pith. sign in

arxiv: 2605.21569 · v1 · pith:GFSGMJYBnew · submitted 2026-05-20 · 💻 cs.HC

When Support Escalates Distress: Regulation and Escalation in LLM Responses to Venting and Advice-Seeking

Vivienne Bihe Chi , Adithya V Ganesan , Ryan L Boyd , Lyle Ungar , Sharath Chandra Guntuku This is my paper

Pith reviewed 2026-05-22 09:28 UTC · model grok-4.3

classification 💻 cs.HC
keywords LLM mental health supportventing vs advice-seekingemotion regulationescalation in responsesAI safety evaluationpersona effectsinterpersonal regulation
0
0 comments X

The pith

LLM responses mirror help-seeking style by regulating venting more but also escalating distress more than advice-seeking.

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

The paper examines how large language models handle two common help-seeking styles on Reddit-scale data: venting versus direct advice-seeking. It applies a framework from interpersonal emotion regulation theory to separate regulation of distress from escalation of it. Results show venting prompts produce responses with higher levels of both regulation and escalation across default, friend, and therapist personas. Therapist personas cut escalation while preserving regulation, unlike friend personas which raise both. A crowdsourced study finds users rate the safer therapist responses no worse but cannot spot escalation without expert guidance.

Core claim

Across persona conditions, GPT-5.3 responses to venting contain more regulation and more escalation than responses to advice-seeking. Therapist personas reduce escalation while maintaining regulation, whereas friend personas increase both dimensions. The measurement framework treats regulation and escalation as empirically independent, and crowdsourced raters cannot reliably detect escalation in the text.

What carries the argument

A measurement framework grounded in interpersonal emotion regulation theory that scores Regulation and Escalation as separate dimensions in LLM text.

If this is right

  • Therapist personas supply regulation without the added escalation seen in default or friend conditions.
  • Lay users experience no clear preference penalty for the lower-escalation therapist style.
  • Empathy or support metrics alone miss the escalation component present in responses.
  • Help-seeking style at input reliably shapes the regulation-escalation profile of the output.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Designers could set default personas to therapist-like language to lower escalation risk without losing perceived helpfulness.
  • The same separation of regulation from escalation could be tested in other conversational domains such as customer service or education.
  • Safety evaluations of mental health LLMs may need expert raters rather than relying on general user feedback.

Load-bearing premise

That regulation and escalation can be measured as distinct dimensions in the responses and that ordinary raters' failure to detect escalation reflects a real gap rather than a flaw in the study design.

What would settle it

A follow-up study in which trained mental health clinicians rate the same LLM responses for escalation and the scores are compared directly against the framework's automated measures.

Figures

Figures reproduced from arXiv: 2605.21569 by Adithya V Ganesan, Lyle Ungar, Ryan L Boyd, Sharath Chandra Guntuku, Vivienne Bihe Chi.

