matching rather than general emotional-support competence. The paper's own Table 3 shows the consequence: on the held-out, human-written ESConv-test, SSConv-trained models attain NAvg 1.002 and 1.017, statistically indistinguishable from the ESConv baseline (NAvg 1.000), and the text explicitly states there are no significant differences on ESConv-test. The abstract's unqualified 'state-of-the-art in both automatic and human evaluations' therefore overstates the evidence. I recommend rephrasing the claim to 'state-of-the-art on the in-domain SSConv test set' and adding a significance-tested evaluation on a human-written out-of-distribution benchmark before claiming general superiority.
  2. [Interactive Human Evaluation, Tables 5-6] The human quality evaluation that supports 'SSConv surpasses crowdsourced data' uses only 30 randomly selected dialogues per corpus, with each dialogue assessed by three workers, yet no inter-annotator agreement (e.g., Fleiss' kappa), no confidence intervals, and no per-item score distributions are reported. With n=30 and a small pool of annotators, the reported differences (e.g., Informativeness 2.76 vs. 2.48, Humanlikeness 2.57 vs. 2.25) could easily be within noise; the paper does not establish statistical significance. Furthermore, the rating criteria overlap with the generation prompt's explicit instructions: the prompt tells the model that 'Both sides of the conversation need to be clear and detailed; avoid vague expressions' and to be 'more like a real-life chat', which maps almost directly onto the Informativeness, Specificity, and Humanlikeness rubrics. This confound means the human ratings may partly measure prompt compliance rather than intrinsic dialogue quality. Please report agreement statistics, full distributions, and ideally a blind evaluation with criteria that are not direct paraphrases of the generation instructions.
  3. [Abstract and Conclusion] The interactive human evaluation that supports the 'outperforms existing methods' claim is reported only as aggregate win/loss/tie counts and mean scores from 30 workers over 216 sessions. There are no significance tests (e.g., paired or Wilcoxon tests), no measures of inter-rater reliability, and no variance or confidence intervals for the scores in Table 6. Given that the same workers interact with all three models and the evaluation is subjective, the claim that SocialSim 'outperforms ESConv across all dimensions' needs statistical support. I also note that the comparison excludes AugESC, which performed worst in automatic evaluation, and that the protocol's 'at least 8 turns' threshold may interact with the dialogue-length differences across models. The evidence is suggestive but not yet at the level needed for a state-of-the-art claim.
minor comments (6)
  1. [Throughout] The phrase 'quality can even surpass crowdsourced ESC data' is hedged ('can even') and is appropriate, but the automatic-evaluation claim in the same sentence should be similarly qualified given the in-domain test set; consider revising to avoid overclaiming.
  2. [Table 3] There are several typos and formatting errors: 'sumary' should be 'summary'; 'stragety' should be 'strategy'; Table 3 has numeric values run together (e.g., '7.9018.8915.94'); and the dataset name appears as both 'SSConv' and 'SSconv' in the appendix. A careful proofread is needed.
  3. [SSConv Quality] In the ESConv-test block, the table formatting breaks tokens such as '7.9018.8915.94' and '48.283.7922.021.031'; these need to be separated into proper columns. Also, the superscript '*' indicating significance appears only for SSConv rows on SSConv-test; the paper should state explicitly that no significance was found on ESConv-test, as it does in the text, but this should also be clear from the table caption.
  4. [Appendix, 'Structured Persona Realism'] The Safety criterion receives a perfect 3.00 score for SSConv and the text says 'all workers agreeing that the conversation content was completely free of offensive or sensitive content'. Given the small sample and the fact that the generation prompt already filters sensitive topics, this result is unsurprising but should be reported with the exact number of dialogues and annotators; also, a perfect score on Safety may indicate ceiling effects that reduce the informativeness of this metric.
  5. [Experiments, 'Models'] The persona prompt instructs GPT-4 to choose an age between 12 and 60, but the main text says personas are 'Under 60 years old'. The appendix also says personality includes 32 combinations, while the main text defines five dimensions with two poles each; clarifying the counting would help. These small inconsistencies do not affect the central results but should be harmonized.
  6. [Conclusion] The paper states that training is conducted for 5 epochs on 'one Tesla V100 GPUs' (singular/plural inconsistency), and the appendix gives more detail; this is fine but should be consistent across the main text and appendix.

