Determinants of rural middle school students' adoption of AI chatbots for mental health

Abstract

Adolescent mental health challenges constitute an important global public health issue. Despite the rapid development of AI technology in various fields, its adoption in rural mental health remains constrained. The purpose of this study is to examine the factors that influence the adoption of AI chatbots for mental health education among rural Chinese secondary school students. Utilizing the UTAUT2 framework, we included Perceived Risk (PR) and Perceived Anthropomorphism (PA) to construct a theoretical model. A questionnaire survey of 317 rural adolescents was conducted, analyzed via SPSS and AMOS. Results showed PE, EE, SI, and PA positively correlated with BI; PR negatively correlated; HM had no effect. Grade level moderated specific paths. The study extends UTAUT2 to marginalized populations, filling a gap in AI-driven rural adolescent mental health interventions.

1 Background

Global mental health is experiencing a significant decline. According to the World Health Organization (WHO), ~1 billion people worldwide suffer from mental disorders (1), with the prevalence continuing to rise (2, 3). This increasing burden not only imposes substantial economic costs (4), but also exacerbates issues such as high unemployment (5) and low educational achievements (6), further straining the global economy (7–9). To address this challenge, the international community has implemented measures to enhance mental health (10, 11). For instance, the 2030 Sustainable Development Goals (12) advocate policy actions to promote mental wellbeing, while WHO's Global Mental Health Action Plan (13) integrates mental health services into public health systems.

Adolescents aged 10–19 are at particularly high risk of developing mental health disorders, yet their needs remain widely under-recognized and under-treated (14). Recent evidence shows that the prevalence of clinically elevated symptoms of depression and anxiety among children and adolescents reached 25.2 and 20.5% respectively during the COVID-19 pandemic, nearly double pre-pandemic levels (15).

In low-and middle-income countries, the prevalence of adolescent mental illness may be even higher (16). In China, a middle-income country, 14.8% of adolescents face mental health challenges, predominantly depression and anxiety (17). The 2022 National Depression Blue Book (18) reports that 30% of depression patients are under 18, 50% of whom are students. The 2023 China Mental Health Blue Book (19) further indicates a 40% depression detection rate among high school students and 30% among junior high students.

Adolescence is a critical period for physical and mental development. Untreated childhood mental health issues significantly increase the risks of adult health problems, criminal behavior, academic failure, and poverty (20).

Despite China's policy initiatives—such as the Special Action Plan for Student Mental Health (21–23), adolescent mental health issues remain severe (24–26). Notably, rural adolescents exhibit poorer mental health than urban peers (27). In rural schools, 20% of students show depression risk, 68% have at least one anxiety symptom (28), and their mental illness scale scores surpass urban students (29). This disparity is largely attributed to parental labor migration, which has left millions of rural students without direct parental care (30, 31).

Although cognitive behavioral therapy (CBT) (32), psychoanalytic therapy (33), pharmacological treatment (34), and complementary interventions (35–37) have proven effective in alleviating mental disorders and restoring patients' normal functioning, their implementation faces significant barriers among rural secondary school students in China. First, the exorbitant cost of professional counseling renders it financially inaccessible (38). Second, over half of the rural students reside in boarding schools (39), leaving limited time for regular counseling sessions. Furthermore, insufficient mental health literacy among parents and teachers in rural areas (40), may lead to the neglect of students' psychological issues or reluctance to seek treatment due to stigma concerns (41). In addition, many rural schools lack professional mental health teachers due to insufficient funding, limited infrastructure, and a shortage of qualified personnel (42, 43). Traditional educational priorities in rural areas emphasize academic achievement over mental health, resulting in limited recognition and support for psychological services within schools (44). Consequently, conventional mental health interventions remain impractical for most rural adolescents. This situation is similarly observed in rural and underserved communities in other countries (45–47).

In recent years, Artificial Intelligence (AI) has emerged as a transformative force, driving significant advancements in industries such as manufacturing, healthcare, architecture, and translation (48–50). As a subset of AI, chatbots are defined as intelligent programs that simulate human-like conversations (51) and have been recognized as a promising solution for mental health interventions (52). Within mental health domains, AI technologies—particularly generative AI—demonstrate capabilities in comprehending and generating natural language, with performance comparable to or exceeding human expertise in medical diagnosis, communication, and therapeutic practices (53). These AI-driven chatbots are based on robust decision-making framework. According to Guo and Hou (54) their research analyzes the conversational context to identify users' emotional states and subsequently utilize risk stratification models to assess mental health conditions and deliver personalized interventions. Unlike human therapists, AI chatbots offer round-the-clock support (55), providing a cost-effective solution to address mental health resource shortages in rural areas and potentially enhancing students' psychological wellbeing.

In summary, AI holds significant promise for delivering mental health education to rural secondary school students. However, existing research predominantly focuses on AI applications in universities (56) and urban primary/secondary schools (57), while targeted studies on rural adolescents remain scarce. Given the low adoption rates of AI technologies in rural settings, students' acceptance of such systems may significantly influence intervention efficacy. This study therefore aims to investigate rural students' acceptance of AI chatbots and its determinants, addressing current research gaps and proposing a scalable intervention framework for global implementation.

Part 1 reviews global mental health challenges and examines mental health issues prevalent among secondary school students in Chinese educational contexts. Part 2 examines key studies from existing literature, demonstrating the feasibility of AI applications in mental health education. Part 3 details the questionnaire-driven methodology and UTAUT2-derived research model. Parts 4–5 detail the survey implementation, data analysis procedures, and empirical findings. Part 6 discusses the implications of these findings and proposes evidence-based recommendations. The paper concludes with a synthesis of major contributions and future directions.

2 Related research

2.1 Application of artificial intelligence in educational contexts

Recent studies indicate that artificial intelligence (AI) offers transformative opportunities for the education sector, with applications extensively penetrating core domains such as curriculum development, learning guidance, and instructional assistance.

First, research on curriculum development—the cornerstone of educational practice—has yielded AI-powered frameworks and tools to enhance pedagogical design. For instance, Dickey and Bejarano (58) proposed the GAIDE framework, enabling educators to efficiently develop diverse, high-quality, and learner-centric materials through AI, thereby alleviating workload pressures. Heo and Kang (59) introduced the AIESTEP platform, which integrates learning content with practical modules and facilitates real-time teacher reflection for curriculum optimization. Sun et al. (60) further confirmed AI's efficacy in improving teachers' technical proficiency, self-efficacy, and instructional outcomes through experimental evidence.

The second research direction focuses on personalized learning guidance. In educational environments, students exhibit diverse learning experiences, proficiency levels, and individual characteristics. To enhance both engagement and knowledge retention, AI technology can dynamically adjust instructional content and methodologies based on learners' behavioral patterns, cognitive styles, and learning preferences (61), delivering tailored educational experiences (62). Representative implementations include Becerra et al. (63), who utilized generative AI to analyze MOOC data for personalized interventions and dropout prevention, and Qiu et al. (64), who developed the “Academic Quick Guide” system to streamline educational administration through 24/7 academic support.

The third direction emphasizes AI-augmented pedagogical support. As interactive practice constitutes a crucial component for assessing knowledge acquisition, some researchers have conducted related studies on this recently. Lin and Ye (65) developed a biology-focused chatbot to enhance academic performance via extracurricular scaffolding; Xu et al. (66) implemented a digital game-based chatbot system that boosts motivation through gameplay mechanics; while Banjade et al. (67) created an adaptive learning environment integrating AI-generated images with text-to-speech technology.

Notably, AI demonstrates particular efficacy in addressing educational disparities within China's resource-constrained rural regions. Niu et al. (68) mixed-methods study involving 130 teachers and students across nine schools, validated the effectiveness of AI platforms as effective pedagogical tools, particularly in resource-limited contexts.

2.2 Application of artificial intelligence in psychotherapy

Artificial intelligence has been explored in the mental health domain for decades. Pioneering systems include Eliza, which simulates psychotherapeutic dialogues to investigate problem origins (69). Woebot implements cognitive behavioral therapy to detect and mitigate depression (70), and Tess delivers emotional support to reduce anxiety and depression levels (71). Additionally, iHelpr facilitates depressive symptom self-assessment with improvement recommendations (72). Contemporary AI applications exhibit enhanced specialization and comprehensiveness in psychotherapy.

The primary consideration lies in supplementing or replacing traditional treatment methods. M and Nallasamy (73) demonstrate that chatbots delivering behavioral therapy through virtual coaching can enhance clinical outcomes, reduce social stigma, and bridge treatment gaps. Vahedifard et al. (74) analyze ChatGPT's potential in psychiatry, acknowledging its emotional support capabilities while examining privacy and ethical concerns. Eid et al. (75) leverage AI-powered patient data analysis to optimize depression treatment plans to enable precision medication customization.

Special population factors constitute another critical dimension. Habicht et al. (76) reveal through studies on marginalized groups (e.g., bisexual individuals) that AI chatbots significantly lower treatment barriers compared to conventional approaches, promoting equitable access to mental health services. Wang and Li (77) establish through comparative experiments that AI interventions not only alleviate geriatric depression but also reduce economic burdens.

