Skip to main content
NCBI home page
Australian Journal of Psychology logoLink to Australian Journal of Psychology
. 2026 Sep 27;78(1):2737094. doi: 10.1080/00049530.2026.2737094

From boredom to problematic AI chatbot use: a three-wave half-longitudinal mediation study of loneliness and dual-system self-control

Jiahui Yuan a, Kai Zhou a,✉, Chen Zhang b
PMCID: PMC13618079  PMID: 42807895

ABSTRACT

Background

Problematic use of AI chatbots (PUAIC) is an emerging concern. Drawing on the I-PACE model and the dual-system model of self-control, this study examined the mediating roles of loneliness, the control system, and the impulsive system.

Methods

A three-wave half-longitudinal mediation design was used with 551 Chinese college students at baseline. Boredom proneness was assessed at Time 1, loneliness and the control and impulsive systems at Time 2, and PUAIC at Time 3. Missing data were handled using 20 multiply imputed datasets, and PROCESS Model 81 was estimated with sex and grade level as covariates.

Results

    Pooled analyses showed that boredom proneness was associated with loneliness (β = .57), the impulsive system (β = .31), the control system (β = −.38), and PUAIC directly (β = .25; all p < .001). Significant indirect effects occurred through loneliness (.073), the impulsive system (.077), and the serial loneliness–impulsive-system pathway (.038).

Conclusion

The findings suggest that boredom proneness may be associated with later PUAIC directly and indirectly through loneliness and impulsive tendencies. These associations may inform the development of prevention strategies for PUAIC that address loneliness and impulsive tendencies among college students.

KEYWORDS: Boredom proneness, problematic use of AI chatbots, loneliness, dual-system self-control, half-longitudinal mediation

KEY POINTS

What is already known about this topic:

  1. Boredom proneness is an established risk factor for problematic engagement with digital technologies and may increase reliance on online activities for stimulation and emotional relief.

  2. Loneliness and impaired self-control are closely associated with problematic technology use, while the impulsive and control systems represent two distinct pathways through which individual vulnerabilities may influence maladaptive digital behaviours.

  3. The I-PACE model proposes that person-level vulnerabilities interact with affective and execution-related processes in the development of problematic technology use.

What this topic adds:

  1. Using a three-wave half-longitudinal design, this study provides temporally ordered evidence that boredom proneness is associated with subsequent problematic use of AI chatbots among college students.

  2. By integrating the I-PACE model with the dual-system model of self-control, the study identifies loneliness and the impulsive system as key explanatory mechanisms, including a sequential pathway from boredom proneness to loneliness, heightened impulsive tendencies, and problematic AI chatbot use.

  3. The findings reveal an asymmetric role of the dual-system processes: the impulsive system significantly predicted problematic AI chatbot use, whereas the control system did not. This pattern identifies loneliness and impulsive tendencies as candidate targets for future evaluation without implying that control-system interventions are ineffective.

Introduction

Boredom proneness and problematic AI chatbot use

With the rapid integration of generative artificial intelligence into higher education, AI chatbots have become increasingly embedded in university students’ academic learning, writing, information seeking, problem solving, and daily decision-making (Badali et al., 2026; Ma et al., 2025). Although these tools offer considerable educational convenience, emerging research has raised concerns about their potentially maladaptive use. For example, AI chatbot use and dependence have been linked to poorer mental health outcomes among university students, including higher levels of depression and anxiety (Chavez Sosa & Huancahuire-Vega, 2026; Zhang et al., 2025, 2026). Beyond mental health risks, excessive reliance on generative AI may also undermine students’ critical thinking, writing proficiency, academic self-efficacy, autonomy, and self-regulated learning, while increasing the likelihood of academic dishonesty and cognitive offloading (Abbas et al., 2025; T. Wang et al., 2026; Williams, 2024). University students are a particularly important group to examine because they are frequent users of emerging technologies and are simultaneously facing academic pressure, identity exploration, and increasing demands for independent learning. Moreover, recent discussions of AI addiction have identified college students as a key population for understanding the boundary between adaptive academic use and problematic dependence on AI tools (Al-Obaydi & Pikhart, 2026). Therefore, investigating problematic use of AI chatbots (PUAIC) among university students is both theoretically and practically important. Such research can extend existing knowledge of technology-related behavioural dependence to the AI era and provide empirical evidence for developing responsible AI-use guidelines, mental health prevention strategies, and targeted interventions in higher education.

PUAIC generally refers to excessive or poorly controlled chatbot use characterised by psychological dependence, emotional compensation, impaired control, and negative consequences for daily functioning (Ciudad Fernández et al., 2025; Maral et al., 2026). A growing body of research suggests that PUAIC is shaped by both risk and protective factors. Risk factors include poor mental health, anxiety, depression, loneliness, low self-esteem, escapism, impulsivity, self-regulation difficulties, and pre-existing dependence on digital technologies such as the Internet and smartphones (Al-Obaydi & Pikhart, 2026; Shen et al., 2026; Yao et al., 2025; Zhang et al., 2026). In contrast, protective factors may include resilience, intentional and information-oriented use, social reconnection, and the use of AI chatbots as supportive tools rather than substitutes for offline relationships or autonomous coping (Carter, 2026; Lee, 2026; Zhang et al., 2026). At the same time, Ciudad Fernández et al. (2025) cautioned that labelling frequent ChatGPT use as an addiction may pathologize intensive engagement without demonstrating impaired control or functional harm. Accordingly, the present study uses PUAIC as an operational, dimensional construct rather than as a formally recognised addictive disorder. Frequency or duration alone is not treated as evidence of problematic use; the construct instead emphasises impaired control, emotional dependence, and negative consequences.

I-PACE model as the theoretical framework

The Interaction of Person–Affect–Cognition–Execution (I-PACE) model provides a process-oriented framework for explaining the development of problematic technology use. It proposes that such behaviour arises from dynamic interactions among relatively stable person-level predispositions, affective and cognitive responses to internal or external triggers, and execution-related processes such as inhibitory control and decision-making (Brand et al., 2019). By distinguishing distal vulnerabilities from proximal psychological and regulatory mechanisms, the model explains why similar technology-related experiences may lead to problematic use among some individuals but not others.