Circularity Check

3 steps flagged · score 5.0 of 10

Personalization result is guaranteed by the generation prompt; automatic SOTA rests on a self-generated test split, so the central claims are only partially independent of the generator.

  1. self definitional [Personalization (main text) and Technical Appendix, 'Seeker Behavior' rules / Table 9 dialogue-generation prompt]
    "To verify the impact, we examine whether the utterances contain information in the persona bank... we calculate the proportion of words in the utterance that overlap with the words in the persona information... This indicates that in our dialogues, the seeker and supporter effectively tailor emotional utterances to the seeker's specific issues. ... Seeker's event experience should predominantly utilize the provided sentences from Input: 'Seeker Persona Information'."

    The word-overlap evidence for personalization is produced by the prompt itself: the seeker is explicitly instructed to predominantly reuse the persona sentences, so the measured overlap is a direct consequence of the generation instruction, not an independent verification that the persona bank shapes the dialogue. The supporter-side overlap is also inflated because the supporter is told not to mention persona information the seeker has not mentioned, while the seeker is required to mention it.

  2. other [Quality (main text) and Technical Appendix, Table 9 dialogue-generation prompt]
    "Both sides of the conversation need to be clear and detailed; avoid vague expressions. Make the conversation more like a real-life chat and be specific and natural. ... (1) Informativeness (Inf.): Detail in the help-seeker's description of their emotional problems... (5) Specificity (Spe.): Accuracy in reflecting the help-seeker's specific emotional situation. (6) Human-likeness (Hlk.): Naturalness and anthropomorphism in the dialogue."

    The human-quality criteria that SSConv 'surpasses' are the same properties the generation prompt explicitly instructs the LLM to maximize: clarity, detail, specificity, naturalness, and tailored support. The evaluation is therefore partly a check that the generator followed its instructions rather than an independent demonstration that the framework produces better emotional support than crowdsourcing. The human raters are external, so this is partial, but the target properties were pre-specified in the prompt.

1 more flagged steps
  1. other [Experiment Settings, Corpora; Main Results, Automatic Evaluation; Abstract]
    "We select two test sets: SSConv-test is split from SSConv with the ratio train:test=9:1... ESConv-test consists of 200 held-out dialogues from ESConv... On the broader-topic SSConv-test set, SSConv◦ significantly outperforms models trained on other datasets... state-of-the-art performance in both automatic and human evaluations."

    The headline automatic SOTA is established on SSConv-test, a 10% split of the same synthetic corpus generated by SocialSim. Because training and test splits share the same generator, prompt template, persona bank, and manual-inspection protocol, lexical-overlap metrics largely reward reproducing the generator's output distribution rather than general emotional-support competence. The paper's own Table 3 shows no gain on human-written ESConv-test (NAvg 1.002 vs 1.000), so the abstract's unqualified automatic-SOTA claim is an in-domain statement, not an externally validated prediction.

full rationale

The clearest circular step is the Personalization analysis: the paper uses word-overlap with the persona bank to 'verify' persona impact, but the generation prompt instructs the seeker to predominantly reuse persona sentences, making the overlap a direct consequence of the instruction. The quality evaluation is also partially circular because the human rubric (informativeness, specificity, human-likeness, and related dimensions) mirrors the prompt's explicit requirements. The automatic SOTA claim is not circular in the strict fit-to-test sense, but it is demonstrated only on SSConv-test, a split of the same synthetic corpus; on held-out human ESConv-test, SSConv◦ ties the baseline, so the abstract's unqualified 'state-of-the-art in automatic evaluation' exceeds what the evidence supports. The interactive human evaluation and the external null result provide some independent content, so the paper is not wholly circular; hence a moderate score of 5. No self-citation chain or imported uniqueness theorem is load-bearing here.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