Regarding therapeutic efficacy for student mental health, Mahmud and Porntrakoon (78) propose AI as a viable complement or alternative to traditional treatments for Thai university students, though AI requires enhanced user-friendly designs and privacy safeguards. Moreover, Oghenekaro and Okoro (79) validate through combined quantitative-qualitative assessments that AI technologies provide personalized support significantly improving psychological states. Klos et al. (80) confirmed Tess's (https://tess.x2ai.com/) intervention potential for anxiety and depression in Argentine university students via controlled trials. Liu et al. (81) demonstrated that chatbot-based self-help interventions outperformed bibliotherapy in depression management, and Wang et al. (82) documented high trust levels of middle school students in using AI wristbands for mental health intervention.

Notably, current research predominantly focuses on hospital patients and university populations (83). Research on middle school students, particularly in rural areas, remain critically underexplored.

2.3 Unified theory of acceptance and use of technology (UTAUT) in AI applications for mental health

The UTAUT model provides robust theoretical foundation for investigating users‘ acceptance and adoption behaviors toward emerging technologies. Through its core constructs of Performance Expectancy, Effort Expectancy, and Social Influence, this framework systematically analyzes key drivers underlying rural secondary school students' utilization of AI-powered psychotherapeutic tools. The model enables systematic understanding and prediction of this population's acceptance of AI mental health interventions, offering theoretical guidance for future implementation designs. Several scholars have employed UTAUT to explore AI applications in mental health contexts. For instance, Alojail (84) utilized UTAUT to examine user acceptance of digital mental health interventions, revealing that perceived usefulness, perceived ease of use, and facilitating conditions exerted positive effects on user attitudes and behavioral intentions, whereas social influence demonstrated non-significant effects. Henkel et al. (85) extended UTAUT to LGBTQIA+ populations in mental health chatbot adoption studies, identifying Performance Expectancy, Social Influence, Willingness to Self-disclose, and Trust as critical determinants of behavioral intention, with gender showing no moderating effects. Li et al. (136) investigated Americans with depression/anxiety, finding Performance Expectancy, Price Value, and Descriptive Norms were all positively related to behavioral intention for both treatment-naïve and experienced cohorts, while Effort Expectancy was negatively associated with behavioral intention.

These findings suggest AI adoption intentions correlate closely with perceived usefulness, perceived ease of use, and risk perceptions. Therefore, applying the UTAUT framework to investigate influencing factors in rural adolescents' acceptance of AI-assisted mental health education proves methodologically appropriate.

3 Research methods

3.1 Theoretical model

The UTAUT is an Integrated Technology Acceptance Model proposed by Venkatesh et al. (86) through synthesizing and refining prior Technology Acceptance Model (TAM) research. Venkatesh et al. (87) further refined and validated the UTAUT framework, introducing the UTAUT2 model. This enhanced framework incorporates additional variables to comprehensively explain and predict users' acceptance and adoption behaviors toward emerging technologies. The UTAUT2 model operationalizes seven core constructs including Performance Expectancy, Effort Expectancy, Social Influence, Hedonic Motivation, Price Value, Facilitating Conditions, and Habit, along with three moderators which are Gender, Age, and Experience. Empirical validation demonstrates that this model explains 74% of the variance in behavioral intention across diverse cultural and social contexts, significantly outperforming the original UTAUT framework (88). Consequently, UTAUT2 serves as the theoretical foundation for this study.

3.2 Variable selection and hypotheses

3.2.1 Variable selection

This study retains four core constructs from the original UTAUT2 model: Performance Expectancy (PE), Effort Expectancy (EE), Social Influence (SI), and Hedonic Motivation (HM). Facilitating Conditions (FC) and Price Value (PV) were excluded because education departments will provide schools with equipment and training for free to facilitate students' use. Given the exploratory nature of AI-assisted mental health education in rural Chinese secondary schools, we removed habit and experience variables from UTAUT2. This decision was also influenced by students' limited experience and developing autonomy. However, we retained gender as a moderator. The age variable has been replaced by the grade moderating variable to examine behavioral intention variations across different academic stages. Considering that students in rural Chinese secondary schools have less exposure to AI technology, they may be concerned about privacy risks when using such unfamiliar tools. Moreover, as mental health education involves emotional interactions, human-like chatbots may affect students' attitudes toward usage. Therefore, Perceived Risk (PR) and Perceived Anthropomorphism (PA) are included as variables. Our research model is shown in Figure 1.

Figure 1

3.2.2 Hypotheses development

Performance expectancy is defined as the degree to which an individual believes using the system will enhance job performance (89). Huang et al. (90) confirmed in their research on health chatbots that performance expectancy is a crucial variable. As a well-established predictor of behavioral intention (86), performance expectancy in this context reflects rural adolescents' perceptions that AI chatbots can improve mental health outcomes. Students who perceive AI as effective in addressing psychological issues or enhancing wellbeing are more likely to demonstrate higher behavioral intention. We derive the following hypothesis from this:

• H1. PE positively influences rural students' BI toward AI-assisted mental health education.

Effort expectancy is defined as the perceived ease or difficulty of utilizing a technology (87). Prior studies have shown that effort expectancy plays a vital role in the use of Learning Management System (LMS) and AI Tools (91, 92). Specifically, effort expectancy in this study measures the effort and time investment required for rural students to achieve desired psychological improvements through AI chatbot interactions. If students perceive the AI system as user-friendly, easy to understand and demanding minimal learning effort, their willingness to adopt the technology will markedly increase. Thus, the following hypothesis was proposed:

• H2. EE positively influences rural students' BI toward AI-assisted mental health education.

Social influence refers to the extent to which an individual is influenced by the behavior of groups from outside (87). Nurkhin (93) confirmed that social influence has a positive impact on the willingness to use online learning. In this study, social influence refers to whether rural secondary school students are influenced by their peers, parents, and teachers when using AI chatbots. If students' peers or people around them widely recognize and actively use AI chatbots, students themselves may also be more inclined to adopt them. Therefore, the following hypothesis was proposed:

• H3. SI positively influences rural students' BI toward AI-assisted mental health education.

Hedonic motivation denotes the fun and pleasure consumers feel when using a technology (87). Earlier investigations have found that hedonic motivation has the strongest influence on the behavioral intention to use mobile applications (94, 95). In this study, hedonic motivation refers to whether or not rural secondary school students feel pleasure when using an AI chatbot. If students find that using the chatbot leads to a pleasurable experience, it may increase their willingness to continue using it. We deduce the following hypothesis:

• H4. HM positively influences rural students' BI toward AI-assisted mental health education.

Perceived risk refers to users' uncertainty regarding potential risks associated with technology adoption (87). For example, concerns about data privacy vulnerabilities (96). Empirical evidence indicates a negative correlation between perceived privacy loss and chatbot acceptance (97). Wu et al. (98) substantiated that perceived risk (PR) significantly negatively impacts students' AI-assisted learning. In this study, students' behavioral intentions may diminish if they perceive AI technologies as posing privacy threats. Thus, we propose:

• H5. PR negatively influences rural students' BI toward AI-assisted mental health education.

Perceived anthropomorphism in AI chatbots refers to endowing machines with human-like traits, thereby enabling natural language conversational capabilities (99). For users, the presence of empathy and empathic abilities in therapeutic chatbots is positively correlated with their behavioral intention (100). In this study, the degree of anthropomorphism influences students' trust and emotional affinity toward AI chatbots. If a chatbot exhibits higher anthropomorphic features, students may show greater willingness to interact with it, thereby increasing their adoption intention. This leads to the following hypothesis:

• H6. PA positively influences rural students' BI toward AI-assisted mental health education.

Gender and age are recognized moderators of technology acceptance (101). In this study, age is operationalized as grade level, with both gender and grade hypothesized to shape attitudes toward chatbots. Hence, the hypotheses:

• H7. Gender moderates the effects of HM (H7a), PR (H7b), and PA (H7c) on BI.

• H8. Grade moderates the effects of HM (H8a), PR (H8b), and PA (H8c) on BI.

4 Scale design and data collection

This article designed scales based on the characteristics of each variable and collected data through the distribution of questionnaires.

4.1 Questionnaire design

Based on the above hypotheses and theoretical framework, this study developed a survey instrument by adapting validated measurement scales from prior literature. The constructs of performance expectancy, effort expectancy, social influence, hedonic motivation, and behavioral intention were drawn directly from the Unified Theory of Acceptance and Use of Technology 2 (UTAUT2) proposed by Venkatesh et al. (87). The original items were slightly reworded to fit the context of AI chatbots for mental health—for example, by replacing general terms such as “technology” or “system” with “AI chatbot.” These subscales have been widely validated and applied in studies involving both general consumers and student populations. The perceived risk scale was adapted from Liu and Tao (102), originally developed in the context of smart healthcare services, while the perceived anthropomorphism scale was adapted from Liu and Cao (103), who examined users' responses to human-like virtual chatbots. In both cases, domain-specific terms (e.g., “smart healthcare services” or “Alex/Robo”) were replaced with references to AI chatbots, while preserving the original constructs. All items were further reviewed to ensure clarity, contextual relevance, and age appropriateness for secondary school students.