In the present study, the placement of the study variables within the I-PACE framework was guided by both the model’s specifications and construct-specific research. Boredom proneness is positioned as a person-level predisposition (Tam et al., 2021), loneliness as a proximal affective pathway (Kardefelt-Winther, 2014), and the control and impulsive systems as execution-related regulatory and action-tendency processes (Hofmann et al., 2009). PUAIC represents the subsequent behavioural outcome. This mapping organises the proposed associations into a person–affect–execution sequence and provides the theoretical basis for the pathways examined in the following sections. However, the present study does not operationalise all I-PACE components, such as specific cognitive biases, use-related expectancies, cue reactivity, craving, or reinforcement processes; it therefore tests a focused person–affect–execution pathway rather than the complete model.

Boredom proneness as a person-level vulnerability

PUAIC should be distinguished from frequent but functional use because it involves maladaptive engagement characterised by emotional reliance, impaired self-regulation, and compensatory or escapist motivations (Latikka et al., 2026; Zhai et al., 2025). Within this framework, boredom proneness warrants particular attention as a person-level vulnerability. As a relatively stable tendency to experience boredom more frequently or intensely and to perceive life as insufficiently engaging (Tam et al., 2021), boredom proneness may increase individuals’ preference for immediate stimulation, emotional relief, and low-effort engagement, thereby heightening their susceptibility to PUAIC.

Although direct empirical evidence linking boredom proneness specifically to PUAIC remains limited, research in related technological contexts consistently associates boredom with maladaptive patterns of use. A recent systematic review identified trait boredom as both a direct and indirect contributor to problematic smartphone, Internet, social media, and gaming use (Tagliaferri et al., 2025). In the context of generative AI, Zhong et al. (2025) found that boredom was associated with college students’ willingness to use generative AI through perceived usefulness, perceived ease of use, and novelty experience, providing preliminary evidence that boredom may shape AI-related engagement. From the perspective of compensatory Internet use theory, AI chatbot dependence may arise not only from instrumental needs but also from attempts to alleviate aversive states such as boredom, loneliness, and anxiety (Zhai et al., 2025). Research on social chatbots has similarly linked chatbot engagement with psychological distress and loneliness, suggesting that individuals with unmet emotional or social needs may be particularly vulnerable to excessive use (Latikka et al., 2026). Taken together, these findings suggest that boredom proneness may contribute to PUAIC directly by increasing the motivation to seek frequent and stimulating AI interaction and indirectly through associated affective and self-regulatory vulnerabilities. Nevertheless, because much of the existing evidence concerns adjacent forms of problematic technology use or general AI adoption rather than PUAIC itself, the role of boredom proneness in problematic AI chatbot use requires direct empirical examination.

Loneliness as an affective pathway

Loneliness may serve as a key psychological mechanism linking boredom proneness to PUAIC. Boredom proneness reflects a stable tendency to experience insufficient stimulation and difficulty maintaining meaningful engagement (Tam et al., 2021), which may increase individuals’ vulnerability to negative affective states and unmet social needs. Prior research on problematic technology use suggests that boredom and loneliness are closely intertwined and may jointly motivate compensatory digital engagement, in which individuals use online technologies to alleviate aversive internal states (Błachnio et al., 2026; Tagliaferri et al., 2025). In this sense, boredom-prone individuals may be more likely to experience loneliness because they often struggle to obtain satisfying stimulation and social connection from offline contexts. Loneliness, in turn, may directly increase the risk of PUAIC. AI chatbots can provide immediate responses, emotional comfort, perceived companionship, and a low-risk space for self-disclosure, making them especially attractive to individuals who feel socially disconnected (Hu et al., 2023; Huang et al., 2026). However, although chatbot interaction may temporarily relieve loneliness, excessive reliance on AI-based companionship may reinforce emotional dependence, reduce motivation for offline social interaction, and contribute to maladaptive or problematic use (Adewale & Muhammad, 2025; Crawford et al., 2024; Liu et al., 2025). Consistent with compensatory Internet use theory, lonely individuals may turn to conversational AI as a coping strategy for unmet relational needs, but this compensatory use may become problematic when it replaces rather than supports real interpersonal connection (Hu et al., 2023; Huang et al., 2026). Therefore, loneliness provides a theoretically meaningful mediating pathway through which boredom proneness may contribute to problematic AI chatbot use: boredom proneness may heighten loneliness, and loneliness may subsequently drive excessive or dependent engagement with AI chatbots.

Dual-system self-control as an execution-related pathway

The dual-system model of self-control provides an execution-related explanation for the association between boredom proneness and PUAIC. According to this perspective, self-control-related behaviour is shaped by the balance between a reflective control system, which supports deliberate regulation and long-term goal pursuit, and an impulsive system, which favours immediate gratification and reward-seeking responses (Hofmann et al., 2009). Boredom proneness may disrupt this balance by increasing the desire for immediate stimulation and emotional relief, thereby strengthening impulsive tendencies while weakening reflective control. Research on problematic technology use has identified boredom as an internal trigger that can activate impulsive responses and promote immediate-gratification behaviours in digital environments (Qahri-Saremi & Turel, 2023; Tagliaferri et al., 2025). This mechanism may be particularly relevant to AI chatbot use because chatbots provide instant feedback, personalised responses, emotional support, and low-effort interaction, making continued engagement highly accessible and potentially difficult to regulate. Research has consistently associated problematic digital technology use with an imbalance between strong impulsive tendencies and weak regulatory control (Turel & Qahri-Saremi, 2016; Zahrai et al., 2022). Emerging evidence in AI-related contexts similarly suggests that self-control deficits increase vulnerability to PUAIC, whereas stronger self-control may protect against maladaptive engagement (Maral et al., 2026; Prayoga & Wakhid, 2024). AI chatbot dependence may also involve both efficiency-oriented reliance and emotion-driven attachment, indicating that deliberative and affective processes can coexist in AI engagement (Huang et al., 2026; Zhai et al., 2025). Accordingly, the control and impulsive systems may mediate the relationship between boredom proneness and PUAIC: boredom proneness may weaken regulatory control and intensify impulsive motivation, thereby increasing the likelihood of excessive or dependent AI chatbot use.