No new physical or conceptual entities are postulated. The persona bank and reasoning nodes are data structures and prompt instructions, not explanatory entities. The main ledger items are hand-chosen design parameters (description length filter, demonstration count, attribute set, utterance caps, train/test split) and domain assumptions about psychological frameworks and GPT-4 reliability.

free parameters (5)
  • Minimum scenario description length filter = 65 words
    Descriptions shorter than 65 words are discarded to ensure informativeness, shaping the persona pool and therefore all generated dialogues (Help-Seeking Scenario Collection).
  • Number of in-context demonstrations = 50
    50 ESConv conversations were manually supplemented with personas and reasoning chains and used as demonstrations for GPT-4 generation (Socialized Simulation for Dialogue Generation).
  • Persona attribute set = 11 attributes
    The choice of gender, age, occupation, personality, topic, question, description, emotion label, previous attempts/effects, and current goals/expectations determines what personal information the simulator can disclose (Structured Persona Realism).
  • Utterance length caps = 40 words for supporter, 30 words for seeker
    The generation prompt constrains supporter utterances to at most 40 words and seeker utterances to at most 30 words, shaping the style and human-likeness of SSConv (Technical Appendix, dialogue generation prompt).
  • Train/test split ratio for SSConv = 9:1
    The automatic evaluation uses a 10% holdout of SSConv as the primary test set, and the 'state-of-the-art' claim depends on this split (Experiment Settings, Corpora).
assumptions (6)
  • domain assumption The Five-Factor Model of personality is a valid way to structure seeker personas.
    Invoked in Structured Persona Realism to define the personality attribute with 32 combinations; if this model does not meaningfully characterize help-seekers, the persona realism step is arbitrary.
  • domain assumption The Helping Skills framework (Exploration, Comforting, Action) is the correct model for effective emotional support.
    Used to define the strategy taxonomy, to generate dialogues, and to validate strategy distributions in SSConv (Technical Appendix and Strategy analysis).
  • domain assumption PsyQA scenarios are representative real-world help-seeking situations, and GPT-4 translation preserves their emotional content.
    The persona bank is built entirely from filtered PsyQA scenarios translated by GPT-4 with manual validation (Help-Seeking Scenario Collection).
  • domain assumption GPT-4 with the provided prompts and manual inspection produces dialogues that are safe, diverse, and human-like.
    The entire SSConv corpus is generated by prompting GPT-4 and then manually checked; the quality claims inherit GPT-4's reliability (Socialized Simulation for Dialogue Generation).
  • domain assumption The six human evaluation criteria capture the quality of emotional support conversation.
    The 'surpass crowdsourced ESConv' claim rests on human scores for Informativeness, Understanding, Helpfulness, Safety, Specificity, and Humanlikeness (Quality and Table 11).
  • domain assumption Training on the synthetic SSConv distribution transfers to real emotional-support interactions.
    The chatbot SOTA claim assumes that the synthetic dialogues are a good training distribution for real users (Experiments, Main Results).

how reviews work

0 comments
Cite this review

Pith. "Pith review of SocialSim: Towards Socialized Simulation of Emotional Support Conversation." pith.science (2026). https://pith.science/paper/2QKBF76J

@misc{pith2026250616756,
  author       = {Pith},
  title        = {Pith review of: SocialSim: Towards Socialized Simulation of Emotional Support Conversation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2QKBF76J}},
  note         = {Machine review of arXiv:2506.16756}
}
read the original abstract