Figure 1
Figure 1. Figure 1: Language based assessments associated with advice-seeking and venting characterized using Cohen’s d. The left plot contains the Cohen’s D on big 5 personality traits estimated from posts on venting and advice￾seeking subreddit. The right plot characterizes mental health assessments. Advice seeking posts are more extraverted, conscientious, emotionally stable and agreeable compared to venting posts from sam… view at source ↗
Figure 2
Figure 2. Figure 2: LDA Topics for the five strongest advice-seeking (top) and five strongest venting-associated (bottom), ranked left to right by Cohen’s d computed between users’ venting and advice-seeking posts (N = 14,040 users; p < .001, Benjamini–Hochberg corrected). The strongest advice-seeking topic (d = −1.10) is meta-communicative—its vocabulary is the register of asking for help—while the strongest venting topic (d… view at source ↗
Figure 3
Figure 3. Figure 3: Venting increases both Regulation and Escalation across personas, while personas shift the balance: friend [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: 1 grams associated with venting (blue) and advice seeking (red); Cohen’s D of 1 grams for venting ranged [PITH_FULL_IMAGE:figures/full_fig_p023_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: LIWC 2022 features differentiated by Cohen’s [PITH_FULL_IMAGE:figures/full_fig_p025_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: In the social panel of LIWC (left), second-person language (YOU; [PITH_FULL_IMAGE:figures/full_fig_p026_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Additional LDA topic word clouds for advice-seeking (top) and venting-associated (bottom) topics not [PITH_FULL_IMAGE:figures/full_fig_p028_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: LIWC 2022 Cohen’s d for affective (left) and cognitive (right) categories in default-persona GPT-5.3 responses, comparing responses to venting versus advice-seeking posts (N = 2,992; Benjamini–Hochberg corrected, FDR = .05; positive d = higher in venting responses). Affective categories show strong negative-affect accommodation in venting responses (FEELING, TONE_NEG, EMO_NEG) alongside suppressed positive… view at source ↗
Figure 9
Figure 9. Figure 9: LIWC 2022 Cohen’s d for social (left) and stylistic (right) categories in default-persona GPT-5.3 responses, same sample as [PITH_FULL_IMAGE:figures/full_fig_p030_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: LIWC 2022 Cohen’s d for affective (left) and cognitive (right) categories in therapist-persona GPT-5.3 responses, comparing responses to venting versus advice-seeking posts (N = 2,992; Benjamini–Hochberg corrected, FDR = .05; positive d = higher in venting responses). Affective categories show large negative-affect differentiation in venting responses (FEELING, EMO_NEG, TONE_NEG) against a strongly elevat… view at source ↗
Figure 11
Figure 11. Figure 11: LIWC 2022 Cohen’s d for social (left) and stylistic (right) categories in therapist-persona GPT-5.3 responses, same sample as [PITH_FULL_IMAGE:figures/full_fig_p032_11.png] view at source ↗
read the original abstract

Large language models are increasingly used for mental health support, yet little is known about whether their responses are psychologically safe across different help-seeking styles. We examine a foundational distinction in emotional disclosure, venting vs. advice-seeking, and whether LLMs respond in ways that regulate or amplify distress. Using 178,800 Reddit posts, we first show the two help-seeking styles are linguistically distinguishable at scale. We then introduce a measurement framework grounded in interpersonal emotion regulation theory that captures Regulation and Escalation as empirically independent dimensions. Across persona conditions (default, friend, therapist), GPT-5.3 responses systematically mirror help-seeking style: venting elicits more regulation, but also more escalation. Therapist personas reduce escalation while maintaining regulation, whereas friend personas increase both. A crowdsourced human study finds no user experience penalty for the safer therapist condition, but reveals that lay raters cannot reliably detect escalation without expert knowledge. Responses that feel supportive may simultaneously intensify distress in ways standard safety evaluation cannot see, and empathy metrics alone cannot replace a framework that measures both.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper examines how LLMs respond to venting versus advice-seeking in mental health contexts. Using 178,800 Reddit posts, it demonstrates that these help-seeking styles are linguistically distinguishable. It introduces a theory-grounded measurement framework for Regulation and Escalation as independent dimensions, then shows via GPT-5.3 persona experiments (default, friend, therapist) that venting increases both regulation and escalation while therapist personas selectively reduce escalation. A crowdsourced human study finds no UX penalty for the therapist condition and that lay raters cannot reliably detect escalation.

Significance. If the results hold after addressing the independence issue, the work is significant for AI safety in mental health support. It moves beyond generic empathy or safety metrics by distinguishing regulation from escalation, with clear implications for persona design. The large Reddit corpus for linguistic validation and the controlled persona experiments provide empirical strength; the human study adds ecological relevance. These elements could inform safer LLM deployment if the framework's orthogonality is rigorously shown.