Loneliness is closely related to the dual-system model of self-control because it may both reflect and intensify self-regulatory imbalance. Specifically, low self-control can increase loneliness by making it more difficult for individuals to inhibit selfish or impulsive behaviours and maintain positive social bonds, whereas loneliness may further weaken executive control and activate impulsive, emotion-driven responses to cope with unmet social needs (Sin et al., 2021; Stavrova et al., 2022). From a dual-system perspective, loneliness can therefore be understood as an internal emotional cue that strengthens the impulsive system while placing greater demands on the control system, increasing vulnerability to maladaptive behaviours such as problematic technology use (Turel & Qahri-Saremi, 2016). The hypothesised sequence from loneliness to self-control was guided by the I-PACE model, which conceptualises affective responses as proximal factors that may influence subsequent self-regulatory processes. However, because loneliness and self-control were both assessed at Time 2, their temporal ordering cannot be established empirically, and reciprocal associations remain possible. Thus, the proposed ordering should be interpreted as theoretically informed rather than temporally demonstrated.

Present study and hypotheses

Despite the growing scholarly attention to PUAIC, several important gaps remain. First, existing research has mainly approached AI chatbot use from the perspectives of technology acceptance, learning efficiency, or general psychological outcomes, whereas relatively limited attention has been paid to the psychological risk mechanisms through which stable individual vulnerabilities may contribute to PUAIC. In particular, although boredom proneness has been identified as an important antecedent of problematic digital technology use, direct empirical evidence linking boredom proneness to PUAIC remains scarce. Second, most prior studies on problematic technology use and AI chatbot engagement have relied on cross-sectional designs, which limits the ability to clarify temporal ordering among psychological risk factors, mediating mechanisms, and later problematic use. Third, previous studies have seldom integrated affective and self-regulatory mechanisms within a unified temporally ordered framework. Loneliness may explain how boredom-prone individuals turn to AI chatbots for emotional compensation, whereas the dual-system model of self-control may further clarify why such use becomes difficult to regulate when impulsive tendencies are strengthened and reflective control is weakened. Therefore, a half-longitudinal examination of loneliness, control system, and impulsive system as mediating mechanisms may provide a more comprehensive understanding of how boredom proneness contributes to PUAIC.

To address these gaps, the present study employed a three-wave half-longitudinal mediation design to examine whether boredom proneness at Time 1 was associated with PUAIC at Time 3, and whether this association was mediated by loneliness and the dual-system components of self-control at Time 2 (see Figure 1). By incorporating both affective and self-regulatory pathways, this study aimed to extend the literature on problematic technology use to the emerging context of AI chatbot use and to identify psychological mechanisms that may increase university students’ vulnerability to PUAIC. Accordingly, the following hypotheses were proposed:

Figure 1.

Flowchart: boredom, loneliness, control, impulsivity, AI chatbot use relationship.

Open in a new tab

The hypothesized model.

Hypothesis 1.

Boredom proneness at Time 1 will be positively associated with PUAIC at Time 3.

Hypothesis 2.

Loneliness at Time 2 will mediate the temporally ordered association between boredom proneness at Time 1 and PUAIC at Time 3.

Hypothesis 3a.

The control system at Time 2 will mediate the temporally ordered association between boredom proneness at Time 1 and PUAIC at Time 3.

Hypothesis 3b.

The impulsive system at Time 2 will mediate the temporally ordered association between boredom proneness at Time 1 and PUAIC at Time 3.

Hypothesis 4.

Loneliness and each dual-system component of self-control will form serial mediating pathways between boredom proneness and PUAIC.

Methods

Participants and procedure

Given the growing prevalence of AI chatbot use among university students, participants were recruited from two universities in China using a convenience sampling method. A three-wave half-longitudinal design was implemented across the fall semester of 2024, with approximately 30 days between consecutive waves. At Wave 1, conducted in October 2024, 551 students completed measures of demographic characteristics and boredom proneness. The baseline sample consisted of 57.5% female students, with a mean age of 19.02 years (SD = 1.17). At Wave 2, conducted in November 2024, 533 participants completed measures of loneliness and the dual-system model of self-control. At Wave 3, conducted in December 2024, 521 participants provided valid responses on problematic AI chatbot use and reported their AI chatbot usage patterns. At the final wave, participants were also asked to identify the AI chatbot they used most frequently. Complete data across all three waves were available for 503 participants. Results showed that 68.2% of participants primarily used Doubao, 21.9% reported using DeepSeek, and 9.9% used other AI tools, including Yuanbao, Qianwen, ChatGPT, and Gemini. Participants reported using AI chatbots for an average of 2.98 hours per week (SD = 1.05).

The study protocol was reviewed and approved by the ethics committee of the researchers’ institution. All participants took part voluntarily and were informed that their responses would remain confidential and that they could withdraw at any point without penalty. Participants generated a self-created matching code based on the last four digits of their mobile phone number. The code was used exclusively to match responses across the three waves, and the analytical dataset contained no direct personal identifiers. The three-wave design, conducted over one academic semester, enabled the examination of temporally ordered associations among the study variables.

Measures

Boredom proneness

Boredom proneness was measured using the Chinese short form of the Boredom Proneness Scale (Li et al., 2016; Vodanovich et al., 2005). The scale contains 12 items assessing insufficient internal and external stimulation. Participants rated each item on a 7-point Likert scale, ranging from 1 (strongly disagree) to 7 (strongly agree). Higher mean scores reflected greater boredom proneness. The scale demonstrated good internal consistency (Cronbach’s α = .83; McDonald’s ω = .84). A sample item was “I often feel that my surroundings are monotonous and boring”.

Dual-system self-control

Dual-system self-control at T2 was assessed using the Dual-System of Self-Control Scale developed by Xie et al. (2014). The instrument consists of a 9-item control-system subscale and a 12-item impulsive-system subscale. Participants rated all items on a 5-point Likert scale, ranging from 1 (strongly disagree) to 5 (strongly agree). Higher control-system scores indicated greater self-regulatory capacity, whereas higher impulsive-system scores reflected stronger impulsive tendencies. Both subscales demonstrated good internal consistency (control system: Cronbach’s α = .85, McDonald’s ω = .86; impulsive system: Cronbach’s α = .86, McDonald’s ω = .87). Sample items for the impulsive and control systems were “I often act or speak without thinking” and “I believe we should plan our day in the morning”, respectively.

Loneliness

Loneliness was assessed using the Chinese short-form version of the UCLA Loneliness Scale, originally developed by Hays and DiMatteo (1987), following the Chinese-context use reported by K. Yu and Zheng (2020). This unidimensional measure consists of eight items rated on a 4-point scale from 1 (never) to 4 (always), with higher scores indicating greater loneliness. The scale demonstrated good internal consistency (Cronbach’s α = .86; McDonald’s ω = .87). A sample item was “I lack companionship”.