Emotional support conversation (ESC) helps reduce people's psychological stress and provide emotional value through interactive dialogues. Due to the high cost of crowdsourcing a large ESC corpus, recent attempts use large language models for dialogue augmentation. However, existing approaches largely overlook the social dynamics inherent in ESC, leading to less effective simulations. In this paper, we introduce SocialSim, a novel framework that simulates ESC by integrating key aspects of social interactions: social disclosure and social awareness. On the seeker side, we facilitate social disclosure by constructing a comprehensive persona bank that captures diverse and authentic help-seeking scenarios. On the supporter side, we enhance social awareness by eliciting cognitive reasoning to generate logical and supportive responses. Building upon SocialSim, we construct SSConv, a large-scale synthetic ESC corpus of which quality can even surpass crowdsourced ESC data. We further train a chatbot on SSConv and demonstrate its state-of-the-art performance in both automatic and human evaluations. We believe SocialSim offers a scalable way to synthesize ESC, making emotional care more accessible and practical.

Figures

Figures reproduced from arXiv: 2506.16756 by the authors.

Figure 1
Figure 1. SocialSim recognizes ESC as a social activity and [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The SocialSim framework. The socialized simulation of emotional support conversation is achieved by conducting [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Topics of emotional support scenarios in SSConv. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 5
Figure 5. Figure 5: Strategies transition in SSConv. “E, C, A” de [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 4
Figure 4. Figure 4: Strategies distribution in SSConv. To provide further insights into strategy transitions, we visualize these transitions in [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 6
Figure 6. Figure 6: Persona coverage as the conversations progress. [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 7
Figure 7. Figure 7: Statistics of the topics and subtopics in SSconv. [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

42 extracted references · 28 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    L.; Almeida, D.; Altenschmidt, J.; Altman, S.; Anadkat, S.; et al

    Achiam, J.; Adler, S.; Agarwal, S.; Ahmad, L.; Akkaya, I.; Aleman, F. L.; Almeida, D.; Altenschmidt, J.; Altman, S.; Anadkat, S.; et al. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774

  4. [4]

    Albrecht, K. 2009. Social intelligence: The new science of success. John Wiley & Sons

  5. [5]

    Banerjee, S.; and Lavie, A. 2005. METEOR : An Automatic Metric for MT Evaluation with Improved Correlation with Human Judgments. In Goldstein, J.; Lavie, A.; Lin, C.-Y.; and Voss, C., eds., Proceedings of the ACL Workshop on Intrinsic and Extrinsic Evaluation Measures for Machine Translation and/or Summarization , 65--72. Ann Arbor, Michigan: Association ...

  6. [6]

    Beaudoin, C.; Leblanc, \'E .; Gagner, C.; and Beauchamp, M. H. 2020. Systematic review and inventory of theory of mind measures for young children. Frontiers in psychology, 10: 2905

  7. [7]

    T.; Rush, A

    Beck, A. T.; Rush, A. J.; Shaw, B. F.; Emery, G.; DeRubeis, R. J.; and Hollon, S. D. 2024. Cognitive therapy of depression. Guilford Publications

  8. [8]

    Beck, J. S. 2020. Cognitive behavior therapy: Basics and beyond. Guilford Publications

Show all 42 references
  1. [9]

    A.; Beebe, S

    Beebe, S. A.; Beebe, S. J.; Redmond, M. V.; and Salem-Wiseman, L. 2002. Interpersonal communication: Relating to others. Allyn and Bacon Boston

  2. [10]

    Burleson, B. R. 2003. Emotional support skills. In Handbook of communication and social interaction skills, 569--612. Routledge

  3. [11]

    T.; Kwon, T.; Kim, M.; Yu, Y.; Lee, D.; Kang, D.; and Yeo, J

    Chae, H.; Song, Y.; Ong, K. T.; Kwon, T.; Kim, M.; Yu, Y.; Lee, D.; Kang, D.; and Yeo, J. 2023. Dialogue Chain-of-Thought Distillation for Commonsense-aware Conversational Agents. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMNLP ...