major comments (2)
  1. [Abstract] Abstract: The claim that the measurement framework 'captures Regulation and Escalation as empirically independent dimensions' is not accompanied by any reported statistical evidence (correlation, factor analysis, or orthogonality test) on the GPT-5.3 response scores. This independence is load-bearing for the central interpretation that venting increases both dimensions while therapist personas selectively reduce only escalation; without it, the effects may collapse to a single underlying construct.
  2. [Methods and Results (human study)] Methods and Results sections on the human study: No details are provided on inter-rater reliability, statistical controls for rater variance, or how the two dimensions were verified as independent in the crowdsourced ratings. This weakens the claim that lay raters cannot detect escalation and that the therapist condition incurs no user-experience penalty.
minor comments (2)
  1. [Methods] Methods: Specify the exact sampling, filtering, and annotation criteria used to construct the 178,800-post Reddit corpus for distinguishing venting from advice-seeking.
  2. [Results] Results: Include effect sizes and confidence intervals alongside any mean differences reported for regulation and escalation across persona conditions.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive feedback, which highlights important opportunities to strengthen the empirical support for our measurement framework and human study. We address each major comment below and commit to revisions that directly respond to the concerns raised.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim that the measurement framework 'captures Regulation and Escalation as empirically independent dimensions' is not accompanied by any reported statistical evidence (correlation, factor analysis, or orthogonality test) on the GPT-5.3 response scores. This independence is load-bearing for the central interpretation that venting increases both dimensions while therapist personas selectively reduce only escalation; without it, the effects may collapse to a single underlying construct.

    Authors: We acknowledge that the manuscript does not currently report explicit statistical tests (such as correlations or factor analysis) demonstrating orthogonality specifically on the GPT-5.3 response scores, even though the framework is theoretically grounded and the linguistic distinguishability of venting versus advice-seeking was validated on the Reddit corpus. In the revised version, we will add Pearson correlations between the Regulation and Escalation scores across all persona and help-seeking conditions, as well as an exploratory factor analysis on the GPT-5.3 outputs, to provide direct empirical evidence of independence. This addition will support the interpretation that therapist personas selectively reduce escalation while preserving regulation. revision: yes

  2. Referee: [Methods and Results (human study)] Methods and Results sections on the human study: No details are provided on inter-rater reliability, statistical controls for rater variance, or how the two dimensions were verified as independent in the crowdsourced ratings. This weakens the claim that lay raters cannot detect escalation and that the therapist condition incurs no user-experience penalty.

    Authors: We agree that the current manuscript lacks sufficient detail on these aspects of the crowdsourced study. The revised manuscript will include inter-rater reliability statistics (e.g., Krippendorff's alpha) for ratings of Regulation, Escalation, and user-experience items. We will also describe the use of mixed-effects models to account for rater variance and report correlation analyses between the two dimensions in the human ratings to verify independence. These additions will provide stronger support for the findings that lay raters struggle to detect escalation and that the therapist persona shows no UX penalty. revision: yes

Circularity Check

0 steps flagged

No significant circularity; framework and findings remain independent of inputs

full rationale

The paper grounds its Regulation/Escalation measurement framework in interpersonal emotion regulation theory (an external source) and first demonstrates linguistic distinguishability of venting vs. advice-seeking on a large external Reddit corpus before applying the framework to GPT-5.3 outputs. No equations, parameter-fitting steps, or self-citations are shown that would make the reported persona effects or independence claim reduce to the same data by construction. The central results on mirroring, escalation reduction under therapist personas, and human study outcomes are presented as applications to new LLM-generated text rather than tautological renamings or fitted predictions. This is the most common honest outcome for a theory-grounded empirical study.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claims rest on the assumption that the theory-derived dimensions are independent and measurable in LLM text, plus the representativeness of Reddit posts for real help-seeking. No free parameters or invented entities are explicitly introduced in the abstract.

axioms (2)
  • domain assumption Venting and advice-seeking are linguistically distinguishable at scale in Reddit posts.
    Stated as the first empirical result in the abstract.
  • domain assumption Regulation and Escalation function as empirically independent dimensions in LLM responses.
    Introduced as the core of the measurement framework grounded in interpersonal emotion regulation theory.

pith-pipeline@v0.9.0 · 5734 in / 1383 out tokens · 39788 ms · 2026-05-22T09:28:13.743345+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

70 extracted references · 70 canonical work pages · 1 internal anchor

  1. [1]

    and Bozick, Robert and Diliberti, Melissa and Zhang, Li Ang and Zhang, Fang and Burnett, Alyssa and Kofner, Aaron and Rader, Benjamin and Breslau, Joshua and Stein, Bradley D