Problematic AI chatbot use

Problematic AI chatbot use was measured with an adapted version of the Problematic ChatGPT Use Scale developed by S. C. Yu et al. (2024). To make the scale applicable to a broader range of AI chatbot platforms, “ChatGPT” was replaced with “AI chatbot” in the item wording. The 11 items, including “I progressively spend more time using AI chatbots”, were rated on a 5-point Likert scale; higher scores indicated more problematic use. The adapted scale demonstrated good internal consistency (Cronbach’s α = .89; McDonald’s ω = .90). Confirmatory factor analysis indicated acceptable fit (CFI = .95, TLI = .94, RMSEA = .062, SRMR = .047).

Data analyses

All statistical analyses were conducted using SPSS 27.0 and the PROCESS macro. Little’s MCAR test (Little, 1988) was conducted for boredom proneness, loneliness, the impulsive system, the control system, and PUAIC. The result was nonsignificant, χ2(6) = 4.22, p = .647, indicating that the observed missing-data pattern was consistent with MCAR. Missing values on the study variables were handled using multiple imputation by chained equations with predictive mean matching. Twenty imputed datasets were generated. PROCESS Model 81 was estimated separately in each completed dataset, and the resulting regression parameters were pooled using Rubin’s rules. Boredom proneness at T1 was entered as the predictor, PUAIC at T3 as the outcome, and loneliness and the two self-control systems at T2 as mediators; sex and grade level were included as covariates. Weekly AI-chatbot usage time was not included as a covariate because it was measured concurrently with PUAIC at T3 and may partly reflect the behavioural manifestation of the outcome; adjusting for it could therefore constitute overcontrol. Indirect effects were evaluated using 5,000 bootstrap resamples within each completed dataset, and the resulting estimates and confidence intervals were summarised across the 20 analyses. Ordinary pooled 95% confidence intervals were reported for regression coefficients. Harman’s single-factor test showed that the first unrotated factor explained 21.92% of the variance, below the commonly used 40% criterion (Tang & Wen, 2020).

Assumption checks indicated no substantial violations. Skewness values ranged from −0.04 to 0.28 and kurtosis values ranged from −0.42 to 1.43, suggesting no substantial departures from normality. VIF values ranged from 1.01 to 1.75, and tolerance values ranged from .57 to .99, indicating no serious multicollinearity. Although some observations showed relatively large standardised residuals or Mahalanobis distances, Cook’s distance values were all below .08, suggesting that no individual case exerted undue influence on the regression estimates. For loneliness, 2.3% of observations were at the minimum scale value, indicating no substantial floor effect.

Results

Descriptive and correlation analysis

As shown in Table 1, boredom proneness at T1 was positively associated with loneliness, impulsive system, and PUAIC, but negatively associated with control system. Loneliness was positively related to impulsive system and PUAIC, and negatively related to control system. In addition, impulsive system was positively associated with PUAIC, whereas control system was negatively associated with PUAIC.

Table 1.

Summary statistics and bivariate correlations.

  1 2 3 4 5
1. Boredom proneness T1 1        
2. Loneliness T2 .57** 1      
3. Impulsive system T2 .46** .44** 1    
4. Control system T2 −.42** −.29** −.33** 1  
5. PUAIC T3 .39** .36** .39** −.15** 1
M (SD) 3.62 (0.90) 2.07 (0.60) 2.39 (0.66) 3.55 (0.57) 2.66 (0.77)
Open in a new tab

Note: **p < .01. PUAIC = problematic use of AI chatbots.

Testing the multiple mediation model

As reported in Figure 2 and Table 2, boredom proneness was positively associated with loneliness (β = .57, p < .001) and the impulsive system (β = .31, p < .001), and negatively associated with the control system (β = −.38, p < .001). Loneliness was positively associated with the impulsive system (β = .27, p < .001), but not with the control system (β = −.06, p = .244). In the model predicting PUAIC, boredom proneness (β = .25, p < .001), loneliness (β = .13, p = .010), and the impulsive system (β = .24, p < .001) were significant positive predictors, whereas the control system was not significant (β = .08, p = .072). The models explained 33% of the variance in loneliness, 27% in the impulsive system, 17% in the control system, and 23% in PUAIC.

Figure 2.

Diagram: boredom, loneliness, impulsiveness, control, AI chatbot use relationships. The diagram illustrates the relationships between various psychological factors and the problematic use of AI chatbots. Boredom proneness T1 is positively associated with loneliness T2 (0.57) and the impulsive system T2 (0.31) and negatively associated with the control system T2 (-0.38). Loneliness T2 is positively associated with the impulsive system T2 (0.27) and problematic use of AI chatbots T3 (0.13), but not significantly with the control system T2 (-0.06). The impulsive system T2 is positively associated with problematic use of AI chatbots T3 (0.24). The control system T2 has a non-significant positive association with problematic use of AI chatbots T3 (0.08). Boredom proneness T1 directly influences problematic use of AI chatbots T3 (0.25).

Open in a new tab

The multiple mediation model.

Note. Values are standardised regression coefficients pooled across 20 imputed datasets (N = 551). Sex and grade level were included as covariates. Dashed lines indicate nonsignificant paths. *p < .05, **p < .01, ***p < .001.

Table 2.

Multiple mediation model analysis results.

Outcomes Predictors R2 F β SE 95% CI
Loneliness   .33 F(3, 519.02) = 86.22***      
  Boredom proneness     .57*** 0.04 [.50, .64]
Impulsive system   .27 F(4, 522.52) = 48.55***      
  Boredom proneness     .31*** 0.05 [.22, .40]
  Loneliness     .27*** 0.05 [.18, .36]
Control system   .17 F(4, 509.90) = 27.25***      
  Boredom proneness     −.38*** 0.05 [−.48, −.29]
  Loneliness     −.06 0.05 [−.16, .04]
PUAIC   .23 F(6, 486.24) = 24.13***      
  Boredom proneness     .25*** 0.05 [.14, .35]
  Loneliness     .13* 0.05 [.03, .23]
  Impulsive system     .24*** 0.05 [.15, .34]
  Control system     .08 0.04 [−.01, .16]
Open in a new tab

Note: Values are standardized regression coefficients pooled across 20 multiply imputed datasets using Rubin’s rules. Sex and grade were included as covariates in all models but are not displayed. R2 values were pooled using Fisher’s z transformation, and the reported F statistics are pooled Wald tests; therefore, denominator degrees of freedom are fractional. *p < .05, **p < .01, ***p < .001. PUAIC = problematic use of AI chatbots. Confidence intervals are ordinary pooled 95% CIs.