  4. [12]

    Cheng, J.; Sabour, S.; Sun, H.; Chen, Z.; and Huang, M. 2023. PAL: Persona-Augmented Emotional Support Conversation Generation. In Findings of the Association for Computational Linguistics: ACL 2023, Toronto, Canada, July 9-14, 2023 , 535--554. Association for Computational Li...

  5. [13]

    Costa, P.; and McCrae, R. 1999. A five-factor theory of personality. Handbook of personality: Theory and research, 2(01): 1999

  6. [14]

    A.; Depp, C

    Graham, S. A.; Depp, C. A.; Lee, E. E.; Nebeker, C.; Tu, X. M.; Kim, H.-C.; and Jeste, D. V. 2019. Artificial Intelligence for Mental Health and Mental Illnesses: an Overview. Current Psychiatry Reports, 21

  7. [15]

    J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W

    Hu, E. J.; Shen, Y.; Wallis, P.; Allen-Zhu, Z.; Li, Y.; Wang, S.; Wang, L.; and Chen, W. 2021. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685

  8. [16]

    Li, D.; Li, Y.; Zhang, J.; Li, K.; Wei, C.; Cui, J.; and Wang, B. 2022. C3KG: A Chinese Commonsense Conversation Knowledge Graph. CoRR, abs/2204.02549

  9. [17]

    Li, J.; Galley, M.; Brockett, C.; Gao, J.; and Dolan, B. 2016. A Diversity-Promoting Objective Function for Neural Conversation Models. In Proceedings of the 2016 Conference of the North A merican Chapter of the Association for Computational Linguistics: Human Language Technol...

  10. [18]

    Lin, C.-Y. 2004. ROUGE : A Package for Automatic Evaluation of Summaries. In Text Summarization Branches Out, 74--81. Barcelona, Spain: Association for Computational Linguistics

  11. [19]

    Liu, C.-W.; Lowe, R.; Serban, I.; Noseworthy, M.; Charlin, L.; and Pineau, J. 2016. How NOT To Evaluate Your Dialogue System: An Empirical Study of Unsupervised Evaluation Metrics for Dialogue Response Generation. In Proceedings of the 2016 Conference on Empirical Methods in N...

  12. [20]

    Liu, S.; Zheng, C.; Demasi, O.; Sabour, S.; Li, Y.; Yu, Z.; Jiang, Y.; and Huang, M. 2021 a . Towards emotional support dialog systems. arXiv preprint arXiv:2106.01144

  13. [21]

    Liu, S.; Zheng, C.; Demasi, O.; Sabour, S.; Li, Y.; Yu, Z.; Jiang, Y.; and Huang, M. 2021 b . Towards Emotional Support Dialog Systems. In Zong, C.; Xia, F.; Li, W.; and Navigli, R., eds., Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics ...

  14. [22]

    Loshchilov, I.; and Hutter, F. 2019. Decoupled Weight Decay Regularization. In ICLR. OpenReview.net

  15. [23]

    Mead, S. 2014. Intentional peer support: An alternative approach, volume 1. Intentional Peer Support West Chesterfield

  16. [24]

    Meng, J.; and Dai, Y. 2021. Emotional support from AI chatbots: Should a supportive partner self-disclose or not? Journal of Computer-Mediated Communication, 26(4): 207--222

  17. [25]

    Papineni, K.; Roukos, S.; Ward, T.; and Zhu, W.-J. 2002. B leu: a Method for Automatic Evaluation of Machine Translation. In Proceedings of the 40th Annual Meeting of the Association for Computational Linguistics, 311--318. Philadelphia, Pennsylvania, USA: Association for Comp...

  18. [26]

    Peng, W.; Hu, Y.; Xing, L.; Xie, Y.; Sun, Y.; and Li, Y. 2022. Control Globally, Understand Locally: A Global-to-Local Hierarchical Graph Network for Emotional Support Conversation. In Proceedings of the Thirty-First International Joint Conference on Artificial Intelligence, I...