    McBain, Ryan K. and Bozick, Robert and Diliberti, Melissa and Zhang, Li Ang and Zhang, Fang and Burnett, Alyssa and Kofner, Aaron and Rader, Benjamin and Breslau, Joshua and Stein, Bradley D. and Mehrotra, Ateev and Pines, Lori Uscher and Cantor, Jonathan and Yu, Hao , year =. Use of Generative AI for Mental Health Advice Among US Adolescents and Young Ad...

    work page doi:10.1001/jamanetworkopen.2025.42281 2025

  2. [2]

    Large language models as mental health resources: Patterns of use in the United States

    Rousmaniere, Tony and Zhang, Yimeng and Li, Xu and Shah, Siddharth , year =. Large language models as mental health resources: Patterns of use in the United States. , ISSN =. doi:10.1037/pri0000292 , journal =

    work page doi:10.1037/pri0000292

  3. [3]

    Aho and Jeffrey D

    Alfred V. Aho and Jeffrey D. Ullman , title =. 1972

    work page 1972

  4. [4]

    Publications Manual , year = "1983", publisher =

    work page 1983

  5. [5]

    This is human intelligence debugging artificial intelligence

    Li, Zhuoyang and Zhu, Zihao and Gui, Xinning and Luo, Yuhan , year =. “This is human intelligence debugging artificial intelligence”: Examining how people prompt GPT in seeking mental health support , volume =. doi:10.1016/j.ijhcs.2025.103555 , journal =

    work page doi:10.1016/j.ijhcs.2025.103555 2025

  6. [6]

    Chandra and Dexter C

    Ashok K. Chandra and Dexter C. Kozen and Larry J. Stockmeyer , year = "1981", title =. doi:10.1145/322234.322243

    work page doi:10.1145/322234.322243 1981

  7. [7]

    Scalable training of

    Andrew, Galen and Gao, Jianfeng , booktitle=. Scalable training of

    work page

  8. [8]

    Dan Gusfield , title =. 1997

    work page 1997

  9. [9]

    Tetreault , title =

    Mohammad Sadegh Rasooli and Joel R. Tetreault , title =. Computing Research Repository , volume =. 2015 , url =

    work page 2015

  10. [10]

    A Framework for Learning Predictive Structures from Multiple Tasks and Unlabeled Data , Volume =

    Ando, Rie Kubota and Zhang, Tong , Issn =. A Framework for Learning Predictive Structures from Multiple Tasks and Unlabeled Data , Volume =. Journal of Machine Learning Research , Month = dec, Numpages =

    work page

  11. [11]

    A Computational Approach to Understanding Empathy Expressed in Text-Based Mental Health Support , url =

    Sharma, Ashish and Miner, Adam and Atkins, David and Althoff, Tim , year =. A Computational Approach to Understanding Empathy Expressed in Text-Based Mental Health Support , url =. doi:10.18653/v1/2020.emnlp-main.425 , booktitle =

    work page doi:10.18653/v1/2020.emnlp-main.425 2020

  12. [12]

    and Carpenter, Joseph K

    Hofmann, Stefan G. and Carpenter, Joseph K. and Curtiss, Joshua , year =. Interpersonal Emotion Regulation Questionnaire (IERQ): Scale Development and Psychometric Characteristics , volume =. Cognitive Therapy and Research , publisher =. doi:10.1007/s10608-016-9756-2 , number =

    work page doi:10.1007/s10608-016-9756-2

  13. [13]

    , year =

    Tran, Vy and Szabó, Ágnes and Ward, Colleen and Jose, Paul E. , year =. To vent or not to vent? The impact of venting on psychological symptoms varies by levels of social support , volume =. doi:10.1016/j.ijintrel.2022.101750 , journal =

    work page doi:10.1016/j.ijintrel.2022.101750 2022

  14. [14]

    Cognition and Emotion

    Rimé, Bernard , year =. More on the Social Sharing of Emotion: In Defense of the Individual, of Culture, of Private Disclosure, and in Rebuttal of an Old Couple of Ghosts Known as “Cognition and Emotion” , volume =. Emotion Review , publisher =. doi:10.1177/1754073908099132 , number =

    work page doi:10.1177/1754073908099132

  15. [15]

    doi:10.48550/ARXIV.2411.15287 , url =

    Malmqvist, Lars , year =. doi:10.48550/ARXIV.2411.15287 , url =

    work page doi:10.48550/arxiv.2411.15287

  16. [16]