The questionnaire comprises two primary parts. Firstly, it includes demographic information, namely gender and grade level. Secondly, it contains influencing factor metrics. These metrics are structured across seven dimensions, and there are a total of 21 measurement items in this section. For detailed information about the instrument and source references, please refer to Table 1. All items were quantified using a 5-point Likert scale, where 1 indicates “strongly disagree” and 5 denotes “strongly agree.” Prior to formal data collection, a pilot test was conducted with 65 respondents to assess the questionnaire's reliability and validity. The instrument was subsequently refined based on pilot test outcomes to establish the final version for official administration. Table 1 displays the finalized items.

4.2 Data collection

China has established the world's largest education system (104). Henan Province, which has the largest basic education population nationwide (105), represents a typical educational demographic. For instance, Zhoukou City contains a rural population exceeding 50% of its total residents with 676,900 students enrolled in general secondary schools (106, 107). These statistical profiles effectively reflect shared characteristics of educational contexts in China's rural areas. Consequently, Zhoukou was selected as the case study area to investigate rural middle school students' behavioral intentions toward AI chatbots through questionnaire surveys.

The formal survey via the Questionnaire Star platform (https://www.wjx.cn/) consisted of two phases. In Phase I, questionnaires were electronically distributed via WeChat parent-teacher groups, yielding 202 responses (all participants from Zhoukou rural schools). Before answering the questionnaire, all participants were required to watch an embedded instructional video introducing the concept, functions, and limitations of AI chatbots for mental health support. This ensured that participants had a consistent and objective understanding of the chatbot. After excluding invalid responses (e.g., short-duration submissions and uniform answers), 187 valid questionnaires were retained. In Phase II, a stratified random sampling method was employed (25 students per grade, none of whom participated in the first survey) to conduct an in-person survey during noon study sessions. Participants first viewed an AI chatbot instructional video, followed by onsite questionnaire completion under researcher supervision. From 150 collected responses, 130 met the validity criteria after data cleansing.

This study was conducted ethically. All participants signed an informed consent form prior to the survey. Participation was voluntary and anonymous. Since the AI chatbot was presented only in a simulated video, no real interaction or psychological risk was involved.

As shown in Table 2, the sample comprised 185 male participants (58.36%) and 132 female participants (41.64%). Regarding academic progression, 51.42% were junior high school students, while 48.58% were senior high school students.

5 Data analysis

This study employed SPSS 26 and AMOS 23 for data processing and hypothesis testing. The analytical procedure consisted of five stages. First, reliability analysis of the scales was conducted using SPSS. Subsequently, confirmatory factor analysis (CFA) was performed in AMOS to assess structural validity, and exploratory factor analysis (EFA) in SPSS was employed concurrently to examine dimensionality and discriminant validity. Next, descriptive statistics and normality tests were conducted through SPSS. Then, structural equation modeling (SEM) was constructed via AMOS. Finally, hierarchical regression analysis was implemented to evaluate the impacts of demographic variables on research outcomes.

5.1 Reliability analysis

To ensure measurement quality, scale reliability was initially verified as a prerequisite for subsequent analyses. Internal consistency across dimensions was evaluated using Cronbach's alpha coefficient (α), where values range from 0 to 1 with higher coefficients indicating superior reliability. The established thresholds were: < 0.6 (unacceptable), 0.6–0.7 (acceptable), 0.7–0.8 (good), 0.8–0.9 (excellent) and >0.9 (ideal). Using 317 completed questionnaires, the calculated α values for all seven latent variables exceeded 0.7 (Table 3). These results confirm the measurement scale demonstrates satisfactory internal consistency and meets psychometric standards.

5.2 Validity test

Validity tests were conducted to verify whether the observed variables accurately reflect the latent variables, ensuring the validity of the questionnaire. Since all observed variables in this questionnaire were derived from previously validated scales, confirmatory factor analysis (CFA) was employed to examine the alignment between factor-observed variable relationships and theoretical assumptions. In this study, confirmatory factor analysis (CFA) was conducted using AMOS software (IBM SPSS, Chicago, OH, USA), with the analytical results illustrated in Figure 2. As shown in Table 4, the model fit indices demonstrated satisfactory performance: the CMIN/DF (chi-square/degrees of freedom ratio) of 2.502 fell within the recommended range of 1–3, and the RMSEA (root mean square error of approximation) of 0.059 remained below the 0.08 threshold, indicating acceptable model fit. Furthermore, other indices including the NFI (normed fit index) and CFI (comparative fit index) achieved good levels exceeding 0.9. These results confirm the model's strong fit, validating the questionnaire's effectiveness.

Figure 2

CFA encompasses convergent validity and discriminant validity. Convergent validity was assessed via composite reliability (CR) and average variance extracted (AVE). The AVE value, a critical metric for evaluating convergent validity, determines the strength of relationships between measurement items and their corresponding factors. Higher AVE values indicate greater reliability and convergent validity of the construct, with an ideal threshold exceeding 0.5. Composite reliability (CR) reflects the internal consistency of observed variables, where higher CR values signify stronger internal consistency and convergence. As shown in Table 5, all seven latent variables exhibited CR values above 0.7 and AVE values exceeding 0.5, confirming satisfactory composite reliability and convergent validity. Additionally, all standardized factor loadings surpassed 0.7, indicating strong explanatory power of individual items for their respective dimensions. Collectively, these results demonstrate robust convergent validity and composite reliability across all dimensions.

The data were subjected to varimax rotation to elucidate factor-item correlations. Given the seven predefined variables in the theoretical model, principal component analysis extracted seven common factors, yielding the rotated component matrix in Table 6. Analysis of factor loading coefficients (highlighted in bold) revealed communality values exceeding 0.4 for all items, indicating strong item-factor associations and effective information extraction by the factors. Furthermore, the grouping patterns of measurement indicators under each variable aligned with theoretical expectations.

Discriminant validity, which ensures the distinct measurement of different constructs, was evaluated by comparing the square roots of AVE values with inter-factor correlation coefficients. As presented in Table 7, all AVE square roots exceeded the absolute values of corresponding inter-factor correlations, confirming adequate discriminant validity of the measurement model.

5.3 Descriptive statistics and normality test

Table 8 presents descriptive statistics and normality test results for the measured constructs in this study. The analysis of descriptive statistics reveals that all variables' mean scores range between 3 and 4. Given the 1–5 positive scoring scale, these results indicate that participants' awareness of using AI chatbots for mental health education assistance is above medium level. Normality tests for measurement items were conducted through skewness and kurtosis analyses. Following Kline's (137) criteria (absolute skewness < 3; absolute kurtosis < 8), the data were considered approximately normally distributed. As evidenced in Table 8, all measurement items' absolute skewness and kurtosis values fall within these thresholds, confirming their compliance with normal distribution assumptions.

5.4 Structural equation analysis

As shown in Table 9, PE (β = 0.261, p < 0.001), EE (β = 0.281, p < 0.001), SI (β = 0.136, p < 0.01), and PA (β = 0.176, p < 0.01) exerted significant positive effects on BI. Thus, H1, H2, H3 and H6 were supported. PR (β = −0.154, p < 0.01) demonstrated a negative association with BI, validating H5. Contrary to expectations, HM (β = −0.052, p > 0.10) showed no significant relationship with BI, leading to the rejection of H4.

The effect size for each predictor was evaluated using Cohen's f², which indicates the unique contribution of each variable to the explained variance in BI. According to Cohen's (108) benchmarks, f² values of 0.02, 0.15, and 0.35 represent small, medium, and large effects respectively. In this study, PE, EE, and PA showed medium to large effect sizes, SI and PR exhibited a small effect size.

5.5 Effect of moderating variables

The path analysis results indicate that the hypothesized relationship between HM and BI (H4) was not supported. Consequently, hypotheses H7a and H8a were rejected, and their moderation effects were not further examined.

This study employed hierarchical regression analysis to investigate the moderating effects of gender and grade. The analytical framework positioned PR and PA as independent variables, and BI as the dependent variable, with gender and grade as moderators. Interaction terms were systematically incorporated into sequential regression models.

As shown in Table 10, with gender included in Model 2, neither the PR × Gender (β = 0.008, p > 0.05) nor PA × Gender (β = −0.03, p > 0.05) interaction terms reached statistical significance. These results confirm that there are no gender moderation effects on the relationship between PR and BI. Similarly, there are no such effects on the relationship between PA and BI. As a result, hypotheses H7b and H7c are rejected.

