Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record, with a small difference in title or year; 59 of the 63 references it cites were found in a registry or an index; 8 of the 30 sources its claims cite were checked for support.

World Trade Center responders in their own words: predicting PTSD symptom trajectories with AI-based language analyses of interviews

Open access

Abstract

Abstract Background Oral histories from 9/11 responders to the World Trade Center (WTC) attacks provide rich narratives about distress and resilience. Artificial Intelligence (AI) models promise to detect psychopathology in natural language, but they have been evaluated primarily in non-clinical settings using social media. This study sought to test the ability of AI-based language assessments to predict PTSD symptom trajectories among responders. Methods Participants were 124 responders whose health was monitored at the Stony Brook WTC Health and Wellness Program who completed oral history interviews about their initial WTC experiences. PTSD symptom severity was measured longitudinally using the PTSD Checklist (PCL) for up to 7 years post-interview. AI-based indicators were computed for depression, anxiety, neuroticism, and extraversion along with dictionary-based measures of linguistic and interpersonal style. Linear regression and multilevel models estimated associations of AI indicators with concurrent and subsequent PTSD symptom severity (significance adjusted by false discovery rate). Results Cross-sectionally, greater depressive language ( β = 0.32; p = 0.049) and first-person singular usage ( β = 0.31; p = 0.049) were associated with increased symptom severity. Longitudinally, anxious language predicted future worsening in PCL scores ( β = 0.30; p = 0.049), whereas first-person plural usage ( β = −0.36; p = 0.014) and longer words usage ( β = −0.35; p = 0.014) predicted improvement. Conclusions This is the first study to demonstrate the value of AI in understanding PTSD in a vulnerable population. Future studies should extend this application to other trauma exposures and to other demographic groups, especially under-represented minorities.

Overview

CoChat Summary

TL;DR

In 124 World Trade Center responders’ oral history interviews, AI-based language features were associated with concurrent PTSD severity and, in a subset, predicted future PTSD symptom trajectories. Depressive/self-focused language tracked higher symptoms, while more anxious language predicted worsening and more plural pronouns/longer words predicted improvement.

Key findings

  • In cross-sectional analyses, greater depressive language and first-person singular use were each associated with higher PTSD symptom severity: β = 0.32 and β = 0.31, both p = 0.049.
  • Longitudinally, anxious language predicted future worsening in PTSD symptoms: β = 0.30, p = 0.049.
  • First-person plural pronoun use and longer word usage predicted improvement in PTSD symptoms: β = −0.36 and β = −0.35, both p = 0.014.
  • The paper reports that higher PCL scores were also associated with anxious, neurotic, and less extraverted language patterns, and with greater word count; results remained after adjustment for age, gender, occupation, and years after 9/11.

Methods

  • Observational study of 124 WTC responders from the Stony Brook WTC Health & Wellness Program who completed ~1-hour oral history interviews; analysis sample was restricted to 113 responders with at least one valid PTSD Checklist (PCL) score within 2 years of interview and at least one pre-interview PCL, and a longitudinal subset of 75 had at least three post-interview assessments at least 2 years later.
  • Audio interviews were transcribed and converted into language features; nine predictors were derived: AI-based depression, anxiety, neuroticism, extraversion; LIWC-style first-person singular, first-person plural, and article use; plus word count and average word length.
  • Associations with concurrent PTSD severity and future symptom trajectories were estimated using linear regression and multilevel/trajectory models, adjusting for age, gender, occupation, and years since 9/11; significance used false discovery rate adjustment.

Limitations

  • The authors state the sample was a specific WTC responder cohort that was predominantly male and largely police, which may limit generalizability to other occupations and demographic groups.
  • They state the mechanisms behind the predictive effects were not tested directly.
  • They state the sample was relatively small and the pre-trained AI models could not be retrained for this population; this likely limits model tailoring and predictive power.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1017/s0033291721002294

