The effects of responsiveness, perceived warmth, and anthropomorphism on university students' use of conversational AI for learning support: a chain mediation analysis based on S-O-R framework
Open accessThe DOI matches its Crossref record; 128 of the 128 references it cites were found in a registry or an index; 2 of the 73 sources its claims cite were checked for support.
Abstract
Introduction: Conversational artificial intelligence (C-AI) is increasingly used by university students for learning support, yet the mechanisms through which its affective attributes shape adoption behaviors remain insufficiently understood. Drawing on the Stimulus-Organism-Response (S-O-R) framework, this study examines how AI responsiveness, anthropomorphism, and perceived warmth influence students' adoption of C-AI through AI attachment and AI trust. Methods: A cross-sectional survey was conducted among 538 Chinese university students. The proposed model tested the relationships among AI responsiveness, anthropomorphism, perceived warmth, AI attachment, AI trust, and adoption-related learning behaviors. Results: The results showed that AI responsiveness and anthropomorphism significantly strengthened students' AI attachment and AI trust, which in turn promoted their adoption of C-AI for learning support. Perceived warmth also facilitated sustained interaction and learning engagement through attachment and trust. Overall, AI attachment and AI trust served as key mediating mechanisms linking affective AI attributes to students' learning behaviors. Discussion: The findings suggest that university students' adoption of C-AI is shaped not only by technological functionality but also by emotional and relational cues embedded in AI interaction. This study extends the S-O-R framework in the context of educational AI and offers practical implications for designing human-centered, emotionally responsive, and pedagogically effective intelligent learning systems.
Overview
CoChat Summary
TL;DR
In a cross-sectional survey of 538 Chinese university students, conversational AI responsiveness, anthropomorphism, and perceived warmth predicted AI attachment and AI trust, which in turn predicted AI use for learning support. Attachment and trust mediated these relationships, including a modest serial mediation pathway, within a Stimulus-Organism-Response framework.
Key findings
- AI anthropomorphism had the strongest effect on AI attachment (β = 0.507, p < 0.001) and also predicted AI trust (β = 0.194, p < 0.001); AI responsiveness predicted attachment (β = 0.132, p < 0.01) and trust (β = 0.280, p < 0.001); perceived warmth predicted attachment (β = 0.322, p < 0.001) and trust (β = 0.217, p < 0.001).
- AI attachment predicted AI trust (β = 0.297, p < 0.001) and AI usage behavior (β = 0.460, p < 0.001); AI trust predicted AI usage behavior (β = 0.314, p < 0.001). The paper reports attachment’s association with usage as stronger than trust’s.
- AI attachment and AI trust mediated the links from responsiveness, anthropomorphism, and perceived warmth to usage behavior. Serial mediation through attachment → trust was significant for anthropomorphism (β = 0.047, p < 0.001), responsiveness (β = 0.012, p < 0.01), and perceived warmth (β = 0.030, p < 0.001), though described as modest.
- Model statistics reported: R² = 0.54–0.726; Q² = 0.301–0.543; SRMR = 0.054 saturated / 0.062 estimated; NFI = 0.839 saturated / 0.833 estimated.
Methods
- Cross-sectional online survey of Chinese university students; 565 questionnaires distributed, 27 invalid responses excluded, final N = 538 (95.2% response rate). Recruitment used non-probability convenience sampling via Wenjuanxing and WeChat; data collected November–December 2025. Sample was 40.5% male and 59.5% female; third-year students were the largest group (44.1%).
- Measures were adapted from prior validated scales and used five-point Likert items. Cronbach’s α ranged from 0.813 to 0.925; lowest AVE was 0.671; highest HTMT was 0.893; all inner-model VIFs were below 3.3.
- Analysis used PLS-SEM in SmartPLS 4.1 with 5,000 bootstrap resamples, plus SPSS for descriptive analysis. The model framed responsiveness, anthropomorphism, and perceived warmth as stimuli; attachment and trust as organism states; and AI usage behavior as the response.
Limitations
- Stated: cross-sectional SEM without qualitative data; sample drawn mainly from Fujian Province, China, limiting generalizability; omitted variables such as AI literacy, mindfulness, loneliness, and personality traits; no longitudinal tracking of attachment, trust, or usage over time.
- Stated/evident: the serial mediation effects were modest, and NFI = 0.839 is below the conservative 0.90 baseline, though the paper argues it is acceptable for PLS-SEM.