As shown in Table 3, the total indirect effect of boredom proneness on PUAIC was significant (effect = .156, SE = .03, 95% CI [.101, .237]). Significant specific indirect effects occurred through loneliness (effect = .073, SE = .03, 95% CI [.033, .142]), the impulsive system (effect = .077, SE = .02, 95% CI [.040, .117]), and the serial pathway through loneliness and the impulsive system (effect = .038, SE = .01, 95% CI [.019, .057]). Indirect effects involving the control system were nonsignificant because their confidence intervals included zero. The total effect was also significant (effect = .401, SE = .04, 95% CI [.314, .471]).

Table 3.

Results of multiple mediation effects.

Model paths Effect value SE 95% CI
Boredom proneness→Loneliness→PUAIC .073 0.03 [.033, .142]
Boredom proneness→Control system→PUAIC −.030 0.02 [−.061, .004]
Boredom proneness→Impulsive system→PUAIC .077 0.02 [.040, .117]
Boredom proneness→Loneliness→Control system→PUAIC −.003 0.01 [−.009, .002]
Boredom proneness→Loneliness→Impulsive system→PUAIC .038 0.01 [.019, .057]
Total indirect effects .156 0.03 [.101, .237]
Total effect .401 0.04 [.314, .471]
Open in a new tab

Note: PUAIC = problematic use of AI chatbots. Indirect-effect intervals are 95% bootstrap CIs based on 5,000 resamples within each imputed dataset and summarized across 20 imputed datasets. The CI for the total effect is an ordinary pooled 95% CI.

Sensitivity analyses

A strict complete-case analysis (n = 503) yielded the same substantive conclusions as the primary multiple-imputation analysis. The direct effect of boredom proneness on PUAIC remained significant (β = .243, p < .001), as did the total indirect effect (effect = .154, SE = .036, 95% CI [.084, .227]). The indirect effects through loneliness, the impulsive system, and the loneliness–impulsive-system serial pathway remained significant, whereas effects involving the control system remained nonsignificant.

In the reverse-order model, the impulsive system predicted loneliness (β = .221, p < .001), whereas the control system did not (β = −.011, p = .777). The reverse serial indirect effect through the impulsive system and loneliness was significant (effect = .013, 95% CI [.004, .028]), but the corresponding control-system pathway was not (effect = .001, 95% CI [−.004, .006]). The indirect effect through the impulsive system was significantly stronger than that through the control system (contrast = .107, SE = .028, 95% CI [.054, .163]). The sensitivity analyses supported the robustness of the associations involving loneliness and the impulsive system, but the significant reverse-order pathway indicated that their directional ordering could not be established. All sensitivity models controlled for sex and grade level and used 5,000 bootstrap resamples.

Discussion

Guided primarily by the I-PACE model (Brand et al., 2019), with compensatory Internet use theory and the dual-system model of self-control as complementary accounts (Hofmann et al., 2009; Kardefelt-Winther, 2014), the present study used a three-wave half-longitudinal mediation design to examine temporally ordered associations between boredom proneness and PUAIC among university students. Boredom proneness was associated with later PUAIC, supporting Hypothesis 1. Loneliness and the impulsive system showed significant indirect effects, supporting Hypotheses 2 and 3b, whereas the control-system indirect effect was not significant and Hypothesis 3a was not supported. The serial pathway through loneliness and the impulsive system was significant, partially supporting Hypothesis 4. In I-PACE terms, the findings are consistent with a focused pathway from a person-level predisposition (boredom proneness), through an affective response (loneliness) and an execution-related impulsive tendency, to later problematic behaviour. Because baseline levels were unavailable, these results indicate temporal ordering but not change or causality.

Consistent with Hypothesis 1, boredom proneness was positively associated with subsequent PUAIC. This finding is in line with prior studies suggesting that boredom is an important risk factor for maladaptive engagement with digital technologies, including problematic smartphone use, excessive Internet use, social media overuse, and gaming-related addictive behaviours (Tagliaferri et al., 2025). The result also echoes emerging evidence that boredom may increase students’ willingness to use generative AI by enhancing perceived usefulness, perceived ease of use, and novelty experience (Zhong et al., 2025). From the perspective of compensatory Internet use theory (Kardefelt-Winther, 2014), individuals with high boredom proneness may be more likely to turn to AI chatbots as an easily accessible source of stimulation, distraction, and emotional relief. Compared with many traditional digital tools, AI chatbots provide immediate feedback, personalised responses, low-effort interaction, and a sense of conversational engagement (Boyd & Markowitz, 2026; Han, 2025). These features may be especially attractive to boredom-prone students who often experience insufficient internal and external stimulation (Camerini et al., 2023). Therefore, the direct association between boredom proneness and PUAIC suggests that AI chatbot use may become problematic when it is repeatedly used as a convenient strategy to escape boredom or compensate for a lack of meaningful engagement in offline life.

The mediation findings further clarify the psychological pathways through which boredom proneness contributes to PUAIC. First, loneliness significantly mediated the association between boredom proneness and PUAIC, supporting Hypothesis 2. This result is consistent with previous research suggesting that boredom and loneliness are closely intertwined and may jointly motivate compensatory digital engagement (Błachnio et al., 2026; Tagliaferri et al., 2025). Boredom-prone individuals may have difficulty obtaining meaningful stimulation and satisfying social connection from their offline environment (Tagliaferri et al., 2025), which may increase their feelings of loneliness. In turn, lonely students may be more inclined to interact with AI chatbots because such tools can provide instant responses, emotional comfort, perceived companionship, and a low-risk space for self-disclosure (Hu et al., 2023; Huang et al., 2026). Second, the impulsive system, but not the control system, significantly mediated the relationship between boredom proneness and PUAIC. Thus, Hypothesis 3b was supported, whereas Hypothesis 3a was not supported. This pattern suggests that the risk pathway from boredom proneness to PUAIC may be more strongly driven by heightened impulsive motivation than by weakened reflective control. In other words, students with higher boredom proneness may be especially motivated to seek immediate gratification and effortless emotional relief from AI chatbot interaction. However, the nonsignificant coefficient does not provide evidence that the control system was directly associated with PUAIC once loneliness and impulsive tendencies were considered simultaneously. Finally, Hypothesis 4 was partially supported: the serial pathway through loneliness and the impulsive system was significant, whereas the serial pathway involving the control system was not. This finding suggests that loneliness may function as an emotional trigger that further activates impulsive, reward-seeking responses (Y. Wang et al., 2023), thereby increasing vulnerability to PUAIC. This interpretation is consistent with the dual-system perspective, which emphasises that maladaptive technology use is more likely to occur when immediate affective needs and impulsive tendencies override long-term self-regulatory goals (Hofmann et al., 2009; Turel & Qahri-Saremi, 2016; Zahrai et al., 2022). Notably, the control system correlated negatively with PUAIC at the bivariate level (r = −.15) but became positive and nonsignificant in the multivariate model (β = .08). This adjustment-induced sign reversal may reflect statistical suppression or predictor overlap; however, because the adjusted coefficient was nonsignificant, it should not be interpreted as evidence of either a direct protective or a direct harmful effect.