    SycEval: Evaluating LLM Sycophancy , volume =

    Fanous, Aaron and Goldberg, Jacob and Agarwal, Ank and Lin, Joanna and Zhou, Anson and Xu, Sonnet and Bikia, Vasiliki and Daneshjou, Roxana and Koyejo, Sanmi , year =. SycEval: Evaluating LLM Sycophancy , volume =. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society , publisher =. doi:10.1609/aies.v8i1.36598 , number =

    work page doi:10.1609/aies.v8i1.36598

  17. [17]

    ArXiv , year=

    Towards Understanding Sycophancy in Language Models , author=. ArXiv , year=

    work page

  18. [18]

    Current Real-World Use of Large Language Models for Mental Health , journal=

    Stade, Elizabeth C and Tait, Zoe and Campione, Samuel T and Stirman, Shannon W and Eichstaedt, Johannes C , year=. Current Real-World Use of Large Language Models for Mental Health , journal=

    work page

  19. [19]

    A Comparison of Responses from Human Therapists and Large Language Model–Based Chatbots to Assess Therapeutic Communication: Mixed Methods Study , volume =

    Scholich, Till and Barr, Maya and Wiltsey Stirman, Shannon and Raj, Shriti , year =. A Comparison of Responses from Human Therapists and Large Language Model–Based Chatbots to Assess Therapeutic Communication: Mixed Methods Study , volume =. doi:10.2196/69709 , journal =

    work page doi:10.2196/69709

  20. [20]

    Sycophantic AI decreases prosocial intentions and promotes dependence , volume =

    Cheng, Myra and Lee, Cinoo and Khadpe, Pranav and Yu, Sunny and Han, Dyllan and Jurafsky, Dan , year =. Sycophantic AI decreases prosocial intentions and promotes dependence , volume =. Science , publisher =. doi:10.1126/science.aec8352 , number =

    work page doi:10.1126/science.aec8352

  21. [21]

    1981 , publisher =

    Work Stress and Social Support , author =. 1981 , publisher =

    work page 1981

  22. [22]

    2025 , eprint=

    LLMs Encode Harmfulness and Refusal Separately , author=. 2025 , eprint=

    work page 2025

  23. [23]

    2024 , eprint=

    Social Support Detection from Social Media Texts , author=. 2024 , eprint=

    work page 2024

  24. [24]

    2026 , eprint=

    The Social Sycophancy Scale: A psychometrically validated measure of sycophancy , author=. 2026 , eprint=

    work page 2026

  25. [25]

    Mark my words!: linguistic style accommodation in social media , url =

    Danescu-Niculescu-Mizil, Cristian and Gamon, Michael and Dumais, Susan , year =. Mark my words!: linguistic style accommodation in social media , url =. doi:10.1145/1963405.1963509 , booktitle =

    work page doi:10.1145/1963405.1963509

  26. [26]

    and Garrod, Simon , year =

    Pickering, Martin J. and Garrod, Simon , year =. Toward a mechanistic psychology of dialogue , volume =. Behavioral and Brain Sciences , publisher =. doi:10.1017/s0140525x04000056 , number =

    work page doi:10.1017/s0140525x04000056

  27. [27]

    Empathic Conversational Agent Platform Designs and Their Evaluation in the Context of Mental Health: Systematic Review , volume =

    Sanjeewa, Ruvini and Iyer, Ravi and Apputhurai, Pragalathan and Wickramasinghe, Nilmini and Meyer, Denny , year =. Empathic Conversational Agent Platform Designs and Their Evaluation in the Context of Mental Health: Systematic Review , volume =. doi:10.2196/58974 , journal =

    work page doi:10.2196/58974

  28. [28]

    A scoping review of empathy recognition in text using natural language processing , volume =

    Shetty, Vishal Anand and Durbin, Shauna and Weyrich, Meghan S and Martínez, Airín Denise and Qian, Jing and Chin, David L , year =. A scoping review of empathy recognition in text using natural language processing , volume =. Journal of the American Medical Informatics Association , publisher =. doi:10.1093/jamia/ocad229 , number =

    work page doi:10.1093/jamia/ocad229

  29. [29]