- Evident: the study relies on self-report, non-probability convenience sampling, and assesses common-method variance only through VIF values reported in the text.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures
Verification
Claims the paper makes
2 Valid · 0 Not valid · 59 Not checked · 105 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Likewise, the Theory of Planned Behavior (TPB), the Expectation-Confirmation Model of Information Systems Continuance (ECM-IS), and Social Cognitive Theory (SCT) remain centered on performance beliefs, expectations, and social norms, offering comparatively limited insight into the emotional fluctuations and relational experiences emerging during technology use (Ajzen, 1991; Bhattacherjee, 2001).”
§ The effects of responsiveness, perceived warmth, and anthropomorphism on university students' use of conversational AI for learning support: a chain mediation analysis based on S-O-R framework > Theoretical framework and research hypothesis > Stimulus-Organism-Response model (S-O-R)
What the sources say
- Partly supportedcites Ajzen, 1991 sourceExperimental — accuracy not yet measured
Ajzen (1991) supports the descriptive core: the Conclusions state that 'the theory of planned behavior traces attitudes, subjective norms, and perceived behavioral control to an underlying foundation of beliefs about the behavior,' with the relations described 'in terms of expectancy-value models' (also Abstract, p.1; Normative Beliefs and Subjective Norms; Control Beliefs and Perceived Behavioral Control) — i.e. centered on beliefs/expectancies and normative pressure from salient referents. What the source does not support is the second half of the attribution: our copy nowhere addresses technology use, 'emotional fluctuations,' or 'relational experiences during technology use,' and it in fact devotes a section ('Affect versus Evaluation') to affective beliefs and affective attitude components, noting only that the earlier theory of reasoned action drew no clear distinction between affective and evaluative responses. So the 'comparatively limited insight into emotion and relationships in technology use' is the citing authors' extension (and their comparison with ECM-IS/SCT), not a claim Ajzen makes; the 'performance beliefs' phrasing also narrows Ajzen's broader behavioral/instrumental beliefs. Searches run: 'affect versus evaluation attitude emotional component' and 'behavioral beliefs outcome evaluations subjective norm normative beliefs expectancy value'; no passage on technology continuance was found. (LLM, conf=0.72)
- Not checked — no full text of the sourcecites Bhattacherjee, 2001 source
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Prior to data collection, an a priori statistical power analysis was conducted using G*Power (Faul et al., 2009) to determine the minimum required sample size for regression-based path analysis. Following Cohen (1988) guidelines, a medium effect size (f2 = 0.15), a Type I error rate of 0.05, a statistical power of 0.80, and five predictor variables were specified. The analysis indicated that at le”
§ The effects of responsiveness, perceived warmth, and anthropomorphism on university students' use of conversational AI for learning support: a chain mediation analysis based on S-O-R framework > Materials and methods > Participants
What the sources say
- Supportedcites Faul et al., 2009 source
The citation is attached to G*Power itself (an identification/tool citation), and our copy is exactly that reference: the Abstract's parameter list defines a priori analysis as computing 'the necessary sample size ... as a function of user-specified values for the required significance level, the desired statistical power, and the to-be-detected population effect size,' and Section 4.1 ('Linear multiple regression: Fixed model, R2 deviation from zero') plus its Effect size measure section specify Cohen's f2 with input parameters 'Effect size f2,' 'err prob,' and 'Total number of predictors' m, with f2 values of .02, .15, .35 labelled small, medium, large. The specific inputs (f2 = 0.15, alpha = .05, power = .80, five predictors) and the resulting N >= 92 are the citing authors' own run of the program (and the medium-effect convention is separately credited to Cohen 1988), so nothing beyond the tool and its a priori regression procedure is attributed to Faul et al. (2009). (LLM, conf=0.88)
- Not matchedcites Cohen (1988)
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
hedgedn = 92- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The results showed that AI responsiveness and anthropomorphism significantly strengthened students' AI attachment and AI trust, which in turn promoted their adoption of C-AI for learning support.”
§ The effects of responsiveness, perceived warmth, and anthropomorphism on university students' use of conversational AI for learning support: a chain mediation analysis based on S-O-R framework > Abstract > Results
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Perceived warmth also facilitated sustained interaction and learning engagement through attachment and trust.”