Implications

Theoretically, this study identifies boredom proneness as a temporally ordered risk correlate for PUAIC and highlights affective and impulsive mechanisms in this process. This finding is consistent with evidence that boredom may promote problematic digital media use by motivating individuals to seek immediate stimulation (Camerini et al., 2023). It also supports the I-PACE model, which emphasises the combined roles of personal predispositions, affective responses, and executive processes in problematic technology use (Brand et al., 2019). Practically, potential prevention efforts may consider addressing not only excessive chatbot use but also its underlying psychological drivers. Students may benefit from greater boredom tolerance, emotional awareness, offline social connection, and adaptive coping with loneliness. Universities should integrate mental health education and digital literacy to help students distinguish productive AI-assisted learning from compulsive or emotionally dependent use. This is especially relevant because loneliness may help explain the use of conversational AI for companionship among psychologically distressed students (Lai et al., 2025). Universities could also provide engaging activities, peer interaction, and counselling support. In addition, AI providers should adopt responsible design features, including usage reminders, reflective prompts, well-being feedback, and referrals to psychological support when emotional dependence is detected (Coghlan et al., 2023). Preventing PUAIC therefore requires coordinated efforts by students, universities, and AI service providers to promote balanced and psychologically healthy AI use. The nonsignificant control-system pathway does not demonstrate that control-focused interventions are ineffective, because the study was observational and did not test any intervention. Practical recommendations should therefore be treated as tentative targets for future experimental evaluation.

Limitations and future directions

First, this was a half-longitudinal mediation design: boredom proneness, the mediators, and PUAIC were each measured at only one wave. Baseline loneliness, control and impulsive systems, and PUAIC were unavailable, so autoregressive stability could not be controlled and pre-existing PUAIC may have contributed to the observed associations. Because loneliness and both self-control systems were measured concurrently at Time 2, the proposed serial ordering is theoretical rather than temporally established. These constraints preclude causal and within-person change conclusions; future studies should measure every construct at every wave and use cross-lagged or random-intercept longitudinal models. Second, the approximately 30-day intervals were short and may not capture longer-term development. Third, reliance on self-report measures may have introduced common-method and social-desirability bias; behavioural usage logs and functional-impairment indicators should be incorporated. Fourth, convenience sampling from two Chinese universities limits generalisability. Fifth, future work should distinguish academic, recreational, supportive, and emotionally dependent chatbot use and examine mechanisms such as FoMO, academic stress, social anxiety, emotion regulation, anthropomorphism, and emotional attachment.

Conclusion

This three-wave half-longitudinal study found temporally ordered associations linking boredom proneness with later PUAIC directly and indirectly through loneliness and heightened impulsive tendencies; the indirect effect through the control system was not supported. The pattern is consistent with a focused I-PACE pathway involving person-level vulnerability, affective response, and execution-related impulsivity. However, the absence of autoregressive controls prevents causal or change-over-time conclusions. Interventions addressing loneliness, boredom coping, and impulsive engagement warrant experimental evaluation rather than being inferred as effective from these observational results.

Acknowledgements

The authors sincerely thank all individuals who participated in this research.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

Ethical approval

The study procedures complied with the Declaration of Helsinki and were approved by the Institutional Review Board of the authors’ institution (Approval No. 2,024,027).

Informed consent statement

All participants provided informed consent before taking part in the study.

