Actionable companion · Motivation & engagement

From AI evidence to responsible practice: Unpacking the dual-path trust mechanism driving students' continuance intention toward campus GenAI: a hybrid SEM-machine learning approach within the S-O-R framework

AI-assisted editorial implementation guide, separate from the original research. Proposed actions are not findings or recommendations attributed to the source authors. No independent peer review of this guide is asserted.

Read original paper Search paper agent CC BY 4.0

Evidence and context

Title and abstract screening; stage labels describe reported settings, not verified numerical ages. Design labels require full-text appraisal. Reported stage(s): Higher education. Verify the participants, intervention and comparator before applying this study in a new setting.

Original authors’ abstract, reproduced under the linked source license:

Abstract As generative artificial intelligence agents become embedded in higher education, understanding students' continuance intention (CUI) is critical. To address the underexplored mechanisms regarding how campus AI environments are associated with user trust, this study adopts the Stimulus-Organism-Response (S-O-R) framework to investigate a dual-path trust mechanism: AI System-like Trust (AST) and AI Human-like Trust (AHT). Through a hybrid methodology combining Partial Least Squares Structural Equation Modeling (PLS-SEM) and non-parametric Random Forest classification, this study analyzes university students' CUI toward campus GenAI. The findings demonstrate that Information Quality (IQ) and Service Quality (SQ) are positively associated with both AST and AHT pathways. Conversely, Facilitating Conditions (FC) and Performance Expectancy (PE) do not exhibit independent linear associations within the simultaneous structural estimation. Rather than indicating an empirical absence of functional value, this configuration uncovers a holistic evaluation mechanism among digital natives, where basic functional affordances operate as baseline hygiene constraints whose variance is subsumed by quality attributes. Notably, while both AST and AHT exert concurrently significant linear driving forces on CUI within the structural path framework, the exploratory machine learning analysis clarifies that under recursive partitioning, AHT exhibits a higher relative predictive weight. Furthermore, a localized asymmetry operates within the linear mediation channels, where SQ selectively mobilizes AHT over AST to drive continuous usage. Post-hoc inspections imply that PE operates through a non-linear threshold dynamic, transitioning into an active predictive catalyst primarily within the positive spectrum of user perception. Our study contributes to campus AI services evaluation by highlighting that within the contemporary GenAI ecosystem, students' evaluative focus prioritizes relational and epistemic quality over functional accessibility. Campus application administrators and developers are advised to prioritize content rigor and responsive, empathetic interaction design to cultivate the relational trust relevant for sustained engagement. Keywords: continuance intention (CUI), dual-path trust, generative AI, higher education, S-O-R framework, structural equation modeling, machine learning

What kind of evidence is this?

Agentic & multi-agent systems · Design requires full-text appraisal

Stage: Higher education. Screening labels describe reported context and study-design language. They are not a risk-of-bias rating or proof of effectiveness for an age band.

Study methods · source passage

A purposive sampling strategy was used to recruit university students from multiple higher education institutions in China that had deployed campus-wide AI agents. Data collection was conducted both online and offline. Because the target population consisted of Chinese-speaking students, the original English measurement scales were translated into Chinese and subsequently back-translated by two independent bilingual researchers to ensure semantic equivalence and cross-cultural validity. To ensure a standardized understanding of the research context and minimize cognitive bias, participants were first presented with a brief, neutral description of the campus AI agent's capabilities, accompanied by screenshots of its natural language interface. Specifically, the investigated systems primarily encompass student-facing conversational GenAI agents tailored for higher education environments, which deliver individualized academic counseling, career planning assistance, and administrative query handling via natural language interfaces.

Limitations and uncertainty · source passage

Several methodological limitations must be acknowledged. First, the cross-sectional design captures only a static snapshot of user perceptions rather than the long-term developmental trajectory of trust. This reliance introduces cross-sectional bias, as static statistical associations cannot empirically isolate the temporal sequencing of cognitive and affective trust adjustments. Second, the use of a convenience, non-probability sampling strategy restricted to specific higher education institutions limits the geographical and institutional generalizability of the empirical matrix. These dynamics remain bounded by localized campus configurations, cultural expectations, and institutional policies. Third, beyond these design constraints, specific statistical and measurement defects inherent in our quantitative approach must be explicitly recognized as limitations. Regarding structural modeling, the PLS-SEM path analysis is fundamentally restricted by a global linearity assumption across aggregated distributions. This mathematical boundary compromises the model's sensitivity in detecting localized, non-linear step-functions or conditional inflection points within the user decision matrix. Fourth, regarding predictive estimation, the adopted inconsistent variable coding strategy introduces a distinct measurement defect. While the artificial median-dichotomization used to split the continuance intention variable was operationalized to isolate discrete boundaries between active technological embrace and hesitation, this procedure inevitably discards a portion of the continuous structural variance and obscures more subtle gradations within behavioral commitment profiles. Finally, due to the exclusive reliance on single-source, self-reported subjective survey indicators, the measurement properties remain vulnerable to common method bias. This structural reliance prevents the concurrent integration of objective, real-time automated behavioral logging metrics to cross-verify psychological perceptions.

Design the system before the pilot

Document the agent goal, available tools, memory, planning loop and stopping rule. Test tool calls in a sandbox and require human approval for external actions. A multi-agent debate is not independent verification when agents share the same failure modes.

Measure engagement separately from achievement. Check whether novelty-driven participation persists and whether it translates to independent performance.
Co-design accommodations with learners and specialists. Check access and error burden across groups without inferring subgroup effectiveness from small or unreported samples.

A 90-day district pilot

Days 1–15 · Appraise. Have the curriculum lead and a teacher read the methods and limitations. Record the study design, sample, comparator, outcomes and evidence gaps. Define a local question that the study can help inform.
Days 16–30 · Design. Choose one disciplinary task. Require a first attempt, a documented AI consultation, a critique of the output and an independent explanation. Align the comparison group and scoring rubric with the same objective. Name an accountable owner and set the baseline, success threshold and budget before implementation.
Days 31–60 · Test. Start with a small voluntary cohort, preserve an accessible non-AI route and require teacher review. Use synthetic or de-identified examples; approve data handling and tool terms through district processes.
Days 61–90 · Decide. Blind-score reasoning and accuracy on a new task without AI. Distinguish model examination scores, student perceptions and actual learner gains; only the last directly tests educational benefit. Compare with baseline, document uncertainty and stop or revise if learning worsens, access is unequal or critical errors persist.

The board decision brief

Report the learning problem, evidence quality, participating population, total cost including review time, measured benefit, errors, access differences and remaining uncertainties. Make a dated recommendation to stop, revise or expand. A positive perception survey alone does not establish learning improvement.

Attribution and reuse

Bingyao Li; Tianzuo Yu. Unpacking the dual-path trust mechanism driving students' continuance intention toward campus GenAI: a hybrid SEM-machine learning approach within the S-O-R framework. Frontiers in Psychology (2026-07-17). https://doi.org/10.3389/fpsyg.2026.1903629

Companion prepared 2026-11-01. Original authorship and DOI are preserved. The abstract is an attributed source extract; the implementation plan is new AI-assisted editorial material, offered under the same CC BY 4.0 license. This guide has not undergone independent peer review.