As shown in Table 11, after incorporating grade into the model, the interaction term PR × Grade showed a significantly positive coefficient (β = 0.132, p < 0.05), indicating that grade exerts a significant positive moderating effect on the relationship between PR and BI, thus supporting Hypothesis H8b. This moderation effect corresponds to a ΔR² of 0.017, indicating a moderate level of effect size. In contrast, the PA × Grade interaction term had no significant effect on BI (β = −0.014, p > 0.05), demonstrating that the moderator grade does not significantly influence the relationship between PA and BI, leading to the rejection of Hypothesis H8c.

To comprehensively understand the specific mechanism of grade moderation and the overall model configuration, we conducted further analyses which involving three nested models. Model 1 includes the independent variable (PR). Model 2 adds the moderator variable (grade) to Model 1. Model 3 enhances explanatory power by incorporating the interaction term (PR × Grade) into Model 2. Here, we use the unstandardized coefficient B to report the specific moderating effects of grade. As shown in Table 12, model 3 reveals significant coefficients for the interaction terms “PR × Grade 3.0” (B = 0.543, t = 2.436, p = 0.015^*) and “PR × Grade 5.0” (B = 0.697, t = 3.056, p = 0.002^**), demonstrating grade-specific moderation effects on the PR-BI relationship at grades 9 and 11.

6 Discussion

The primary objective of this study was to investigate key determinants influencing rural secondary school students' adoption of AI chatbots for mental health education support. This discussion juxtaposes empirical findings with research hypotheses to elucidate mechanisms underlying behavioral intention toward AI chatbot utilization.

6.1 Performance expectancy, effort expectancy, social influence and behavioral intention

As hypothesized, PE, EE, and SI demonstrated significant positive correlations with BI. Consistent with prior research, PE emerged as the strongest predictor of technology adoption (109). This suggests rural adolescents' conviction in AI chatbots' efficacy for mental health improvement constitutes a pivotal adoption driver.

As AI chatbots are digital devices, their ease of use strongly impacts users' BI to adopt and continue usage (110). While general technological proficiency with digital devices among Chinese students is relatively high (111), rural students' limited access to AI technologies creates unique usability challenges. Overly complex interaction mechanisms may result in user disengagement and eventual discontinuation of use. Furthermore, inadequate usability design of the chatbots may hinder teachers' capacity to deliver essential technical assistance.

According to the findings, the more rural secondary school students perceived social support for chatbot use, the higher their BI. Peer influence, teacher support, and parental input are critical in rural settings. Therefore, AI chatbots should be embedded in rural educational practices by creating a supportive community in rural school settings and encouraging students to use the devices.

6.2 Perceived anthropomorphism and behavioral intention

Moreover, anthropomorphic cues—such as facial expressions, human-like voice tone, and emotionally resonant responses—can enhance adolescents' emotional engagement with chatbots by triggering social cognitive mechanisms similar to those activated in human–human interaction (112). These features reduce perceived psychological distance and promote trust by making the AI appear more intentional and empathetic (113). In therapeutic contexts, trust is crucial for fostering self-disclosure and sustained use (114–116). When adolescents feel the chatbot “understands” them emotionally, they are more likely to share private thoughts and continue using the tool for support—especially in environments where human contact is limited or stigmatized.

6.3 Perceived risk and behavioral intention

PR exhibited a negative association with BI, where privacy concerns significantly deterred adoption. This aligns with existing evidence (117). From a neuropsychological perspective, perceived risk may disrupt adolescents' emotional development by overstimulating brain regions associated with threat processing (118). These areas, when hyperactivated by digital threats like privacy breaches, impair emotion regulation control systems (119). This imbalance can heighten anxiety, reduce trust, and lead to withdrawal from AI-assisted tools intended for mental health support (120–122), thereby limiting their therapeutic effectiveness among rural adolescent.

6.4 Hedonic motivation and behavioral intention

Contrary to our hypotheses, HM demonstrated no significant correlation with BI. This suggests rural adolescents do not prioritize recreational features as critical adoption determinants—a finding contradicting prior investigations (123, 124). A potential explanation is that, as Digital Natives (125), students maintain frequent exposure to digital technologies and demonstrate a high level of interest in AI systems (126), which may shape their expectations of smart systems' inherent hedonic properties. Concurrently, instrumental utility may outweigh recreational value, with students prioritizing practical therapeutic assistance over entertainment features. However, this does not imply that hedonic design should be neglected. Rather, AI chatbots should incorporate basic entertainment functionalities commonly found in electronic products, while emphasizing practical effectiveness.

6.5 The moderating effects of gender and grade

First, we examine the moderating factors affecting the relationship between PR and BI. Our study found that gender did not significantly moderate the relationship between PR and BI. Owing to diminishing gender stereotyping and the structural transition of Chinese rural communities to “semi-familiarity societies” (127–130), women's need for rigorous privacy management as a strategy to maintain social image has decreased. Furthermore, digital-native adolescents exhibit technology usage patterns characterized by diminished gender disparities, resulting in comparable technology acceptance behaviors across genders among rural students. The moderating effect of grade level can be ascribed to the structural pressures that students in grade 9 and grade 11 are confronted with. At this stage in China's education system, they are at the critical juncture of universal-vocational streaming and subject selection (131, 132). As a result, they are in desperate need of mental health support. Meanwhile, to prevent new pressures caused by privacy violations, the potential risks of AI technology.

Then, Regarding the relationship between PA and BI, rural participants' limited AI exposure indicated stronger functional orientation in technology adoption decisions. PA demonstrated an inverted U-shaped relationship with BI, where excessive anthropomorphism decreased adoption likelihood (133). These patterns showed no significant variation by gender or grade level, confirming the absence of moderation effects.

6.6 Suggestions

Based on the aforementioned findings, it is evident that multiple factors must be considered when developing a mental health chatbot for rural middle school students.

First, it is essential to enhance the technical performance of the chatbot. It is necessary to ensure that the algorithm can accurately identify early signs of mental health disorders (e.g., depression, anxiety) and provide targeted interventions such as relaxation training and time management. Concurrently, real-time monitoring of psychological states and adaptive feedback mechanisms are essential. Collaboration between educational authorities and certified mental health institutions should be established to develop teleconsultation modules for clinical-grade support. Next, cartoon-style icons and hierarchically simplified operational workflows aligned with adolescent cognitive patterns are required to reduce usability barriers. Supplementary graphic manuals and instructional videos must be provided to facilitate technical proficiency among teachers and students. To expand social influence, schools and education departments should promote chatbot adoption through teacher training workshops and parental awareness programs, thereby improving credibility and acceptance in educational settings. A home-school coordination platform is recommended to enable real-time parental access to students' psychological data for early intervention. Anthropomorphic features such as simulated gestures and emotional language processing can significantly strengthen user emotional bonds and retention rates (134). Therefore, natural language processing should prioritize colloquial expressions over technical jargon to align with adolescents' linguistic habits. Given rural students' dialect preferences (135), dialect recognition modules must be integrated to ensure interaction accuracy. Dynamic effects mimicking human behaviors like nodding and blinking can further enhance user immersion. Robust data protection mechanisms are critical. During initial onboarding, clearly communicate the scope of data collection, usage protocols, and encryption procedures to users. End-to-end encryption technology should be employed to secure chat logs and personal information during transmission and storage. Additionally, anonymization options are necessary to alleviate stigma-related concerns. Finally, implement personalized content delivery systems. Leveraging machine learning algorithms, tailor interventions for subgroups with distinct traits (e.g., high academic stress, introversion), thereby addressing diverse needs and improving user trust and engagement.

7 Research limitations and future directions

This study explores the emerging field of AI chatbot applications in mental health education for rural secondary school students. To better align with the specific conditions of the educational context, two external variables, perceived risk (PR) and perceived anthropomorphism (PA), have been integrated based on the UTAUT2 model. Valid data were obtained through standardized questionnaires distributed to students in rural secondary schools in Zhoukou, China, and a structural equation model was developed for validation. Findings revealed that PE, EE, SI, and PA were positively correlated with BI, while PR showed a significant negative correlation. HM did not reach statistical significance. Additionally, gender and grade exhibited significant moderating effects. Practical recommendations include optimizing chatbot functionality, streamlining operational workflows, amplifying social influence, and enhancing user experience.

However, this study has limitations. First, the sample was geographically concentrated in Zhoukou City, a single rural area in central China. This site was chosen due to its large rural population and representative educational characteristics, which broadly reflect the conditions of many rural regions in China. While this provides valuable insights, it may not fully capture the diversity of rural areas nationwide in terms of culture, infrastructure, and digital access. Therefore, future research should consider more geographically and demographically diverse sampling to enhance external validity and improve the generalizability of the findings. Second, the questionnaire-based quantitative methodology, while generating robust empirical data, failed to capture qualitative dimensions of students' psychological experiences. In response, subsequent studies could adopt mixed-methods approaches such as in-depth interviews and ethnographic observation. These methods will allow researchers to gain deeper insights into students' unmet needs and emotional dynamics. Furthermore, a limitation of the present study is the absence of qualitative interviews, which could have offered more nuanced insights into participants' experiences with the AI chatbot. Future research should consider incorporating such qualitative methods, alongside comparisons with traditional mental health education, to gain a deeper understanding of user perceptions and preferences.

