Parental Emotional Neglect and Internet Addiction Among Left-Behind Children: The Parallel Mediating Roles of Deviant Peer Affiliation and Experiential Avoidance

Abstract

Purpose

Adopting a socio-ecological perspective,This study examined the associations between parental emotional neglect and internet addiction among Chinese left-behind children (LBCs) within an approach/avoidance motivation framework. Specifically, it tested whether deviant peer affiliation (behavioral approach pathway) and experiential avoidance (emotional avoidance pathway) would parallelly mediate this association, and whether growth mindset would moderate the direct and indirect pathways.

Methods

A cross-sectional survey of 1,526 rural LBCs in Western China was conducted using validated self-report scales. Parental emotional neglect, deviant peer affiliation, experiential avoidance, growth mindset, and internet addiction were measured. Descriptive statistics, correlation analyses, parallel mediation analysis (PROCESS Model 4), and moderated mediation analysis (PROCESS Model 15) were performed, with relevant demographic and family-related variables controlled.

Results

Parental emotional neglect was positively associated with internet addiction. Both deviant peer affiliation and experiential avoidance partially mediated this association in parallel. The indirect effect through experiential avoidance (25.13% of the total effect) was substantially larger than that through deviant peer affiliation (4.24% of the total effect). Growth mindset significantly moderated the direct association between emotional neglect and internet addiction, but in a positive direction: the association was stronger among LBCs with higher growth mindset (simple slope = 0.5509, p < 0.001) than among those with lower growth mindset (simple slope = 0.4088, p < 0.001). Growth mindset did not moderate the indirect pathways through deviant peer affiliation or experiential avoidance.

Conclusion

Parental emotional neglect is associated with internet addiction among LBCs through both behavioral approach (deviant peer affiliation) and emotional avoidance (experiential avoidance) routes, with experiential avoidance serving as the stronger mediator. Growth mindset appears to play a complex and context-dependent role rather than acting as a straightforward protective factor. Given the cross-sectional design, causal inferences cannot be drawn. Integrated interventions targeting family support, peer environment, and cognitive-emotional regulation may be promising directions for future research and practice.

Introduction

The rapid proliferation of the internet, characterized by a “liquid modernity”, has permeated all facets of society and reshaped the socializing environments of adolescents.¹ While digital natives enjoy the conveniences of online life, new technologies have also unleashed a “Pandora’s box” of cyber-related issues.² Among the most prevalent concerns is internet addiction, often driven by algorithm technology.³ Internet addiction refers to excessive involvement in online activities, leading to problems such as behavioral dysfunction and psychological impairments in daily life.⁴ Substantial evidence links internet addiction to various adverse outcomes, including psychological disorders, diminished quality of life, and behavioral problems.⁵ Adolescents, with their still-maturing cognitive and psychological faculties and relatively poor self-control, are particularly susceptible to the allure of online content and thus constitute a high-risk group for internet addiction.⁶

In China, accelerated urbanization has drawn vast numbers of rural laborers to cities for work, leaving their children behind in hometowns and creating a large, distinct population—left-behind children (LBCs).⁷ According to the Seventh National Population Census in China, there were 41.77 million rural LBCs in China in 2020, accounting for 37.9% of the total rural child population—meaning over one in three rural children were left behind.⁸ LBCs face the dual deficiency of parental supervision and emotional care, placing them at heightened risk for internet addiction.⁹ A meta-analysis involving 38,410 participants revealed that the prevalence of internet addiction among LBCs was 15%, and their risk was 1.37 times higher than that of non-left-behind children.¹⁰