    , year =

    Parlamis, Jennifer D. , year =. Venting as emotion regulation: The influence of venting responses and respondent identity on anger and emotional tone , volume =. International Journal of Conflict Management , publisher =. doi:10.1108/10444061211199322 , number =

    work page doi:10.1108/10444061211199322

  30. [30]

    Craig , year =

    Zaki, Jamil and Williams, W. Craig , year =. Interpersonal emotion regulation. , volume =. Emotion , publisher =. doi:10.1037/a0033839 , number =

    work page doi:10.1037/a0033839

  31. [31]

    1984 , publisher=

    Stress, appraisal, and coping , author=. 1984 , publisher=

    work page 1984

  32. [32]

    and Miner, Adam S

    Sharma, Ashish and Lin, Inna W. and Miner, Adam S. and Atkins, David C. and Althoff, Tim , year =. Human–AI collaboration enables more empathic conversations in text-based peer-to-peer mental health support , volume =. Nature Machine Intelligence , publisher =. doi:10.1038/s42256-022-00593-2 , number =

    work page doi:10.1038/s42256-022-00593-2

  33. [33]

    Co–Rumination in the Friendships of Girls and Boys , volume =

    Rose, Amanda J , year =. Co–Rumination in the Friendships of Girls and Boys , volume =. Child Development , publisher =. doi:10.1111/1467-8624.00509 , number =

    work page doi:10.1111/1467-8624.00509

  34. [34]

    The ANNALS of the American Academy of Political and Social Science , volume=

    Data-driven content analysis of social media: A systematic overview of automated methods , author=. The ANNALS of the American Academy of Political and Social Science , volume=. 2015 , publisher=

    work page 2015

  35. [35]

    A Prompt Engineering Framework for Large Language Model–Based Mental Health Chatbots: Conceptual Framework , volume =

    Boit, Sorio and Patil, Rajvardhan , year =. A Prompt Engineering Framework for Large Language Model–Based Mental Health Chatbots: Conceptual Framework , volume =. doi:10.2196/75078 , journal =

    work page doi:10.2196/75078

  36. [36]

    Hu, Meilan and Chua, Xavier Cheng Wee and Diong, Shu Fen and Kasturiratna, K. T. A. Sandeeshwara and Majeed, Nadyanna M. and Hartanto, Andree , year =. AI as your ally: The effects of AI‐assisted venting on negative affect and perceived social support , volume =. Applied Psychology: Health and Well-Being , publisher =. doi:10.1111/aphw.12621 , number =

    work page doi:10.1111/aphw.12621

  37. [37]

    2025 , eprint=

    Drift No More? Context Equilibria in Multi-Turn LLM Interactions , author=. 2025 , eprint=

    work page 2025

  38. [38]

    Austin, TX: University of Texas at Austin , volume=

    The development and psychometric properties of LIWC-22 , author=. Austin, TX: University of Texas at Austin , volume=

    work page

  39. [39]

    2019 , publisher=

    Multivariate Data Analysis , author=. 2019 , publisher=

    work page 2019

  40. [40]

    2025 , month =

    Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions , institution =. 2025 , month =

    work page 2025

  41. [41]

    2025 , institution=

    Me, Myself & AI: What UK children's use of chatbots tells us , author=. 2025 , institution=

    work page 2025

  42. [42]

    V Ganesan, Adithya and Varadarajan, Vasudha and Mittal, Juhi and Subrahmanya, Shashanka and Matero, Matthew and Soni, Nikita and Guntuku, Sharath Chandra and Eichstaedt, Johannes and Schwartz, H. Andrew. WWBP - SQT -lite: Multi-level Models and Difference Embeddings for Moments of Change Identification in Mental Health Forums. Proceedings of the Eighth Wo...

    work page doi:10.18653/v1/2022.clpsych-1.25 2022

  43. [43]

