Skip to content
DraconiticResearch
Search Sources API
203,952 studies31 sourcesUpdated 16 min ago
PreprintCC BY 4.0Zenodo

Agreement between large language models and human coders on psychological and clinical text: a systematic review and multilevel meta-analysis

  • Martin Rekhert1
  1. 1Turan University
Open at sourcePDFDOICite

Abstract

Objective: To synthesize agreement between generative large language models (LLMs) and human coders on psychological and clinical text across four tasks, and benchmark it against human–human agreement on identical texts.

Method: Preregistered on OSF (https://doi.org/10.17605/OSF.IO/Z63BS). Nine bibliographic sources were searched to October 21, 2026, plus citation searching. Eligible studies evaluated a generative LLM and human coders on identical text units using the same scheme. Correlated-and-hierarchical-effects models with cluster-robust variance estimation (CR2) pooled effect sizes.

Results: Across 377 studies (416 reports), pooled chance-corrected agreement (kappa family) was 0.62 (95% CI [0.58, 0.65]; 107 studies, 595 estimates; 95% prediction interval [0.15, 1.00]). Agreement did not differ across tasks (codebook coding 0.63, assessment scoring 0.58, clinical rating 0.62, helper evaluation 0.59). Correlation/ICC pooled at 0.71 [0.66, 0.75] (95 studies). In 23 matched studies, LLMs agreed less with humans than humans with each other (&Delta;κ = −0.11 [−0.18, −0.03]); equivalence within ±0.10 could not be established for models from 2025 (&Delta;κ = −0.12, 90% CI [−0.19, −0.05]). Agreement did not improve with release date (+0.02/year, p = .282) or within-study model contrasts (−0.01 [−0.06, 0.03]); it tracked human reliability (slope = 0.63, p < .001) and was higher for manifest than latent constructs (+0.09, p = .028). Percentage agreement exceeded kappa by 0.24 in the same configurations. Certainty (GRADE) was low to very low.

Conclusions: Generative LLMs achieve moderate-to-substantial agreement but remain, on average, below human inter-rater reliability. Agreement did not change across model release dates. Validation studies must report chance-corrected metrics and human benchmarks on identical units.

Related

  1. Using Vision and Large Language Models for Psychological Intervention Competency Assessment: A Pilot Study of Agreement with Clinician Ratings
    Y. Liu, W. Leever, M. Stanco +7
    medRxivPreprintCC BY-NC-NDFull textEpidemiology & public healthmedRxiv
  2. Psychological outcomes of primary-school physical education interventions: a systematic review and multilevel meta-analysis.
    Lebin Li, Yao Huang, Di Zhang +1
    Frontiers in psychologyPeer-reviewedCC BYFull textEurope PMC
  3. Beyond Landmark Error: Agreement Between AI-Assisted and Manual Cephalometric Tracing at the Level of Clinical Measurements: A Systematic Review
    Hassan Ahmed Assiri, Huda Alnami, Anwar Abdullah Alsaeed +3
    TomographyPeer-reviewedCC BYFull textMedical imaging & radiologyEurope PMC
  4. Vibration Therapy for Poststroke Clinical Balance: A Domain-Specific Systematic Review and Multilevel Meta-Analysis.
    Jeong-Woo Seo, Ji-Woo Seok, Jung-Dae Kim
    Healthcare (Basel, Switzerland)Peer-reviewedCC BYFull textGeneral medicineEurope PMC
  5. Large language models reveal a systematic preference in clinical immunotherapy guidelines: a multi-model comparative analysis
    Zhenyu Chen, Anqi Lin, Junyi Shen +24
    Scientific ReportsPeer-reviewedCC BY 4.0Full textEpidemiology & public healthCrossref

Cite

BibTeXRIS
BibTeX
@misc{rekhert2026agreement,
  title = {{Agreement between large language models and human coders on psychological and clinical text: a systematic review and multilevel meta-analysis}},
  author = {Martin Rekhert},
  howpublished = {Zenodo},
  year = {2026},
  month = oct,
  doi = {10.5281/zenodo.23613737},
  url = {https://zenodo.org/records/23613737},
  note = {CC BY 4.0},
}
RIS
TY  - UNPB
TI  - Agreement between large language models and human coders on psychological and clinical text: a systematic review and multilevel meta-analysis
AU  - Martin Rekhert
PY  - 2026
DA  - 2026/10/24
PB  - Zenodo
DO  - 10.5281/zenodo.23613737
UR  - https://zenodo.org/records/23613737
AB  - Objective: To synthesize agreement between generative large language models (LLMs) and human coders on psychological and clinical text across four tasks, and benchmark it against human–human agreement on identical texts.  Method: Preregistered on OSF (https://doi.org/10.17605/OSF.IO/Z63BS). Nine bibliographic sources were searched to October 21, 2026, plus citation searching. Eligible studies evaluated a generative LLM and human coders on identical text units using the same scheme. Correlated-and-hierarchical-effects models with cluster-robust variance estimation (CR2) pooled effect sizes.  Results: Across 377 studies (416 reports), pooled chance-corrected agreement (kappa family) was 0.62 (95% CI [0.58, 0.65]; 107 studies, 595 estimates; 95% prediction interval [0.15, 1.00]). Agreement did not differ across tasks (codebook coding 0.63, assessment scoring 0.58, clinical rating 0.62, helper evaluation 0.59). Correlation/ICC pooled at 0.71 [0.66, 0.75] (95 studies). In 23 matched studies, LLMs agreed less with humans than humans with each other (&Delta;κ = −0.11 [−0.18, −0.03]); equivalence within ±0.10 could not be established for models from 2025 (&Delta;κ = −0.12, 90% CI [−0.19, −0.05]). Agreement did not improve with release date (+0.02/year, p = .282) or within-study model contrasts (−0.01 [−0.06, 0.03]); it tracked human reliability (slope = 0.63, p < .001) and was higher for manifest than latent constructs (+0.09, p = .028). Percentage agreement exceeded kappa by 0.24 in the same configurations. Certainty (GRADE) was low to very low.  Conclusions: Generative LLMs achieve moderate-to-substantial agreement but remain, on average, below human inter-rater reliability. Agreement did not change across model release dates. Validation studies must report chance-corrected metrics and human benchmarks on identical units.
ER  -