§ The effects of responsiveness, perceived warmth, and anthropomorphism on university students' use of conversational AI for learning support: a chain mediation analysis based on S-O-R framework > Abstract > Results
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The results showed that AI responsiveness and anthropomorphism significantly strengthened students' AI attachment and AI trust, which in turn promoted their adoption of C-AI for learning support.”
§ The effects of responsiveness, perceived warmth, and anthropomorphism on university students' use of conversational AI for learning support: a chain mediation analysis based on S-O-R framework > Abstract > Results
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
56 more claims
Create a free account to see all 166 claims and to verify the 63 references still open.
- It may state two separate claims as one. 59
- Probably not a substantial claim. 76
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.
Rated major: it would touch a result — a value a result rests on, reported and recomputed p-values on opposite sides of the significance level, two passages that conflict with no sign which is right, or a cited work retracted before this paper was published.
In the paper
“| AIPW-AIUB | 0.096 | 0.021 | 0.188 | 0.304 | 8.389 | 0 | < 0.001 | Validated |”
§ The effects of responsiveness, perceived warmth, and anthropomorphism on university students' use of conversational AI for learning support: a chain mediation analysis based on S-O-R framework > Data analysis > Hypothesis testing
Reported 0.096 · recomputed 0.246
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ The effects of responsiveness, perceived warmth, and anthropomorphism on university students' use of conversational AI for learning support: a chain mediation analysis based on S-O-R framework > Theoretical framework and research hypothesis > Research hypothesis
“- H7: AI attachment is positively associated with university students' AI trust toward conversational AI. - H8: AI attachment is positively associated with university students' AI usage behavior related to conversational AI-assisted learning.”
§ The effects of responsiveness, perceived warmth, and anthropomorphism on university students' use of conversational AI for learning support: a chain mediation analysis based on S-O-R framework > Discussion > Affective-cognitive pathways to AI usage behavior: the distinct roles of AI attachment and AI trust
“AI attachment was positively associated with university students' AI usage behavior, providing support for Hypothesis H7.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ The effects of responsiveness, perceived warmth, and anthropomorphism on university students' use of conversational AI for learning support: a chain mediation analysis based on S-O-R framework > Materials and methods > Participants
“The collected data were imported into SPSS 26 for descriptive and demographic analysis.”
§ The effects of responsiveness, perceived warmth, and anthropomorphism on university students' use of conversational AI for learning support: a chain mediation analysis based on S-O-R framework > Data analysis
“Descriptive statistical analysis was conducted using SPSS version 29.”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
“third-year students accounted for the largest proportion of the sample (44.1%), corresponding to 249 participants.”
§ The effects of responsiveness, perceived warmth, and anthropomorphism on university students' use of conversational AI for learning support: a chain mediation analysis based on S-O-R framework > Materials and methods > Participants and sampling
Reported 44.1 · recomputed 46.282528
Citations
Cites
The theory of planned behavior
Ajzen, I. (1991) · Organizational Behavior and Human Decision Processes · doi:10.1016/0749-5978(91)90020-t · Cited by 90,587
Perceived usefulness, perceived ease of use, and user acceptance of information technology
Davis, F. D. (1989) · MIS Quarterly · doi:10.2307/249008 · Cited by 68,483 · cited 1×
“…with the Technology Acceptance Model (TAM), which emphasizes perceived usefulness and perceived ease of use as determinants of usage behavior (Davis, 1989), and the Unified Theory of Acceptance and Use of Technology (UTAUT), which explains technology adoption through performance expectancy…”
User Acceptance of Information Technology: Toward A Unified View1
Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003) · MIS Quarterly · doi:10.2307/30036540 · Cited by 45,035
Statistical power analyses using G∗Power 3.1: tests for correlation and regression analyses
Faul, F., Erdfelder, E., Buchner, A., & Lang, A.-G. (2009) · Behavior Research Methods · doi:10.3758/brm.41.4.1149 · Cited by 37,615
Partial Least Squares Structural Equation Modeling (PLS-SEM) Using R: A Workbook
Hair, J. F., Hult, G. T. M., Ringle, C. M., Sarstedt, M., Danks, N. P., & Ray, S. (2021) · Classroom Companion: Business · doi:10.1007/978-3-030-80519-7 · Cited by 10,343 · cited 2×
“…(PLS-SEM) has been widely applied across multiple disciplines—including education, economics, and computer science—to assess large and complex models (Hair et al., 2021a paradigm and is specifically designed to test the predictive capability of models carefully developed based on theory and logical…”
InfluentialCommon Method Bias in PLS-SEM: A Full Collinearity Assessment Approach
Kock, N. (2015) · International Journal of e-Collaboration · doi:10.4018/ijec.2015100101 · Cited by 10,215
Understanding Information Systems Continuance: An Expectation-Confirmation Model1
Bhattacherjee, A. (2001) · MIS Quarterly · doi:10.2307/3250921 · Cited by 8,767 · cited 1×
“…Systems Continuance (ECM-IS), and Social Cognitive Theory (SCT) remain centered on performance beliefs, expectations, and social norms, offering comparatively limited insight into the emotional fluctuations and relational experiences emerging during technology use (Ajzen, 1991; Bhattacherjee, 2001).”