Cites

59 of 59 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1. Evaluation setup for trajectory prediction. According to equation 3, we can then model the control-adjusted trajectory per user as B1−cntrl, i = (α0 + α1x1i + α2x2i + … + α5x5i). Then, we used the slope of the fitted line as the PCL trajectory of the corresponding subject. Our main outcome was correlations between this trajectory slope and the subject's language patterns. The figure illustrates our trajectory modeling; dots in the figure represent the PTSD scores at the health assessments after the oral history interview of a responder and the red line represents the PTSD future trajectory line which is correlated with his/her language assessment from the interview.
Fig. 1. Evaluation setup for trajectory prediction. According to equation 3, we can then model the control-adjusted trajectory per user as B1−cntrl, i = (α0 + α1x1i + α2x2i + … + α5x5i). Then, we used the slope of the fitted line as the PCL trajectory of the corresponding subject. Our main outcome was correlations between this trajectory slope and the subject's language patterns. The figure illustrates our trajectory modeling; dots in the figure represent the PTSD scores at the health assessments after the oral history interview of a responder and the red line represents the PTSD future trajectory line which is correlated with his/her language assessment from the interview.CC BY · PubMed Central
Fig. 2. Average future PCL score trajectories of top (blue) and bottom (red) terciles of responders based on language-based assessments: word usages of first-person plurals (left), anxious language patterns (right), and average word lengths (bottom). All trajectories have been adjusted for interview (baseline) PCL scores, representing the residual after accounting for the expected trajectory at baseline. All differences are significant at p < 0.05 (see online Supplementary Table S1 for further analysis).
Fig. 2. Average future PCL score trajectories of top (blue) and bottom (red) terciles of responders based on language-based assessments: word usages of first-person plurals (left), anxious language patterns (right), and average word lengths (bottom). All trajectories have been adjusted for interview (baseline) PCL scores, representing the residual after accounting for the expected trajectory at baseline. All differences are significant at p < 0.05 (see online Supplementary Table S1 for further analysis).CC BY · PubMed Central

Verification

Claims the paper makes

7 Valid · 1 Not valid · 8 Not checked · 36 Own results and interpretations

27 more claims

Create a free account to see all 52 claims and every source verdict.

See all claims — free account
  • Probably not a substantial claim. 15
  • It may state two separate claims as one. 9

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Citations

Cites

Show all 63 references

Research works printed but not linked: 1

  1. 10.18653/v1/W19-3003 · doi:10.18653/v1/w19-3003

Cited by

See all 42

Metadata

Authors
Son, Y., Clouston, S. A. P., Kotov, R., Eichstaedt, J. C., Bromet, E. J., Luft, B. J., & Schwartz, H. A.
Affiliations
  • Stony Brook University (US)
  • Stanford University (US)
Corresponding author
Son, Youngseo
Venue
Psychological Medicine
Published
2021-06-22
Volume
53
Issue
3
Pages
918-926
ISSN
  • 0033-2917 (ISSN-L)
  • 1469-8978
OpenAlex
W3174101453
Cited by
42
References
63
Citations by year
2026: 9 · 2025: 10 · 2024: 8 · 2023: 9 · 2022: 4 · 2021: 2
FWCI
5.63
Type
journal article
Language
en
Open access
Locations
  • Psychological Medicine (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • arXiv (Cornell University)
Funders
  • U.S. Department of Defense
  • Stony Brook University
  • National Institutes of Health (NIH-R01)
  • Defense Advanced Research Projects Agency (W911NF-20-1-0306)
  • National Institute on Alcohol Abuse and Alcoholism (R01AA028032)
  • National Institute on Aging (R01AG049953)
  • National Institute for Occupational Safety and Health (U01OH012476)
Topics
  • Social Sciences › Psychology › Social Psychology › Mental Health via Writing
  • Social Sciences › Psychology › Clinical Psychology › Posttraumatic Stress Disorder Research
  • Social Sciences › Psychology › Clinical Psychology › COVID-19 and Mental Health
Keywords
World Trade Center responders, September 11 attacks, post-traumatic stress disorder, PTSD symptom trajectories, natural language processing, depression, anxiety, language style, interpersonal style, resilience
MeSH terms
Artificial Intelligence, Humans, Linguistics, Stress Disorders, Post-Traumatic, September 11 Terrorist Attacks, Emergency Responders

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-11, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-11 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: World Trade Center responders in their own words: predicting PTSD symptom trajectories with AI-based language analyses of interviews. https://app.cochat.ai/papers/10.1017/s0033291721002294
The paper
Son, Y., Clouston, S. A. P., Kotov, R., Eichstaedt, J. C., Bromet, E. J., Luft, B. J., & Schwartz, H. A. (2023). World Trade Center responders in their own words: predicting PTSD symptom trajectories with AI-based language analyses of interviews. Psychological Medicine. https://doi.org/10.1017/s0033291721002294

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.