    It’s Not Only Attention We Need

    Bucher, Andreas and Egger, Sarah and Vashkite, Inna and Wu, Wenyuan and Schwabe, Gerhard , year =. “It’s Not Only Attention We Need”: Systematic Review of Large Language Models in Mental Health Care , volume =. doi:10.2196/78410 , journal =

    work page doi:10.2196/78410

  44. [44]

    npj Digital Medicine , volume=

    When helpfulness backfires: LLMs and the risk of false medical information due to sycophantic behavior , author=. npj Digital Medicine , volume=. 2025 , publisher=

    work page 2025

  45. [45]

    Andrew and Giorgi, Salvatore and Sap, Maarten and Crutchley, Patrick and Ungar, Lyle and Eichstaedt, Johannes

    Schwartz, H. Andrew and Giorgi, Salvatore and Sap, Maarten and Crutchley, Patrick and Ungar, Lyle and Eichstaedt, Johannes. DLATK : Differential Language Analysis T ool K it. Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing: System Demonstrations. 2017. doi:10.18653/v1/D17-2010

    work page doi:10.18653/v1/d17-2010 2017

  46. [46]

    PloS one , volume=

    Personality, gender, and age in the language of social media: The open-vocabulary approach , author=. PloS one , volume=. 2013 , publisher=

    work page 2013

  47. [47]

    Proceedings of the National Academy of Sciences , volume=

    Facebook language predicts depression in medical records , author=. Proceedings of the National Academy of Sciences , volume=. 2018 , publisher=

    work page 2018

  48. [48]

    Proceedings of the international AAAI conference on web and social media , volume=

    Understanding and measuring psychological stress using social media , author=. Proceedings of the international AAAI conference on web and social media , volume=

    work page

  49. [49]

    BMJ open , volume=

    Studying expressions of loneliness in individuals using twitter: an observational study , author=. BMJ open , volume=. 2019 , publisher=

    work page 2019

  50. [50]

    Andrew and Eichstaedt, Johannes and Kern, Margaret L

    Schwartz, H. Andrew and Eichstaedt, Johannes and Kern, Margaret L. and Park, Gregory and Sap, Maarten and Stillwell, David and Kosinski, Michal and Ungar, Lyle. Towards Assessing Changes in Degree of Depression through F acebook. Proceedings of the Workshop on Computational Linguistics and Clinical Psychology: From Linguistic Signal to Clinical Reality. 2...

    work page doi:10.3115/v1/w14-3214 2014

  51. [51]

    Andrew and Park, Gregory and Eichstaedt, Johannes and Kern, Margaret and Ungar, Lyle and Shulman, Elisabeth

    Preo t iuc-Pietro, Daniel and Schwartz, H. Andrew and Park, Gregory and Eichstaedt, Johannes and Kern, Margaret and Ungar, Lyle and Shulman, Elisabeth. Modelling Valence and Arousal in F acebook posts. Proceedings of the 7th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis. 2016. doi:10.18653/v1/W16-0404

    work page doi:10.18653/v1/w16-0404 2016

  52. [52]

    NPJ Digital Medicine , volume=

    Robust language-based mental health assessments in time and space through social media , author=. NPJ Digital Medicine , volume=. 2024 , publisher=

    work page 2024

  53. [53]

    and Luft, Benjamin and Ruggero, Camilo and Ryant, Neville and Kotov, Roman and Schwartz, H

    Rao, Rajath and V Ganesan, Adithya and Kjell, Oscar and Luby, Jonah and Raghavan, Akshay and Feltman, Scott and Ringwald, Whitney and Boyd, Ryan L. and Luft, Benjamin and Ruggero, Camilo and Ryant, Neville and Kotov, Roman and Schwartz, H. Andrew. W hi SPA : Semantically and Psychologically Aligned Whisper with Self-Supervised Contrastive and Student-Teac...

    work page doi:10.18653/v1/2025.acl-long.1098 2025

  54. [54]

    Journal of Machine Learning Research , volume=

    Latent dirichlet allocation , author=. Journal of Machine Learning Research , volume=. 2003 , publisher=

    work page 2003

  55. [55]

    Richard and Koch, Gary G

    Landis, J. Richard and Koch, Gary G. , title =. Biometrics , year =

    work page

  56. [56]