References

• 1.

  World Health Organization . World Mental Health Report: Transforming Mental Health for All. Geneva: World Health Organization (2022)

  • Google Scholar

• 2.

  Santomauro DF Herrera AMM Shadid J Zheng P Ashbaugh C Pigott DM et al . Global prevalence and burden of depressive and anxiety disorders in 204 countries and territories in 2020 due to the COVID-19 pandemic. Lancet. (2021) 398:1700–12. 10.1016/S0140-6736(21)02143-7

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 3.

  Kang CY Yang JZ . Prevalence of mental disorders in China. Lancet Psychiatry. (2022) 9:13. 10.1016/S2215-0366(21)00400-4

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 4.

  Salleh MR . The burden of mental illness: an emerging global disaster. J Clin Health Sci. (2018) 3:1–8. 10.24191/jchs.v3i1.6150

  • CrossRef
  • Google Scholar

• 5.

  Perkins R Rinaldi M . Unemployment rates among patients with long-term mental health problems: a decade of rising unemployment. Psychiatrist. (2002) 26:295–8. 10.1192/pb.26.8.295

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 6.

  Brännlund A Strandh M Nilsson K . Mental-health and educational achievement: the link between poor mental-health and upper secondary school completion and grades. J Ment Health. (2017) 26:318–25. 10.1080/09638237.2017.1294739

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 7.

  Arias D Saxena S Verguet S . Quantifying the global burden of mental disorders and their economic value. eClinicalMedicine. (2022) 54:101675. 10.1016/j.eclinm.2022.101675

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 8.

  Greenberg PE Fournier A-A Sisitsky T Pike C Kessler RC . The economic burden of adults with major depressive disorder in the United States (2005 and 2010). J Clin Psychiatry. (2015) 76:155–62. 10.4088/JCP.14m09298

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 9.

  Hsieh C-R Qin XZ . Depression hurts, depression costs: the medical spending attributable to depression and depressive symptoms in China. Health Econ. (2018) 27:525–44. 10.1002/hec.3604

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 10.

  Tomlinson M . Global mental health: a sustainable post millennium development goal?Int Health. (2013) 5:1–3. 10.1093/inthealth/iht001

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 11.

  Wilson Z . Out of the shadows: making mental health a global development priority. Ment Health Matters. (2016) 3:1–4. 10.10520/EJC189629

  • CrossRef
  • Google Scholar

• 12.

  United Nations Department of Economic and Social Affairs . Transforming our world: The 2030 Agenda for Sustainable Development. United Nations Department of Economic and Social Affairs (2015). Available online at: https://sdgs.un.org/2030agenda (Accessed January 15, 2025).

  • Google Scholar

• 13.

  World Health Organization . (2021). Comprehensive Mental Health Action Plan 2013-2030. World Health Organization. https://www.who.int/zh/publications/i/item/9789240031029

  • Google Scholar

• 14.

  Chen K . Adolescent mental health crisis: current situation, challenges and solutions. Health Care. (2024) 2024:1–7. 10.61173/crqkh247

  • CrossRef
  • Google Scholar

• 15.

  Racine N McArthur BA Cooke JE Eirich R Zhu J Madigan S . Global prevalence of depressive and anxiety symptoms in children and adolescents during COVID-19: a meta-analysis. JAMA Pediatr. (2021) 175:1142–50. 10.1001/jamapediatrics.2021.2482

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 16.

  Lahti M Groen G Mwape L Korhonen J Breet E Chapima F et al . Design and development process of a youth depression screening m-health application for primary health care workers in South Africa and Zambia: an overview of the MEGA project. Issues Ment Health Nurs. (2020) 41:24–30. 10.1080/01612840.2019.1604919

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 17.

  Chen ZY . Blue Book Report | 2022 Youth Mental Health Status Survey Report. China Net Psychology China (2023). Available online at: http://psy.china.com.cn/2023-08/10/content_42459520.html (Accessed August 10, 2023).

  • Google Scholar

• 18.

  Institute for Depression Research . Report | “2022 National Depression Blue Book”. Depression Research Institute (2022). Available online at: https://www.chinadevelopmentbrief.org.cn/news/detail/23683.html (Accessed July 11, 2022).

  • Google Scholar

• 19.

  Good Mood Internet Hospital . “2023 Annual Blue Book on Mental Health in China” Released. Good Mood Internet Hospital (2023). Available online at: https://www.haoxinqing.cn/article/detail31087.html (Accessed October 9, 2023).

  • Google Scholar

• 20.

  Taylor A Leslie S Boddie K . “Promoting mental wellbeing in young people aged 12–18 years: opportunities for design.” In: Proceedings of the 31st International BCS Human Computer Interaction Conference (HCI 2017) (2017). p. 1–12. 10.14236/ewic/HCI2017.33

  • CrossRef
  • Google Scholar

• 21.

  Ministry of Education of the People's Republic of China . Notice on the issuance of the “Comprehensive Strengthening and Improvement of Mental Health Work for Students in the New Era Special Action Plan (2023–2025)”. Gazette of the Ministry of Education of the People's Republic of China (2023). p. 4–9.

  • Google Scholar

• 22.

  The State Council of the People's Republic of China . The Central Committee of the Communist Party of China and the State Council issued the “Healthy China 2030 Planning Outline. The State Council of the People's Republic of China (2016). Available online at: https://www.gov.cn/gongbao/content/2016/content_5133024.htm (Accessed October 25, 2016).

  • Google Scholar

• 23.

  The State Council of the People's Republic of China . Jointly issued by 22 departments: “Guiding Opinions on Strengthening Mental Health Services”. The Central People's Government of the People's Republic of China (2017). Available online at: https://www.gov.cn/xinwen/2017-01/24/content_5162861.htm (Accessed January 24, 2017).

  • Google Scholar

• 24.

  Hu JS Wang DF Li Q . A study of mental health status and its determinants among the rural middle school students. Chin J Clin Psychol. (2005) 13:74–6. 10.3969/j.issn.1005-3611.2005.04.025

  • CrossRef
  • Google Scholar

• 25.

  Ma JH . Analysis of the mental health of rural middle school students and research on countermeasures. Contemp Youth Res. (2005) 23:16–9. 10.3969/j.issn.1006-1789.2005.05.004

  • CrossRef
  • Google Scholar

• 26.

  Yu Q Xiong SM Liu FW Dai LL Wang YF Wang C et al . Investigation on mental health status of left-behind children in Bijie rural middle school. J Zunyi Med Univ. (2018) 41:228–32. 10.3969/j.issn.1000-2715.2018.02.022

  • CrossRef
  • Google Scholar

• 27.

  Chen WL Huang Y Riad A . Gender differences in depressive traits among rural and urban Chinese adolescent students: secondary data analysis of nationwide survey CFPS. Int J Environ Res Public Health. (2021) 18:9124. 10.3390/ijerph18179124

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 28.

  Jiang QW She XS Dill S-E Sylvia SY Singh MK Wang H et al . Depressive and anxiety symptoms among children and adolescents in rural China: a large-scale epidemiological study. Int J Environ Res Public Health. (2022) 19:5026. 10.3390/ijerph19095026

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 29.

  Xue XG . Investigation and study on mental health status of students from different regions. Health Med Res Pract. (2012) 9:26–7.

  • Google Scholar

• 30.

  Meng X Yamauchi C . Children of migrants: the cumulative impact of parental migration on children's education and health outcomes in China. Demography. (2017) 54:1677–714. 10.1007/s13524-017-0613-z

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 31.

  Gao Y Li LP Kim JH Congdon N Lau J Griffiths S . The impact of parental migration on health status and health behaviours among left behind adolescent school children in China. BMC Public Health. (2010) 10:1–10. 10.1186/1471-2458-10-56

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 32.

  Butler AC Chapman JE Forman EM Beck AT . The empirical status of cognitive-behavioral therapy: a review of meta-analyses. Clin Psychol Rev. (2006) 26:17–31. 10.1016/j.cpr.2005.07.003

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 33.

  Gaskin C . The effectiveness of psychoanalysis and psychoanalytic psychotherapy: a literature review of recent international and Australian research. Psychother Couns J Aust. (2016) 4:1–23. 10.59158/001c.71168

  • CrossRef
  • Google Scholar

• 34.

  Adjei K Adunlin G Ali AA . Impact of sertraline, fluoxetine, and escitalopram on psychological distress among United States adult outpatients with a major depressive disorder. Healthcare. (2023) 11:740. 10.3390/healthcare11050740

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 35.

  Motlagh EG Bakhshi M Davoudi N Ghasemi A Moonaghi HK . The physical and psychological outcomes of art therapy in pediatric palliative care: a systematic review. J Res Med Sci. (2023) 28:13. 10.4103/jrms.jrms_268_22

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 36.