References

  1. Abbas, M., Khan, T. I., & Jam, F. A. (2025). Avoid excessive usage: Examining the motivations and outcomes of generative artificial intelligence usage among students. Journal of Academic Ethics, 23(4), 2423–13. 10.1007/s10805-025-09659-3 [DOI] [Google Scholar]
  2. Adewale, M. D., & Muhammad, U. I. (2025). From virtual companions to forbidden attractions: The seductive rise of artificial intelligence love, loneliness, and intimacy-A systematic review. Journal of Technology in Behavioral Science. Advance online publication. 10.1007/s41347-025-00549-4 [DOI] [Google Scholar]
  3. Al-Obaydi, L. H., & Pikhart, M. (2026). Artificial intelligence addiction: Exploring the emerging phenomenon of addiction in the AI age. AI & Society, 41(2), 1577–1593. 10.1007/s00146-025-02535-z [DOI] [Google Scholar]
  4. Badali, M., Shahalizadeh, M., Alimirzaie, M., Noroozi, O., & Kazem Banihashem, S. (2026). Bridging AI chatbot use and critical thinking: The mediating role of AI literacy. Interactive Learning Environments, 34(7), 1–13. 10.1080/10494820.2025.2605487 [DOI] [Google Scholar]
  5. Błachnio, A., Przepiórka, A., Cudo, A., Kot, P., & Gorbaniuk, J. (2026). The role of affect balance and external boredom in the relationship between loneliness and problematic smartphone use. Innovation: The European Journal of Social Science Research. Advance online publication. 1–20. 10.1080/13511610.2026.2618079 [DOI] [Google Scholar]
  6. Boyd, R. L., & Markowitz, D. M. (2026). Artificial intelligence and the psychology of human connection. Perspectives on Psychological Science, 21(2), 192–220. 10.1177/17456916251404394 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Brand, M., Wegmann, E., Stark, R., Müller, A., Wölfling, K., Robbins, T. W., & Potenza, M. N. (2019). The interaction of person-affect-cognition-execution (I-PACE) model for addictive behaviors: Update, generalization to addictive behaviors beyond Internet-use disorders, and specification of the process character of addictive behaviors. Neuroscience and Biobehavioral Reviews, 104, 1–10. 10.1016/j.neubiorev.2019.06.032 [DOI] [PubMed] [Google Scholar]
  8. Camerini, A.-L., Morlino, S., & Marciano, L. (2023). Boredom and digital media use: A systematic review and meta-analysis. Computers in Human Behavior Reports, 11, Article 100313. 10.1016/j.chbr.2023.100313 [DOI] [Google Scholar]
  9. Carter, J. W. (2026). Understanding problematic AI companion use through escapism and resilience. Computers in Human Behavior: Artificial Humans, 8, Article 100308. 10.1016/j.chbah.2026.100308 [DOI] [Google Scholar]
  10. Chavez Sosa, J. V., & Huancahuire-Vega, S. (2026). Anxiety and depression associated with the dependent use of generative AI in medical students: Cross-sectional study. JMIR Formative Research, 10, Article e82667. 10.2196/82667 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Ciudad Fernández, V., von Hammerstein, C., & Billieux, J. (2025). People are not becoming “AIholic”: Questioning the “ChatGPT addiction” construct. Addictive Behaviors, 166, Article 108325. 10.1016/j.addbeh.2025.108325 [DOI] [PubMed] [Google Scholar]
  12. Coghlan, S., Leins, K., Sheldrick, S., Cheong, M., Gooding, P., & D’alfonso, S. (2023). To chat or bot to chat: Ethical issues with using chatbots in mental health. Digital Health, 9, 1–11. 10.1177/20552076231183542 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Crawford, J., Allen, K.-A., Pani, B., & Cowling, M. (2024). When artificial intelligence substitutes humans in higher education: The cost of loneliness, student success, and retention. Studies in Higher Education, 49(5), 883–897. 10.1080/03075079.2024.2326956 [DOI] [Google Scholar]
  14. Han, S. (2025). Learner autonomy and mode selection in dual-mode chatbot: Students’ interaction patterns and learning outcomes in an online course. Computers and Education, 240, 105464. 10.1016/j.compedu.2025.105464 [DOI] [Google Scholar]
  15. Hays, R. D., & DiMatteo, M. R. (1987). A short-form measure of loneliness. Journal of Personality Assessment, 51(1), 69–81. 10.1207/s15327752jpa5101_6 [DOI] [PubMed] [Google Scholar]
  16. Hofmann, W., Friese, M., & Strack, F. (2009). Impulse and self-control from a dual-systems perspective. Perspectives on Psychological Science, 4(2), 162–176. 10.1111/j.1745-6924.2009.01116.x [DOI] [PubMed] [Google Scholar]
  17. Hu, B., Mao, Y., & Kim, K. J. (2023). How social anxiety leads to problematic use of conversational AI: The roles of loneliness, rumination, and mind perception. Computers in Human Behavior, 145, Article 107760. 10.1016/j.chb.2023.107760 [DOI] [Google Scholar]
  18. Huang, H., Shi, L., & Pei, X. (2026). When AI becomes a friend: The emotional and rational mechanism of problematic use in generative AI chatbot interactions. International Journal of Human-Computer Interaction, 42(6), 4006–4024. 10.1080/10447318.2025.2536622 [DOI] [Google Scholar]
  19. Kardefelt-Winther, D. (2014). A conceptual and methodological critique of internet addiction research: Towards a model of compensatory internet use. Computers in Human Behavior, 31, 351–354. 10.1016/j.chb.2013.10.059 [DOI] [Google Scholar]
  20. Lai, L., Pan, Y., Xu, R., & Jiang, Y. (2025). Depression and the use of conversational AI for companionship among college students: The mediating role of loneliness and the moderating effects of gender and mind perception. Frontiers in Public Health, 13, Article 1580826. 10.3389/fpubh.2025.1580826 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Latikka, R., Bergdahl, J., Savolainen, I., Celuch, M., & Oksanen, A. (2026). Individual and well-being factors associated with social chatbot usage: A six-country study. Journal of Social & Personal Relationships, 43(4), 1055–1079. 10.1177/02654075251392956 [DOI] [Google Scholar]
  22. Lee, J. (2026). From adolescence to older adulthood: Lifespan pathways linking AI companion chatbots to mental health. Lifespan Development and Mental Health, 2(1), Article 10005. 10.70322/ldmh.2026.10005 [DOI] [Google Scholar]
  23. Li, X., Xin, T., Shang, W., Liu, Y., Du, Y., Lv, L., & Li, B. (2016). Reliability and validity of the boredom proneness scale-short form in university students. Chinese Journal of Clinical Psychology, 24(6), 1029–1033. [Google Scholar]
  24. Little, R. J. A. (1988). A test of missing completely at random for multivariate data with missing values. Journal of the American Statistical Association, 83(404), 1198–1202. 10.1080/01621459.1988.10478722 [DOI] [Google Scholar]
  25. Liu, A. R., Pataranutaporn, P., & Maes, P. (2025). The heterogeneous effects of AI companionship: An empirical model of chatbot usage and loneliness and a typology of user archetypes. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society, 8(2), 1585–1597. 10.1609/aies.v8i2.36658 [DOI] [Google Scholar]