Bronfenbrenner’s ecological systems theory posits that individual development is profoundly influenced by their nested environmental systems.¹¹ The family, as the most immediate microsystem, exerts a deep and lasting impact on children’s psychology and behavior.¹² Previous research has established a link between parental emotional neglect and adolescent internet addiction.^13,14 However, three critical gaps remain in the existing literature. First, although emotional neglect is recognized as a risk factor, the specific mechanisms through which it translates into internet addiction among left-behind children (LBCs) have not been systematically examined. Most prior studies have focused on the direct association or examined single mediators in isolation, leaving the parallel mediating pathways largely unexplored. Second, while ecological systems theory suggests that family dysfunction can cascade into other developmental contexts,¹⁵ few studies have simultaneously examined how family-level risk (emotional neglect) spills over into peer-level risk (deviant peer affiliation) and individual-level risk (experiential avoidance) within an integrated framework. Third, the potential protective role of growth mindset in moderating the relationship between emotional neglect and internet addiction among LBCs has received limited empirical attention. Thus, the present study aims to address these gaps by proposing a moderated parallel mediation model that examines how parental emotional neglect is associated with internet addiction through both behavioral (deviant peer affiliation) and emotional (experiential avoidance) pathways, and whether growth mindset moderates these relationships.

While internalizing problems such as loneliness, depression, and anxiety have been widely documented as consequences of emotional neglect,^16,17 and factors such as self-control and parental monitoring have been identified as protective mechanisms;^18,19 these variables primarily capture the affective outcomes or regulatory capacities of neglect rather than the motivational processes that drive behavior. The motivational dimensional model of affect proposed by Gable and Harmon-Jones²⁰ suggests that the motivational dimension of emotion—whether an individual is oriented toward approaching positive stimuli or avoiding negative ones—may be more critical than its valence dimension in predicting behavior. Among various motivational frameworks in psychology, the approach/avoidance motivation, reflecting the fundamental human instinct to seek benefits and avoid harms, holds broad applicability.^21,22

Within the traditional research framework, the I-PACE (Interaction of Person-Affect-Cognition-Execution) model, proposed by Brand and colleagues, is one of the most comprehensive and widely applied theoretical frameworks for understanding the development and maintenance of internet-use disorders.^23,24 However, the present study adopts the approach/avoidance motivation framework not as a replacement for the I-PACE model, but as a complementary lens that addresses a fundamentally different level of analysis. The I-PACE model primarily explains how addictive behaviors develop and are maintained over time through dynamic cognitive-affective-executive interactions.²⁴ In contrast, the approach/avoidance framework addresses why individuals are motivated to engage in certain behaviors in the first place—specifically, whether their behavior is oriented toward approaching positive/appetitive stimuli or avoiding negative/aversive ones.²² This distinction is critical for our research question: we are not examining the full developmental trajectory of internet addiction (which would indeed require a comprehensive framework like I-PACE), but rather the basic motivational orientations that mediate the link between emotional neglect and internet use. Moreover, the approach/avoidance framework offers parsimony: it allows us to distinguish between two different motivational routes within a single, testable parallel mediation model. By contrast, applying the full I-PACE model to our cross-sectional design would risk over-specification, as many of its core components require longitudinal data to be properly tested.²⁴

Thus, our selection of the approach/avoidance framework is theoretically justified by its conceptual parsimony, its direct alignment with our specific research question (identifying basic motivational tendencies rather than modeling complex developmental processes), and its complementarity with existing frameworks such as the I-PACE model and the compensatory internet use model.²⁵ Rather than competing with these established models, our approach adds a motivational layer that has been relatively underexplored in the internet addiction literature, particularly among left-behind children. By demonstrating that emotional neglect is associated with internet addiction through both approach-oriented (deviant peer affiliation) and avoidance-oriented (experiential avoidance) pathways, we extend the I-PACE model’s insights by specifying which motivational tendencies may serve as entry points into the addictive cycle.