  Satyawan VE Rusdiana OE Latifah MN . The role of forest therapy in promoting physical and mental health: a systematic review. IOP Conf Ser: Earth Environ Sci. (2022) 959:012027. 10.1088/1755-1315/959/1/012027

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 37.

  Knapen J Vancampfort D Moriën Y Marchal Y . Exercise therapy improves both mental and physical health in patients with major depression. Disabil Rehabil. (2015) 37:1490–5. 10.3109/09638288.2014.972579

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 38.

  Yan Y Tu YQ . The impact of China's urban and rural economic revitalization on the utilization of mental health inpatient services. Front Public Health. (2022) 10:1043666. 10.3389/fpubh.2022.1043666

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 39.

  The Ministry of Education . Response of the Ministry of Education to the Proposal No. 3867 of the Fifth Session of the Twelfth National People's Congress. Ministry of Education of the People's Republic of China (2017). Available online at: http://www.moe.gov.cn/jyb_xxgk/xxgk_jyta/jyta_jijiaosi/201712/t20171219_321861.html (Accessed September 8, 2017).

  • Google Scholar

• 40.

  World Health Organization . The World Health Report: 2001: Mental Health: New Understanding, New Hope. World Health Organization (2001). Available online at: https://books.google.com.tw/books?hl=en&lr=&id=GQEdA-VFSIgC&oi=fnd&pg=PR9&ots=d3UBTDedyA&sig=_db2eTs-5Qf0Fi_VRu3CTeWmeFk&redir_esc=y#v=onepage&q&f=false

  • Google Scholar

• 41.

  Thornicroft G Brohan E Rose D Sartorius N . Global pattern of experienced and anticipated discrimination against people with schizophrenia: a cross-sectional survey. Lancet. (2009) 373:408–15. 10.1016/S0140-6736(08)61817-6

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 42.

  Ren Z Wu W . “Study on student management strategy of northwest rural boarding schools.” In 7th International Conference on Economy, Management, Law and Education (EMLE 2021) (2022). p. 376–82. 10.2991/aebmr.k.220306.056

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 43.

  Cao T Wang F . The research of psychological health education of the primary schools. Open Access Libr. (2020) 7:1–7. 10.4236/oalib.1107015

  • CrossRef
  • Google Scholar

• 44.

  Liu Z . Current situation and improvement measures of mental health education in rural schools. New Curric Mid-Month Ed. (2015) 1:252.

  • Google Scholar

• 45.

  Putri A Hartono R Nabila S . Examining health disparities in access to mental health services in urban and rural areas. Int J Health Soc Behav. (2024) 1:6–10. 10.62951/ijhsb.v1i1.146

  • CrossRef
  • Google Scholar

• 46.

  Palomin A Takishima-Lacasa J Selby-Nelson E Mercado A . Challenges and ethical implications in rural community mental health: the role of mental health providers. Community Ment Health J. (2023) 59:1442–51. 10.1007/s10597-023-01151-9

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 47.

  Shelton AJ Owens EW . Mental health services in the United States public high schools. J Sch Health. (2021) 91:70–6. 10.1111/josh.12976

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 48.

  Shormani MQ . Artificial intelligence contribution to translation industry: looking back and forward. arXiv preprint, arXiv:2411.19855. (2024). 10.48550/arXiv.2411.19855

  • CrossRef
  • Google Scholar

• 49.

  Sawers N Bolster NM Bastawrous A . The contribution of artificial intelligence in achieving the sustainable development goals (SDGs): what can eye health can learn from commercial industry and early lessons from the application of machine learning in eye health programmes. Front Public Health. (2021) 9:752049. 10.3389/fpubh.2021.752049

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 50.

  Wang TL Zeng YH Han YY Hong JH . How does AI technology affect the development of physical manufacturing industry under the background of intelligent economy. Trans Econ Bus Manag Res. (2023) 3:217–22. 10.62051/bkm82q24

  • CrossRef
  • Google Scholar

• 51.

  Abdul-Kader SAa Woods J . Survey on chatbot design techniques in speech conversation systems. Int J Adv Comput Sci Appl. (2015) 6:72–80. 10.14569/IJACSA.2015.060712

  • CrossRef
  • Google Scholar

• 52.

  Olawade DB Wada OZ Odetayo A David-Olawade AC Asaolu F Eberhardt J . Enhancing mental health with artificial intelligence: current trends and future prospects. J Med Surg Public Health. (2024) 3:100099. 10.1016/j.glmedi.2024.100099

  • CrossRef
  • Google Scholar

• 53.

  Siddals S Torous JB Coxon A . “It happened to be the perfect thing”: experiences of generative AI chatbots for mental health. NPJ Ment Health Res. (2024) 3:48. 10.1038/s44184-024-00097-4

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 54.

  Guo Y Hou CY . Application of artificial intelligence to mental healthcare in primary and middle schools. Chin J Sch Health. (2021) 42:1124–8. 10.16835/j.cnki.1000-9817.2021.08.002

  • CrossRef
  • Google Scholar

• 55.

  Chen TH Gascó-Hernández M Esteve M . The adoption and implementation of artificial intelligence chatbots in public organizations: evidence from U.S. state governments. Am Rev Public Adm. (2023) 54:255–70. 10.1177/02750740231200522

  • CrossRef
  • Google Scholar

• 56.

  You XY Zhang QY . Exploring the teaching mode of college students' mental health education under artificial intelligence technology guided by cognitive behavioral theory. Appl Math Nonlinear Sci. (2024) 9:1–17. 10.2478/amns-2024-0897

  • CrossRef
  • Google Scholar

• 57.

  Liu C Liu MH . “A review of the application of artificial intelligence in college mental health education.” In: 4th International Conference on Modern Education and Information Management (ICMEIM 2023), Wuhan, China (2023). 10.4108/eai.8-9-2023.2340096

  • CrossRef
  • Google Scholar

• 58.

  Dickey E Bejarano AM . “GAIDE: a framework for using generative AI to assist in course content development.” In: 2024 IEEE Frontiers in Education Conference (FIE) (2024). p. 1–9. 10.1109/FIE61694.2024.10893132

  • CrossRef
  • Google Scholar

• 59.

  Heo H Kang SC . Development of an online teacher education system for designing AI-integrated lessons. J Appl Instr Des. (2024) 13:45–8. 10.59668/1269.15642

  • CrossRef
  • Google Scholar

• 60.

  Sun JM Ma HL Zeng YY Han D Jin YB . Promoting the AI teaching competency of K-12 computer science teachers: a TPACK-based professional development approach. Educ Inf Technol. (2022) 28:1509–33. 10.1007/s10639-022-11256-5

  • CrossRef
  • Google Scholar

• 61.

  Ni AH Cheung A . Understanding secondary students' continuance intention to adopt AI-powered intelligent tutoring system for English learning. Educ Inf Technol. (2022) 28:3191–16. 10.1007/s10639-022-11305-z

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 62.

  Han JH Zhao W Jiang Q Oubibi M Hu XE . “Intelligent tutoring system trends 2006-2018: a literature review.” In: 2019 Eighth International Conference on Educational Innovation through Technology (EITT) (2019). p. 153–9. 10.1109/EITT.2019.00037

  • CrossRef
  • Google Scholar

• 63.

  Becerra Á Mohseni Z Sanz J Pérez RC . A generative AI-based personalized guidance tool for enhancing the feedback to MOOC learners.” In: 2024 IEEE Global Engineering Education Conference (EDUCON) (2024). p. 1–8. 10.1109/EDUCON60312.2024.10578809

  • CrossRef
  • Google Scholar

• 64.

  Qiu YT Khan MH Zhu SQ Chen SY Choonkit C . Enhancing sustainability in academic guidance: develop an AI-driven agent for education. 50 INTI J. (2024) 2024:1–10. 10.61453/INTIj.202440

  • CrossRef
  • Google Scholar

• 65.

  Lin YT Ye JH . Development of an educational chatbot system for enhancing students? Biology learning performance. J Internet Technol. (2023) 24:275–81. 10.53106/160792642023032402006

  • CrossRef
  • Google Scholar

• 66.

  Xu YQ Zhu JD Wang MK Qian F Yang YL Zhang J . The impact of a digital game-based AI chatbot on students' academic performance, higher-order thinking, and behavioral patterns in an information technology curriculum. Appl Sci. (2024) 14:6418. 10.3390/app14156418

  • CrossRef
  • Google Scholar

• 67.

  Banjade S Patel H Pokhrel S . Empowering education by developing and evaluating generative AI-powered tutoring system for enhanced student learning. J Artif Intell Capsule Netw. (2024) 6:278–98. 10.36548/jaicn.2024.3.003

  • CrossRef
  • Google Scholar

• 68.

  Niu SJ Luo JT Niemi H Li XQ Lu Y . Teachers' and students' views of using an AI-aided educational platform for supporting teaching and learning at Chinese schools. Educ Sci. (2022) 12:858. 10.3390/educsci12120858

  • CrossRef
  • Google Scholar

• 69.