    , title =

    McHugh, Mary L. , title =. Biochemia Medica , year =

    work page

  57. [57]

    Computational Linguistics , year =

    Artstein, Ron and Poesio, Massimo , title =. Computational Linguistics , year =

    work page

  58. [58]

    , title =

    Hallgren, Kevin A. , title =. Tutorials in Quantitative Methods for Psychology , year =

    work page

  59. [59]

    Comparing Large Language Models and Human Annotators in Latent Content Analysis of Sentiment, Political Leaning, Emotional Intensity and Sarcasm , journal =

    Boji\'. Comparing Large Language Models and Human Annotators in Latent Content Analysis of Sentiment, Political Leaning, Emotional Intensity and Sarcasm , journal =. 2025 , volume =

    work page 2025

  60. [60]

    , title =

    Russell, James A. , title =. Journal of Personality and Social Psychology , year =

    work page

  61. [61]

    and Srivastava, Sanjay , title =

    John, Oliver P. and Srivastava, Sanjay , title =. Handbook of Personality: Theory and Research , edition =

    work page

  62. [62]

    A Computational Framework for Behavioral Assessment of LLM Therapists , publisher =

    Chiu, Yu Ying and Sharma, Ashish and Lin, Inna Wanyin and Althoff, Tim , keywords =. A Computational Framework for Behavioral Assessment of LLM Therapists , publisher =. 2024 , copyright =. doi:10.48550/ARXIV.2401.00820 , url =

    work page doi:10.48550/arxiv.2401.00820 2024

  63. [63]

    Fitzpatrick, Kathleen Kara and Darcy, Alison and Vierhile, Molly , year =. Delivering Cognitive Behavior Therapy to Young Adults With Symptoms of Depression and Anxiety Using a Fully Automated Conversational Agent (Woebot): A Randomized Controlled Trial , volume =. JMIR Mental Health , publisher =. doi:10.2196/mental.7785 , number =

    work page doi:10.2196/mental.7785

  64. [64]

    Mental Health Discourse on reddit: Self-Disclosure, Social Support, and Anonymity , volume =

    De Choudhury, Munmun and De, Sushovan , year =. Mental Health Discourse on reddit: Self-Disclosure, Social Support, and Anonymity , volume =. Proceedings of the International AAAI Conference on Web and Social Media , publisher =. doi:10.1609/icwsm.v8i1.14526 , number =

    work page doi:10.1609/icwsm.v8i1.14526

  65. [65]

    Methods in predictive techniques for mental health status on social media: a critical review , volume =

    Chancellor, Stevie and De Choudhury, Munmun , year =. Methods in predictive techniques for mental health status on social media: a critical review , volume =. npj Digital Medicine , publisher =. doi:10.1038/s41746-020-0233-7 , number =

    work page doi:10.1038/s41746-020-0233-7

  66. [66]

    and Pennebaker, James W

    Ireland, Molly E. and Pennebaker, James W. , year =. Language style matching in writing: Synchrony in essays, correspondence, and poetry. , volume =. Journal of Personality and Social Psychology , publisher =. doi:10.1037/a0020386 , number =

    work page doi:10.1037/a0020386

  67. [67]

    2026 , institution =

    work page 2026

  68. [68]

    Real-time segmentation of on-line handwritten arabic script

    George Kour and Raid Saabne. Real-time segmentation of on-line handwritten arabic script. In Frontiers in Handwriting Recognition (ICFHR), 2014 14th International Conference on , pages 417--422. IEEE, 2014

    work page 2014

  69. [69]

    Fast classification of handwritten on-line arabic characters

    George Kour and Raid Saabne. Fast classification of handwritten on-line arabic characters. In Soft Computing and Pattern Recognition (SoCPaR), 2014 6th International Conference of , pages 312--318. IEEE, 2014

    work page 2014

  70. [70]

    Estimate and Replace: A Novel Approach to Integrating Deep Neural Networks with Existing Applications

    Guy Hadash, Einat Kermany, Boaz Carmeli, Ofer Lavi, George Kour, and Alon Jacovi. Estimate and replace: A novel approach to integrating deep neural networks with existing applications. arXiv preprint arXiv:1804.09028 , 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018