Critically, the approach/avoidance framework maps directly onto the two mediators we propose. Approach motivation—defined as energization toward appetitive stimuli or rewards²²—is conceptually aligned with deviant peer affiliation. When LBCs experience emotional neglect, the deprivation of parental warmth and acceptance may activate an approach-oriented drive to seek alternative sources of belonging and social reward.²⁶ In the absence of adequate parental monitoring, this motivational tendency may channel LBCs toward deviant peer groups that offer immediate acceptance, albeit through rule-breaking or antisocial activities.^27,28 Indeed, empirical evidence has shown that deviant peer affiliation serves as a crucial mechanism linking parental emotional neglect to externalizing problem behaviors among LBCs.²⁹ Conversely, avoidance motivation—defined as energization away from threats or aversive stimuli—is conceptually aligned with experiential avoidance. Emotional neglect generates painful internal experiences such as loneliness, sadness, and anxiety.¹⁶ These aversive emotional states trigger an avoidance-oriented drive to escape or suppress unwanted private experiences.³⁰ LBCs, lacking adequate parental guidance in emotion regulation,³¹ may resort to experiential avoidance as a maladaptive coping strategy—attempting to alter the frequency or form of distressing thoughts and feelings rather than confronting them.³² Recent research has confirmed that experiential avoidance partially mediates the relationship between emotional neglect and internet addiction among left-behind adolescents.³³ Thus, selecting deviant peer affiliation and experiential avoidance as parallel mediators allows us to test whether emotional neglect is associated with internet addiction through two qualitatively different motivational routes—one behavioral-approach and one emotional-avoidance—offering a more comprehensive understanding of the underlying mechanisms than examining either mediator in isolation.

In addition to examining these two mediating pathways, the present study introduces growth mindset as a potential moderator. Growth mindset—the belief that personal attributes such as intelligence and ability are malleable and can be developed through effort³⁴—has been shown to buffer the negative effects of various adversities on adolescent outcomes.^35,36 For LBCs specifically, who often lack consistent external support due to parental absence, growth mindset may serve as a particularly valuable internal cognitive resource that alters how they interpret and respond to emotional neglect. Unlike external protective factors such as parental monitoring or school connectedness which may be limited for LBCs due to geographic separation. Growth mindset remains accessible regardless of parental availability. Research has demonstrated that growth mindset is negatively associated with mental health problems and can reduce internet addiction.^37,38 However, whether growth mindset moderates the direct and/or indirect associations between emotional neglect and internet addiction remains empirically untested in this population. The detailed theoretical rationale for why growth mindset would moderate each specific pathway is elaborated in the hypotheses section.

In summary, the present study aims to address the identified research gaps by proposing and testing a moderated parallel mediation model. Specifically, we seek to examine: (a) whether parental emotional neglect is positively associated with internet addiction among LBCs (Hypothesis 1); (b) whether deviant peer affiliation (as an approach pathway) and experiential avoidance (as an avoidance pathway) mediate this association in parallel (Hypotheses 2 and 3); and (c) whether growth mindset moderates the direct effect of emotional neglect on internet addiction, while also exploring whether it moderates the indirect pathways (Hypotheses 4a–4c). By integrating ecological systems theory with the approach/avoidance motivation framework, this study offers several contributions. First, and most importantly, it extends previous research by simultaneously testing two theoretically distinct mediating pathways within a single model. While the individual relationships between emotional neglect and deviant peer affiliation, between emotional neglect and experiential avoidance, and between these mediators and internet addiction have been documented separately in prior literature, no study has integrated these two pathways into a parallel mediation model to compare their relative strengths in the context of left-behind children. This integration allows us to illuminate the parallel mechanisms through which family risk may be associated with adolescent internet addiction and to identify which pathway carries greater explanatory weight. Second, it applies the approach/avoidance framework to the understudied population of LBCs, demonstrating how motivational tendencies triggered by emotional neglect can manifest in both peer-directed approach behaviors and emotion-directed avoidance behaviors. Third, it examines growth mindset as a potential moderator, exploring whether this belief system can alter the direct and indirect associations between emotional neglect and internet addiction. Taken together, the primary contribution of this study lies not in discovering entirely novel relationships, but in providing a more comprehensive and integrative understanding of how multiple mechanisms operate simultaneously to link emotional neglect to internet addiction among left-behind children.