  Weizenbaum J . ELIZA—a computer program for the study of natural language communication between man and machine. Commun ACM. (1966) 9:36–45. 10.1145/365153.365168

  • CrossRef
  • Google Scholar

• 70.

  Fitzpatrick KK Darcy AM Vierhile M . Delivering cognitive behavior therapy to young adults with symptoms of depression and anxiety using a fully automated conversational agent (Woebot): a randomized controlled trial. JMIR Mental Health. (2017) 4:e7785. 10.2196/mental.7785

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 71.

  Fulmer R Joerin A Gentile B Lakerink L Rauws M . Using psychological artificial intelligence (tess) to relieve symptoms of depression and anxiety: randomized controlled trial. JMIR Mental Health. (2018) 5:e64. 10.2196/mental.9782

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 72.

  Cameron G Cameron D Megaw G Bond RR Mulvenna MD O'neill S et al . Best Practices for Designing Chatbots in Mental Healthcare–A Case Study on iHelpr. British HCI Conference 2018, Belfast (2018). 10.14236/ewic/HCI2018.129

  • CrossRef
  • Google Scholar

• 73.

  M SG Nallasamy V . Artificial intelligence (AI) generated health counseling for mental illness patients. Curr Psychiatry Res Rev. (2024) 21:269–83. 10.2174/0126660822277500240109050359

  • CrossRef
  • Google Scholar

• 74.

  Vahedifard F Haghighi AS Dave T Tolouei M Zare FH . Practical use of ChatGPT in psychiatry for treatment plan and psychoeducation. arXiv:2311.09131. (2023). 10.48550/arXiv.2311.09131

  • CrossRef
  • Google Scholar

• 75.

  Eid MM Wang YD Mensah GB Pudasaini P . Treating psychological depression utilising artificial intelligence: AI for precision medicine- focus on procedures. Mesopotam J Artif Intell Healthc. (2023) 2023:76–81. 10.58496/MJAIH/2023/015

  • CrossRef
  • Google Scholar

• 76.

  Habicht J Viswanathan S Carrington B Hauser T Harper R Rollwage M . Closing the accessibility gap to mental health treatment with a conversational AI-enabled self-referral tool. Nat Med. (2024) 30:595–602. 10.1038/s41591-023-02766-x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 77.

  Wang Y Li SY . Tech vs. tradition: ChatGPT and mindfulness in enhancing older adults' emotional health. Behav Sci. (2024) 14:923. 10.3390/bs14100923

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 78.

  Mahmud TR Porntrakoon P . “The use of AI chatbots in mental healthcare for university students in Thailand: a case study.” In: 2023 7th International Conference on Business and Information Management (ICBIM) (2023). p. 48–53. 10.1109/ICBIM59872.2023.10303025

  • CrossRef
  • Google Scholar

• 79.

  Oghenekaro LU Okoro CO . Artificial intelligence-based chatbot for student mental health support. OALib. (2024) 11:1–14. 10.4236/oalib.1111511

  • CrossRef
  • Google Scholar

• 80.

  Klos MC Escoredo M Joerin A Lemos VN Rauws M Bunge EL . Artificial intelligence chatbot for anxiety and depression in university students: a pilot randomized controlled trial. JMIR Form Res. (2021) 5:e20678. 10.2196/20678

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 81.

  Liu H Peng HM Song XY Xu CZ Zhang M . Using AI chatbots to provide self-help depression interventions for university students: a randomized trial of effectiveness. Internet Interv. (2022) 27:100495. 10.1016/j.invent.2022.100495

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 82.

  Wang E Chang W-L Shen J Bian Q Huang PL Lai X et al . Effect of artificial intelligence on one-to-one emotional regulation and psychological intervention system of middle school students. Int J Neuropsychopharmacol. (2022) 25:A62–3. 10.1093/ijnp/pyac032.086

  • CrossRef
  • Google Scholar

• 83.

  Gao Y . The impact and application of artificial intelligence technology on mental health counseling services for college students. J Comput Methods Sci Eng. (2024) 25:1–18. 10.3233/JCM-247358

  • CrossRef
  • Google Scholar

• 84.

  Alojail M . UTAUT model for digital mental health interventions: factors influencing user adoption. Int J Adv Comput Sci Appl. (2024) 15:602–10. 10.14569/IJACSA.2024.0150860

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 85.

  Henkel T Linn AJ van der Goot MJ . Understanding the Intention to Use Mental Health Chatbots Among LGBTQIA+ Individuals: Testing and Extending the UTAUT. In: FølstadAAraujoTPapadopoulosSLawELCLugerEGoodwinMBrandtzaegPB editors. Chatbot Research and Design Cham (2023). 10.1007/978-3-031-25581-6_6

  • CrossRef
  • Google Scholar

• 86.

  Venkatesh V Morris MG Davis GB Davis FD . User acceptance of information technology: toward a unified view. Instit Transit Econ: Microecon. Issues eJ. (2003) 27:425–78. 10.2307/30036540

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 87.

  Venkatesh V Thong JYL Xu X . Consumer acceptance and use of information technology: extending the unified theory of acceptance and use of technology. Behav Mark eJ. (2012) 36:157–78. 10.2307/41410412

  • CrossRef
  • Google Scholar

• 88.

  Li SY Yuan QJ . UTAUT and its application and prospect in the field of information system research. J Mod Inf. (2020) 40:168–77. 10.3969/j.issn.1008-0821.2020.10.018

  • CrossRef
  • Google Scholar

• 89.

  Venkatesh V Davis FD Morris MG . Dead or alive? the development, trajectory and future of technology adoption research. J Assoc Inf Syst. (2007) 8:1. 10.17705/1jais.00120

  • CrossRef
  • Google Scholar

• 90.

  Huang CY Yang MC Chen IM Hsu WC . Modeling consumer adoption intention of an ai-powered health chatbot in taiwan: an empirical perspective. Int J Performability Eng. (2022) 18:338–49. 10.23940/ijpe.22.05.p4.338349

  • CrossRef
  • Google Scholar

• 91.

  Rumangkit S Surjandy S Billman A . The effect of performance expectancy, facilitating condition, effort expectancy, and perceived easy to use on intention to using media support learning based on unified theory of acceptance and use of technology (UTAUT). E3S Web of Conferences. (2023) 426:1–6. 10.1051/e3sconf/202342602004

  • CrossRef
  • Google Scholar

• 92.

  Duong CD Bui DT Pham HT Vu AT . How effort expectancy and performance expectancy interact to trigger higher education students' uses of ChatGPT for learning. Interact Technol Smart Educ. (2023) 21:356–80. 10.1108/ITSE-05-2023-0096

  • CrossRef
  • Google Scholar

• 93.

  Nurkhin A . Analysis of factors affecting behavioral intention to use e-learning uses the unified theory of acceptance and use of technology approach. KnE Soc Sci. (2020) 4:1005–25. 10.18502/kss.v4i6.6658

  • CrossRef
  • Google Scholar

• 94.

  Fadzil F . A study on factors affecting the behavioral intention to use mobile apps in Malaysia. SSRN Electron J. (2017). 10.2139/ssrn.3090753. [Epub ahead of print].

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 95.

  Moghavvemi S Paramanathan T Rahin NM Sharabati M . Student's perceptions towards using e-learning via Facebook. Behav Inf Technol. (2017) 36:1081–100. 10.1080/0144929X.2017.1347201

  • CrossRef
  • Google Scholar

• 96.

  Tarhini A El-Masri M Ali M Serrano A . Extending the UTAUT model to understand the customers' acceptance and use of internet banking in Lebanon: a structural equation modeling approach. Inf Technol People. (2016) 29:830–49. 10.1108/ITP-02-2014-0034

  • CrossRef
  • Google Scholar

• 97.

  Lipschitz JM Miller CJ Hogan TP Burdick KE Lippin-Foster R Simon SR et al . Adoption of mobile apps for depression and anxiety: cross-sectional survey study on patient interest and barriers to engagement. JMIR Mental Health. (2019) 6:2–13. 10.2196/11334

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 98.

  Wu WT Zhang B Li ST Liu HH . Exploring factors of the willingness to accept AI-assisted learning environments: an empirical investigation based on the UTAUT model and perceived risk theory. Front Psychol. (2022) 13:870777. 10.3389/fpsyg.2022.870777

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 99.

  Araujo T . Living up to the chatbot hype: the influence of anthropomorphic design cues and communicative agency framing on conversational agent and company perceptions. Comput Human Behav. (2018) 85:183–9. 10.1016/j.chb.2018.03.051

  • CrossRef
  • Google Scholar

• 100.

  Wutz M Hermes M Winter V Köberlein-Neu J . Factors influencing the acceptability, acceptance, and adoption of conversational agents in health care: integrative review. J Med Internet Res. (2023) 25:e46548. 10.2196/46548

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 101.

  Guo YH . “Moderating effects of gender in the acceptance of mobile SNS-based on UTAUT model.” In: 2014 International Conference on Management of e-Commerce and e-Government (2014). p. 163–7. 10.1109/ICMeCG.2014.41

  • CrossRef
  • Google Scholar

• 102.