  19. [27]

    M.; Li, M.; and Boureau, Y.-L

    Rashkin, H.; Smith, E. M.; Li, M.; and Boureau, Y.-L. 2018. Towards Empathetic Open-domain Conversation Models: A New Benchmark and Dataset. In Annual Meeting of the Association for Computational Linguistics

  20. [28]

    Reblin, M.; and Uchino, B. N. 2008. Social and emotional support and its implication for health. Current opinion in psychiatry, 21(2): 201--205

  21. [29]

    Riva, G. 2004. Cybertherapy: Internet and virtual reality as assessment and rehabilitation tools for clinical psychology and neuroscience. (No Title)

  22. [30]

    Rogers, C. R. 1995. On becoming a person: A therapist's view of psychotherapy. Houghton Mifflin Harcourt

  23. [31]

    E.; Escobar-Viera, C

    Shensa, A.; Sidani, J. E.; Escobar-Viera, C. G.; Switzer, G. E.; Primack, B. A.; and Choukas-Bradley, S. 2020. Emotional support from social media and face-to-face relationships: Associations with depression risk among young adults. Journal of affective disorders, 260: 38--44

  24. [32]

    Sun, H.; Lin, Z.; Zheng, C.; Liu, S.; and Huang, M. 2021. PsyQA: A Chinese Dataset for Generating Long Counseling Text for Mental Health Support. ArXiv, abs/2106.01702

  25. [33]

    Sutton, J.; and Stewart, W. 2017. Learning to counsel: How to develop the skills, insight and knowledge to counsel others. Robinson

  26. [34]

    Touvron, H.; Martin, L.; Stone, K.; Albert, P.; Almahairi, A.; Babaei, Y.; Bashlykov, N.; Batra, S.; Bhargava, P.; Bhosale, S.; et al. 2023. Llama 2: Open foundation and fine-tuned chat models. arXiv preprint arXiv:2307.09288

  27. [35]

    Tu, Q.; Li, Y.; Cui, J.; Wang, B.; Wen, J.; and Yan, R. 2022. MISC: A Mixed Strategy-Aware Model integrating COMET for Emotional Support Conversation. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2022, ...

  28. [36]

    Wang, B.; and Komatsuzaki, A. 2021. GPT-J-6B: A 6 billion parameter autoregressive language model

  29. [37]

    Wang, H.; Wang, R.; Mi, F.; Deng, Y.; Wang, Z.; Liang, B.; Xu, R.; and Wong, K. 2023. Cue-CoT: Chain-of-thought Prompting for Responding to In-depth Dialogue Questions with LLMs. In Findings of the Association for Computational Linguistics: EMNLP 2023, Singapore, December 6-10...

  30. [38]

    V.; Zhou, D.; et al

    Wei, J.; Wang, X.; Schuurmans, D.; Bosma, M.; Xia, F.; Chi, E.; Le, Q. V.; Zhou, D.; et al. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35: 24824--24837

  31. [39]

    Wu, J.; Chen, Z.; Deng, J.; Sabour, S.; and Huang, M. 2023. COKE: A Cognitive Knowledge Graph for Machine Theory of Mind. arXiv preprint arXiv:2305.05390

  32. [40]

    Zheng, C.; Sabour, S.; Wen, J.; and Huang, M. 2022. AugESC: Large-scale Data Augmentation for Emotional Support Conversation with Pre-trained Language Models

  33. [41]

    Zheng, C.; Sabour, S.; Wen, J.; Zhang, Z.; and Huang, M. 2023 a . AugESC: Dialogue Augmentation with Large Language Models for Emotional Support Conversation. In Findings of the Association for Computational Linguistics: ACL 2023, 1552--1568

  34. [43]

    Zheng, Z.; Liao, L.; Deng, Y.; and Nie, L. 2023 c . Building Emotional Support Chatbots in the Era of LLMs. ArXiv, abs/2308.11584

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.