Literature Review and Hypotheses

Emotional Neglect and Internet Addiction in Left-Behind Children

Parental acceptance-rejection theory (PARTheory) asserts that all children possess a fundamental need for parental acceptance and love, and when this need is unmet, psychological maladjustment and problematic behaviors, including addiction, may ensue.^39–41 Among the various forms of parental rejection identified by PARTheory—namely coldness, hostility, indifference/neglect, and undifferentiated rejection—emotional neglect corresponds specifically to the indifference/neglect dimension, referring to the failure of caregivers to provide adequate emotional support, warmth, and responsiveness.^42,43 Unlike active forms of rejection (eg, hostility or overt criticism), emotional neglect is characterized by omission—the absence of sufficient care, comfort, and validation—and is particularly prevalent among left-behind children, whose geographical separation from parents results in infrequent and discontinuous emotional communication, rendering parents “observers” rather than active participants in their emotional development.^44,45 While emotional neglect is thus a specific behavioral expression of parental rejection, it is not synonymous with it—a distinction that becomes important when considering the broader relational consequences for children. Specifically, the experience of emotional neglect may foster insecure attachment representations—rooted in Bowlby’s attachment theory^46,47—whereby children develop a lack of confident expectation that their caregivers will be available and comforting in times of need. Insecure attachment is not equivalent to emotional neglect or parental rejection; rather, it is a psychological outcome that may emerge from these caregiving experiences,^48,49 with emotional neglect being particularly associated with avoidant attachment patterns.⁵⁰ Taken together, emotional neglect (a caregiving behavior defined by omission), parental rejection (a broader relational construct that includes emotional neglect as one behavioral expression), and insecure attachment (a psychological adaptation that may result from such experiences) represent distinct but causally linked concepts. Drawing on this conceptual chain, we argue that when LBCs experience emotional neglect and the subsequent formation of insecure attachment representations, their basic psychological needs for belonging and security are frustrated, driving them to seek compensatory fulfillment through alternative means.⁵¹ The model of compensatory internet use posits that when emotional needs are not met in real life, individuals may turn to the internet to satisfy these unmet needs.²⁵ Empirically, numerous studies identify escapism and self-medication motives as key drivers of behavioral addictions,⁵² and a growing body of research demonstrates that psychological distress stemming from poor parent-child relationships can significantly predict maladaptive behaviors like internet addiction.^48,53,54 For instance, a survey from Italy found that children who perceived parental rejection, rather than acceptance, were more prone to developing behavioral problems,⁵⁵ and a study of over 2,000 LBCs in Central China confirmed that parental rejection was a significant factor influencing their internet addiction.⁵⁶ Based on this reasoning, we propose our first hypothesis:

H1: Parental emotional neglect positively predicts internet addiction among left-behind children.

Dual Mediating Pathways Under the Approach/Avoidance Framework

The idea that humans naturally seek pleasure and avoid pain dates back to ancient Greek philosophers like Democritus and Aristippus. The English philosopher Jeremy Bentham further advanced this idea, stating that “Nature has placed mankind under the governance of two sovereign masters: pain and pleasure”.⁵⁷ Later, psychologist William James, in his seminal work The Principles of Psychology, described pleasure and pain as the springs of action, with pleasure acting as a “reinforcer” and pain as an “inhibitor”.⁵⁸ Building on this foundation, Elliot formally proposed the idea of the approach/avoidance motivation framework.⁵⁹ Behavioral approach motivation involves a predisposition toward positive or appetitive stimuli, whereas behavioral avoidance motivation involves a predisposition away from negative or aversive stimuli.⁶⁰ Psychology recognizes that two motivations can be generated relatively independently at the same time due to the diverse stimuli cue in reality.⁶¹ For example, research shows that individuals can experience both approach and avoidance motivations in stressful situations, further influencing their subsequent behavior.⁶² Similarly, this study posits that in response to parental emotional neglect, LBCs may develop both “approach” and “avoidance” motivational tendencies.

The Approach Pathway: The Mediating Role of Deviant Peer Affiliation

To avoid redundancy with the earlier discussion of emotional deprivation and compensatory mechanisms presented in the Introduction and Emotional Neglect and Internet Addiction in Left-Behind Children, we do not reiterate the general arguments about unmet psychological needs here. Instead, we focus specifically on how the approach motivation framework directly explains the mediating role of deviant peer affiliation in the link between emotional neglect and internet addiction among left-behind children.