  Liu KF Tao D . The roles of trust, personalization, loss of privacy, and anthropomorphism in public acceptance of smart healthcare services. Comput Human Behav. (2022) 127:107026. 10.1016/j.chb.2021.107026

  • CrossRef
  • Google Scholar

• 103.

  Liu Y Cao YQ . Influence of anthropomorphic characteristics of virtual chatbot on user acceptance. Chin J Ergon. (2023) 29:8–15. 10.13837/j.issn.1006-8309.2023.01.0002

  • CrossRef
  • Google Scholar

• 104.

  Qian DM . The facts and figures report of China's education development in 2020. ECNU Rev Educ. (2022) 5:540–60. 10.1177/20965311221113868

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 105.

  Wang DR . Leading a new leap in educational equipment with the new development concept. Educ Equip Res. (2021) 37:5.

  • Pubmed Abstract
  • Google Scholar

• 106.

  Zhoukou Bureau of Statistics . 2023 Population Development Report of Zhoukou City. Zhoukou Municipal Bureau of Statistics. Zhoukou Bureau of Statistics (2024). Available online at: https://www.zhoukou.gov.cn/page_pc/sjfb/fxjd/article9e9a8b57522d4e2d8f3a22d182c634aa.html?phlnglfkngdjmgdb

  • Google Scholar

• 107.

  Zhoukou Bureau of Statistics . Statistical Bulletin on the National Economy and Social Development of Zhoukou in 2022. Zhoukou Bureau of Statistics (2023). Available online at: http://tjj.zhoukou.gov.cn/sitesources/tjj/page_pc/tjfw/tjsj/tjgb/article29e05fb76f4e4f0f90064933542aaf07.html

  • Google Scholar

• 108.

  Cohen J . Statistical Power Analysis for the Behavioral Sciences (2nd ed.). Routledge (1988).

  • Google Scholar

• 109.

  Williams MD Rana NP Dwivedi YK . The unified theory of acceptance and use of technology (UTAUT): a literature review. J Enterp Inf Manag. (2015) 28:443–88. 10.1108/JEIM-09-2014-0088

  • CrossRef
  • Google Scholar

• 110.

  Huang F Teo T Zhou MM . Chinese students' intentions to use the Internet-based technology for learning. Educ Technol Res Dev. (2019) 68:575–91. 10.1007/s11423-019-09695-y

  • CrossRef
  • Google Scholar

• 111.

  Huang YM . Examining students' continued use of desktop services: perspectives from expectation-confirmation and social influence. Comput Human Behav. (2019) 96:23–31. 10.1016/j.chb.2019.02.010

  • CrossRef
  • Google Scholar

• 112.

  Waytz A Cacioppo J Epley N . Who sees human? The stability and importance of individual differences in anthropomorphism. Perspect Psychol Sci. (2010) 5:219–32. 10.1177/1745691610369336

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 113.

  De Visser EJ Monfort SS McKendrick R Smith MA McKnight PE Krueger F et al . Almost human: anthropomorphism increases trust resilience in cognitive agents. J Exp Psychol Appl. (2016) 22:331. 10.1037/xap0000092

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 114.

  Steel JL . Interpersonal correlates of trust and self-disclosure. Psychol Rep. (1991) 68:1319–20. 10.2466/pr0.1991.68.3c.1319

  • CrossRef
  • Google Scholar

• 115.

  Wheeless LR Grotz J . The measurement of trust and its relationship to self-disclosure. Hum Commun Res. (1977) 3:250–7. 10.1111/j.1468-2958.1977.tb00523.x

  • CrossRef
  • Google Scholar

• 116.

  Iloen ASN Kusumiati RYE . Social media dynamics: interpersonal trust and self-disclosure on instagram second accounts. Jurnal Bimbingan dan Konseling Terapan. (2024) 8:270–82. 10.30598/jbkt.v8i2.2015

  • CrossRef
  • Google Scholar

• 117.

  Li ZH Sha YZ Song XP Yang KH Zhao K Jiang ZX et al . Impact of risk perception on customer purchase behavior: a meta-analysis. J Bus Ind Mark. (2020) 35:76–96. 10.1108/JBIM-12-2018-0381

  • CrossRef
  • Google Scholar

• 118.

  Maciejewski D Lauharatanahirun N Herd T Lee J Deater-Deckard K King-Casas B et al . Neural cognitive control moderates the association between insular risk processing and risk-taking behaviors via perceived stress in adolescents. Dev Cogn Neurosci. (2018) 30:150–8. 10.1016/j.dcn.2018.02.005

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 119.

  Marusak HA Martin KR Etkin A Thomason ME . Childhood trauma exposure disrupts the automatic regulation of emotional processing. Neuropsychopharmacology. (2015) 40:1250–8. 10.1038/npp.2014.311

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 120.

  George MJ Odgers CL . Seven fears and the science of how mobile technologies may be influencing adolescents in the digital age. Perspect Psychol Sci. (2015) 10:832–51. 10.1177/1745691615596788

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 121.

  Goodwin J Savage E O'Donovan A . “I personally wouldn't know where to go”: Adolescents' perceptions of mental health services. J Adolesc Res. (2024) 39:1384–412. 10.1177/07435584221076056

  • CrossRef
  • Google Scholar

• 122.

  Smith KE Acevedo-Duran R Lovell JL Castillo AV Cardenas Pacheco . V. Youth are the experts! Youth participatory action research to address the adolescent mental health crisis. Healthcare. (2024) 12:592. 10.3390/healthcare12050592

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 123.

  Huang JL Chen J Zhou LY . Motivation crowding effects on the intention for continued use of gamified fitness apps: a mixed-methods approach. Front Psychol. (2024) 14:1286463. 10.3389/fpsyg.2023.1286463

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 124.

  Khatimah H Susanto P Abdullah NL . Hedonic motivation and social influence on behavioral intention of e-money: the role of payment habit as a mediator. Int J Entrep. (2019) 23:1–9.

  • Google Scholar

• 125.

  Prensky M . Digital natives, digital immigrants part 1. On the Horizon. (2001) 9:1–6. 10.1108/10748120110424816

  • CrossRef
  • Google Scholar

• 126.

  Petrescu MA Pop E-L Mihoc TD . Students' interest in knowledge acquisition in artificial intelligence. Procedia Comput Sci. (2023) 225:1028–36. 10.1016/j.procs.2023.10.090

  • CrossRef
  • Google Scholar

• 127.

  Evans A . ‘For the elections, we want women!': closing the gender gap in Zambian politics. Dev Change. (2016) 47:388–411. 10.1111/dech.12224

  • CrossRef
  • Google Scholar

• 128.

  Charlesworth TES Banaji MR . Patterns of implicit and explicit stereotypes III: long-term change in gender stereotypes. Soc Psychol Pers Sci. (2022) 13:14–26. 10.1177/1948550620988425

  • CrossRef
  • Google Scholar

• 129.

  Wang DF . Study on the communicative logics of the acquaintance society. J Yunnan Norm Univ. (2013) 45:79–85.

  • Google Scholar

• 130.

  He XF . On semi-aquaintance society—a perspective to understand village committee elections. CASS J Polit Sci. (2000) 36:61–9.

  • Google Scholar

• 131.

  Cheng HY Pan XF . Equality and excellence: system optimization of education streaming system in China. Educ Sci. (2023) 39:27–33. 10.1515/9781438492803-004

  • CrossRef
  • Google Scholar

• 132.

  Miao XJ Feng MY . Contradiction examination and path reflection on promotion of “optional subject and class” policy in new college entrance examination in past ten years. Teach Adm. (2024) 41:66–70.

  • Google Scholar

• 133.

  Lin MZ Cui XT Wang JQ Wu GH Lin JY . Promotors or inhibitors? Role of task type on the effect of humanoid service robots on consumers' use intention. J Hosp Mark Manag. (2022) 31:710–29. 10.1080/19368623.2022.2062693

  • CrossRef
  • Google Scholar

• 134.

  Kang E Kang YA . Counseling chatbot design: the effect of anthropomorphic chatbot characteristics on user self-disclosure and companionship. Int J Hum-Comput Int. (2024) 40:2781–95. 10.1080/10447318.2022.2163775

  • CrossRef
  • Google Scholar

• 135.

  Zhao CL Su CJ . An empirical study on rural primary and secondary school students' dialect identity. J Nanjing Norm Univ. (2022) 68:106–18. 10.3969/j.issn.1001-4608.2022.02.012

  • CrossRef
  • Google Scholar

• 136.

  Li L Peng W Rheu MMJ . Factors predicting intentions of adoption and continued use of artificial intelligence chatbots for mental health: examining the role of UTAUT model, stigma, privacy concerns, and artificial intelligence hesitancy. Telemedicine e-Health. (2024) 30:722–30. 10.1089/tmj.2023.0313

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 137.

  Kline RB . Principals and Practice of Structural Equation Modeling. New York, NY: Guilford (1998).

  • Google Scholar