(1974) · Medical Entomology and Zoology · Cited by 7,295
Predictive model assessment in PLS-SEM: guidelines for using PLSpredict
Shmueli, G., Sarstedt, M., Hair, J. F., Cheah, J.-H., Ting, H., Vaithilingam, S., & Ringle, C. M. (2019) · European Journal of Marketing · doi:10.1108/ejm-02-2019-0189 · Cited by 4,618
ELIZA—a computer program for the study of natural language communication between man and machine
Weizenbaum, J. (1966) · Communications of the ACM · doi:10.1145/365153.365168 · Cited by 4,452 · cited 1×
“Since the early development of ELIZA, which relied on pattern-matching techniques to simulate dialogue (Weizenbaum, 1966), conversational AI has evolved from text-based systems into interactive agents capable of multimodal communication and contextual feedback (Dehn and Van Mulken, 2000; Wik and…”
Research works printed but not linked: 2
- Cohen, 1988, Statistical Power Analysis for the Behavioral Sciences, 2nd Edn
- Hair, A Primer on Partial Least Squares Structural Equation Modeling (PLS-SEM)
Cited by
Synthetic Attachment Displacement: A Developmental Framework for Adolescent AI Companionship
Timothy Beckman, D. (2026) · The International Journal of Social Sciences and Humanities Invention · doi:10.18535/ijsshi/v13i07.02 · Cited by 1
Perceived control and immersion in AI chatbot interaction: a psychological distance perspective
Wang, M., Li, Y., & Zhang, W. (2026) · Frontiers in Psychology · doi:10.3389/fpsyg.2026.1823399
Do large language models stimulate student motivation? A meta-analysis of experimental studies
Han, X., & Liu, M. (2026) · Educational Technology Research and Development · doi:10.1007/s11423-026-10687-y
Thakkar, S., Bhatt, P., & Parekh, H. (2026) · Counselling and Psychotherapy Research · doi:10.1002/capr.70237
AI Adoption, AI Literacy, and Employee Performance: The Mediating Role of Work Autonomy in MSMEs
Tirtana, A., Sudarmiatin, S., Hermawan, A., & Rahayu, W. P. (2026) · Social Science Academic · doi:10.37680/ssa.10955
Metadata
- Authors
- Lan, Y., Liu, S., & Chen, H.
- Affiliations
- Xiamen University of Technology (CN)
- Corresponding author
- Chen, Hao
- ORCID
- Chen, Hao
- Venue
- Frontiers in Psychology
- Publisher
- Frontiers Media SA
- Published
- 2026-04-07
- Volume
- 17
- Pages
- 1769964-1769964
- ISSN
- 1664-1078 (ISSN-L)
- PMC
- PMC13096099
- OpenAlex
- W7151617022
- Cited by
- 7
- References
- 126
- Citations by year
- 2026: 7
- FWCI
- 20.72
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Frontiers in Psychology (publishedVersion, cc-by)
- PubMed (publishedVersion)
- PubMed Central (publishedVersion, cc-by)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Topics
- Social Sciences › Business, Management and Accounting › Marketing › AI in Service Interactions
- Social Sciences › Education › Artificial Intelligence in Education
- Social Sciences › Psychology › Social Psychology › Social Robot Interaction and HRI
- Keywords
- conversational AI, perceived warmth, anthropomorphism, trust in AI, technology adoption, learning support, university students, stimulus-organism-response model, serial mediation, human-centered AI design
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 2026-11-18, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-11-18 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (qwen3.8-omni-flash).
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
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.

