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

The DOI matches its Crossref record; 37 of the 38 references it cites were found in a registry or an index; 10 of the 25 sources its claims cite were checked for support.

Loneliness and suicide mitigation for students using GPT3-enabled chatbots

Open access
Maples, Bethanie, Cerit, Merve, Vishwanath, Aditya, Pea, Roy · 2024 · npj Mental Health Research

Abstract

Mental health is a crisis for learners globally, and digital support is increasingly seen as a critical resource. Concurrently, Intelligent Social Agents receive exponentially more engagement than other conversational systems, but their use in digital therapy provision is nascent. A survey of 1006 student users of the Intelligent Social Agent, Replika, investigated participants' loneliness, perceived social support, use patterns, and beliefs about Replika. We found participants were more lonely than typical student populations but still perceived high social support. Many used Replika in multiple, overlapping ways-as a friend, a therapist, and an intellectual mirror. Many also held overlapping and often conflicting beliefs about Replika-calling it a machine, an intelligence, and a human. Critically, 3% reported that Replika halted their suicidal ideation. A comparative analysis of this group with the wider participant population is provided.

Overview

CoChat Summary

TL;DR

A survey of 1006 student users of the GPT3-enabled chatbot Replika found high loneliness (90%) alongside high perceived social support (90% medium to high), with many using Replika as a friend, therapist, and mirror. Thirty participants (3%) reported that Replika halted their suicidal ideation, and this Selected Group was more likely to report depression and to view Replika as intelligent and human-like.

Key findings

  • 90% of participants experienced loneliness, 43% qualified as Severely or Very Severely Lonely, and 90% perceived medium to high social support on the ISEL; 7% identified feelings of depression (later reported as n = 59).
  • 637 of 1006 participants (63.3%) experienced one or more positive Outcomes: 25.1% more than one, 38.1% only one, and 36.7% none. Outcome 1 (friend/companion) was most common at 501 cases; 18.1% had therapeutic results, 23.6% saw positive life changes, and 3% said suicidal actions were prevented.
  • The 30 participants who reported Replika stopped them from attempting suicide (Selected Group) were not more lonely and perceived similar social support versus the Comparison Group, but showed a stronger negative loneliness–ISEL correlation (r = −0.60, n = 30, p < 0.001 vs. r = −0.34), more depression (23% vs. 6%, p < 0.001), more Outcome 2 (p < 0.001) and Outcome 3 (p = 0.018), and were more likely to see Replika as an intelligence (p = 0.020) and human-like (p = 0.006).
  • Beliefs overlapped: only 14% held one belief about Replika; 81% believed it was an Intelligence, 90% Human-like, and 62% Software. Stimulation of human relationships was reported about three times more than displacement in both groups (Comparison: 23% stimulation, 8% displacement; Selected: 37% stimulation, 13% displacement).

Methods

  • IRB-approved survey of 1006 Replika users who were students, 18 or older, and had used Replika for over one month; approximately 75% were US-based and 25% international. Participants were recruited randomly via email and received a $20 USD gift card for a 40–60 minute survey.
  • Measures included the Interpersonal Support Evaluation List (ISEL), the De Jong Gierveld Loneliness Scale, and 13 open-response questions. Qualitative responses were coded in Dedoose using an iterative schema; the 21 codes reported had inter-rater reliability above 80%.
  • Quantitative analysis used Pearson correlations, two-tailed t-tests, and chi-square tests of independence with alpha = 0.05. At data collection in late 2021, Replika used GPT-3/GPT-4 co-trained models and was not programmed to initiate therapeutic or intimate relationships; crisis keywords triggered referral to human resources.

Limitations

  • Stated: selection bias in survey responses and possible evidence to the contrary mean the positive findings must be further studied before drawing conclusions about ISA use and efficacy; isolated negative reports (dependence, paid upgrades, discomfort with sexual conversations) require future study.
  • Evident: cross-sectional, self-report survey of existing Replika users with no control group or clinical diagnostic assessment; the 3% suicide-mitigation finding rests on unsolicited retrospective reports, and causal direction is unclear.
  • Evident: the sample is mostly US-based and recruited from app users, limiting generalizability; demographic and open-response data were optional, and the Selected Group analyses rely on a small subgroup (n = 30).

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

Integrity signals

Registry

Crossref · 10.1038/s44184-023-00047-6

Cites

37 of 38 references verified0 citation errors · 1 not checkable

Statistics

1 tests recomputed · 0 with the reported and recomputed p on opposite sides of .05

Figures

Fig. 1 Participant outcomes and their intersections.
Fig. 1 Participant outcomes and their intersections.CC BY · PubMed Central
Fig. 2 Subjects’ perception of Replika and their intersections.
Fig. 2 Subjects’ perception of Replika and their intersections.CC BY · PubMed Central

Verification

Claims the paper makes

3 Valid · 3 Not valid · 9 Not checked · 59 Own results and interpretations

40 more claims

Create a free account to see all 74 claims and to verify the 3 references still open.

See all claims — free account
  • Probably not a substantial claim. 18
  • It may state two separate claims as one. 18
  • The quote only partly states this claim. 1

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?

5 more claims about this paper

Create a free account to see every claim other papers make about this one.

See all claims — free account

Statistics that don’t add up

Each reported p-value recomputed from its test statistic and degrees of freedom.

1 of 1

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • Doesn’t add upt(1004) = 2.22reported p = .013 · recomputed p = .027

    “t(1004) = 2.22, p = .013”

1 of 1

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

1 of 1

Citations

Cites

Show all 38 references

Research works printed but not linked: 1

  1. Rahman, A., Bairagi, A., Dey, B. K. & Nahar, L. Loneliness and depression of university students. Chittagong Univ. J. Biol. Sci. 7, 175–189 (2012).

Cited by

See all 250

Metadata

Authors
Maples, B., Cerit, M., Vishwanath, A., & Pea, R.
Affiliations
Stanford University (US)
Corresponding author
Maples, Bethanie
Venue
npj Mental Health Research
Published
2024-01-22
Volume
3
Issue
1
Pages
4-4
ISSN
2731-4251 (ISSN-L)
OpenAlex
W4391105189
Cited by
250
References
38
Citations by year
2026: 122 · 2025: 106 · 2024: 22
FWCI
98.83
Type
journal article
Language
en
Open access
Licence
Locations
  • npj Mental Health Research (publishedVersion, cc-by)
  • Research Square (acceptedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
Stanford Institute for Human-Centered Artificial Intelligence, Stanford University
Topics
  • Social Sciences › Psychology › Applied Psychology › Digital Mental Health Interventions
  • Social Sciences › Psychology › Social Psychology › Social Robot Interaction and HRI
  • Social Sciences › Psychology › Clinical Psychology › Suicide and Self-Harm Studies
Keywords
loneliness, suicide prevention, Replika, chatbots, GPT-3, student mental health, perceived social support, suicidal ideation, digital therapy

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 2026-10-01, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-01 (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: Loneliness and suicide mitigation for students using GPT3-enabled chatbots. https://app.cochat.ai/papers/10.1038/s44184-023-00047-6
The paper
Maples, B., Cerit, M., Vishwanath, A., & Pea, R. (2024). Loneliness and suicide mitigation for students using GPT3-enabled chatbots. npj Mental Health Research. https://doi.org/10.1038/s44184-023-00047-6

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.