  26. Ma, W., Ma, W., Hu, Y., & Bi, X. (2025). The who, why, and how of AI-based chatbots for learning and teaching in higher education: A systematic review. Education and Information Technologies, 30(6), 7781–7805. 10.1007/s10639-024-13128-6 [DOI] [Google Scholar]
  27. Maral, S., Naycı, N., Bilmez, H., Erdemir, E., & Satici, S. A. (2026). Problematic ChatGPT use scale: AI-human collaboration or unraveling the dark side of ChatGPT. International Journal of Mental Health and Addiction, 24(3), 2369–2395. 10.1007/s11469-025-01509-y [DOI] [Google Scholar]
  28. Prayoga, H., & Wakhid, Z. N. (2024). AI chatbot distractions and academic triumphs: A mediation approach with self-control and coping skills. Journal of Accounting and Investment, 25(2), 673–691. 10.18196/jai.v25i2.20755 [DOI] [Google Scholar]
  29. Qahri-Saremi, H., & Turel, O. (2023). Situational contingencies in susceptibility of social media to phishing: A temptation and restraint model. Journal of Management Information Systems, 40(2), 503–540. 10.1080/07421222.2023.2196779 [DOI] [Google Scholar]
  30. Shen, M. K., Huang, J., Liang, O., Kim, I.-J., & Yoon, D. (2026). The AI genie phenomenon and three types of AI chatbot addiction: Escapist roleplays, pseudosocial companions, and epistemic rabbit holes. In N. Oliver, D. A. Shamma, H. Candello, P. Cesar, P. Lopes, A. Bozzon, T. Kosch, V. Liao, X. Ma, V. Artizzu, F. Draxler, G. López, A. V. Reinschluessel, X. Tong, & P. O. Toups Dugas (Eds.), Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (pp. 1–17). Association for Computing Machinery. 10.1145/3772318.3790896 [DOI] [Google Scholar]
  31. Sin, E., Shao, R., & Lee, T. M. C. (2021). The executive control correlate of loneliness in healthy older people. Aging & Mental Health, 25(7), 1224–1231. 10.1080/13607863.2020.1749832 [DOI] [PubMed] [Google Scholar]
  32. Stavrova, O., Ren, D., & Pronk, T. (2022). Low self-control: A hidden cause of loneliness? Personality and Social Psychology Bulletin, 48(3), 347–362. 10.1177/01461672211007228 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Tagliaferri, G., Martí-Vilar, M., Frisari, F. V., Quaglieri, A., Mari, E., Burrai, J., Giannini, A. M., & Cricenti, C. (2025). Connected by boredom: A systematic review of the role of trait boredom in problematic technology use. Brain Sciences, 15(8), Article 794. 10.3390/brainsci15080794 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Tam, K. Y., van Tilburg, W. A., & Chan, C. S. (2021). What is boredom proneness? A comparison of three characterizations. Journal of Personality, 89(4), 831–846. 10.1111/jopy.12618 [DOI] [PubMed] [Google Scholar]
  35. Tang, D. D., & Wen, Z. L. (2020). Statistical approaches for testing common method bias: Problems and suggestions. Journal of Psychological Science, 43(1), 215–223. [Google Scholar]
  36. Turel, O., & Qahri-Saremi, H. (2016). Problematic use of social networking sites: Antecedents and consequence from a dual-system theory perspective. Journal of Management Information Systems, 33(4), 1087–1116. 10.1080/07421222.2016.1267529 [DOI] [Google Scholar]
  37. Vodanovich, S. J., Wallace, J. C., & Kass, S. J. (2005). A confirmatory approach to the factor structure of the boredom proneness scale: Evidence for a two-factor short form. Journal of Personality Assessment, 85(3), 295–303. 10.1207/s15327752jpa8503_05 [DOI] [PubMed] [Google Scholar]
  38. Wang, T., Li, S., Li, S., Zhang, J., & Gao, J. (2026). Students’ use patterns of generative artificial intelligence during problem-solving in an intelligent learning system: Achievement goal orientation matters. British Journal of Educational Technology, 57(4), 1163–1188. 10.1111/bjet.70059 [DOI] [Google Scholar]
  39. Wang, Y., Hawk, S. T., Wong, N., & Zhang, Y. (2023). Lonely, impulsive, and seeking attention: Predictors of narcissistic adolescents’ antisocial and prosocial behaviors on social media. International Journal of Behavioral Development, 47(6), 540–547. 10.1177/01650254231198034 [DOI] [Google Scholar]
  40. Williams, R. T. (2024). The ethical implications of using generative chatbots in higher education. Frontiers in Education, 8, Article 1331607. 10.3389/feduc.2023.1331607 [DOI] [Google Scholar]
  41. Xie, D., Wang, L., Tao, T., Fan, C., & Gao, W. (2014). Validity and reliability of the Chinese version of the dual-mode of self-control scale for adolescents. Chinese Mental Health Journal, 28(5), 386–391. [Google Scholar]
  42. Yao, R., Qi, G., Sheng, D., Sun, H., & Zhang, J. (2025). Connecting self-esteem to problematic AI chatbot use: The multiple mediating roles of positive and negative psychological states. Frontiers in Psychology, 16, Article 1453072. 10.3389/fpsyg.2025.1453072 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Yu, K., & Zheng, Y. (2020). Sexting and emotional reactions to hooking up among Chinese college students: Moderated mediation effects of loneliness and number of hookup partners. Personality & Individual Differences, 167, Article 110252. 10.1016/j.paid.2020.110252 [DOI] [Google Scholar]
  44. Yu, S. C., Chen, H. R., & Yang, Y. W. (2024). Development and validation the problematic ChatGPT use scale: A preliminary report. Current Psychology, 43(31), 26080–26092. 10.1007/s12144-024-06259-z [DOI] [Google Scholar]
  45. Zahrai, K., Veer, E., Ballantine, P. W., de Vries, H. P., & Prayag, G. (2022). Either you control social media or social media controls you: Understanding the impact of self-control on excessive social media use from the dual-system perspective. Journal of Consumer Affairs, 56(2), 806–848. 10.1111/joca.12449 [DOI] [Google Scholar]
  46. Zhai, N., Ma, X., & Ding, X. (2025). Unpacking AI chatbot dependency: A dual-path model of cognitive and affective mechanisms. Information, 16(12), Article 1025. 10.3390/info16121025 [DOI] [Google Scholar]
  47. Zhang, X., Li, H., Yin, M., Zhang, M., Li, Z., & Chen, Z. (2026). Investigating AI chatbot dependence: Associations with internet and smartphone dependence, mental health outcomes, and the moderating role of usage purposes. International Journal of Human-Computer Interaction, 42(7), 5592–5604. 10.1080/10447318.2025.2545464 [DOI] [Google Scholar]
  48. Zhang, X., Li, Z., Zhang, M., Yin, M., Yang, Z., Gao, D., & Li, H. (2025). Exploring artificial intelligence (AI) chatbot usage behaviors and their association with mental health outcomes in Chinese university students. Journal of Affective Disorders, 380, 394–400. 10.1016/j.jad.2025.03.141 [DOI] [PubMed] [Google Scholar]
  49. Zhong, L., Chen, Y., Zhao, Y., & Yue, Z. (2025). The impact of boredom on college students’ willingness to use generative artificial intelligence [Preprint]. Research Square. 10.21203/rs.3.rs-7370256/v1 [DOI] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Availability Statement

The data supporting the findings of this study are available from the corresponding author upon reasonable request.

Articles from Australian Journal of Psychology are provided here courtesy of Taylor & Francis

RESOURCES