On the “approach” side, Emotional neglect means insufficient parental care and supervision. The estrangement of intimate relationships and the intermittent of emotional communication leave LBCs yearning for interpersonal connection but finding their primary group unresponsive, having to turn to peers for social interaction. However, the lack of parental guidance and monitoring can disrupt the normal socialization process, impairing their development of adequate social skills, proper habits, and moral integrity.⁶³ This socialization failure may leave LBCs lacking the competence to develop and maintain healthy relationships, potentially leading to rejection by normative peers and consequently driving them toward deviant peers for a sense of belonging.^28,64 Deviant peer affiliation refers to associations with peers who engage in rule-breaking or socially deviant behaviors.⁶⁵ A key parental role during adolescence is to mitigate negative peer influences.²⁷ However, LBCs experience not only physical separation but also psychological alienation due to emotional neglect. As a consequence, they have to participate in activities with deviant peer groups to garner emotional support from them.⁶⁶ Existing research indicates that deviant adolescents often exhibit more profound internet use behaviors.⁶⁷ Once LBCs associate with deviant peers, they may feel compelled by peer pressure to join in online activities.⁶⁸ Furthermore, adolescents are highly sensitive to peer opinions and evaluations.⁶⁹ According to Bandura’s social learning theory, prolonged exposure to deviant peer groups, through observation and imitation of their online behaviors and discussions about internet related topics, can lead individuals to develop their own fascination with the internet.⁷⁰ Simultaneously, interactions within deviant peer groups can reinforce an individual’s own online behaviors, leading to increased immersion in the virtual world through peer contagion.⁷¹ Consequently, numerous empirical studies demonstrate that deviant peer affiliation positively predict individual internet addiction.⁷² For example, research has found that elementary school students who associate with peers who play online games are more likely to become addicted to online games.⁷³ Based on this, we propose the following hypothesis:

H2: Deviant peer affiliation mediates the relationship between parental emotional neglect and internet addiction among left-behind children.

The Avoidance Pathway: The Mediating Role of Experiential Avoidance

Experiential avoidance is defined as the individual attempt to avoid, alter, or escape from unwanted private experiences, such as painful thoughts, feelings, and memories, and try to change the frequency or form of these experiences.³² Research identifies experiential avoidance as a core vulnerability factor for psychological distress.³⁰ It is positively correlated with emotional issues like anxiety, depression, and alexithymia.^74,75 As far as this topic is concerned, on one hand, emotional neglect leads LBCs to accumulate painful internal experiences such as loneliness, sadness, inferiority, and anxiety.¹⁶ These aversive emotional states are precisely what LBCs strive to avoid and eliminate, making them prone to develop experiential avoidance. On the other hand, lacking external social support from parents, they may turn inward and develop maladaptive self-regulation strategies. Children typically learn emotion regulation skills through interactions with their caregivers; parents usually help children understand and manage emotions through empathy and soothing.³¹ However, LBCs who experience emotional neglect miss out on this crucial learning process. They often lack adaptive regulation strategies like problem-solving, cognitive reappraisal, or support seeking strategies. When confronted with negative emotions, their only perceived option may be “escape” or “suppression”, eventually fostering a habit of experiential avoidance characterized by persistent efforts to rid themselves of painful emotions and thoughts.¹⁷ Thus, it is reasonable to hypothesize from the above analysis that parental emotional neglect can positively predict experiential avoidance in left behind children.

While experiential avoidance might function as a short-term self-regulation strategy to manage unpleasant feelings, its habitual use can evolve into a rigid, escape-oriented coping style with numerous negative consequences.⁷⁶ This is because, firstly, habitual avoidance leads to psychological inflexibility, reducing an individual’s adaptability in facing life’s challenges.⁷⁷ Secondly, the process of experiential avoidance keeps individuals entangled with negative events, obstructing the pursuit of valued goals.³⁰ The internet, as a means of self-soothing or distraction, can offer temporary relief from stress or negative emotions.⁷⁸ However, habitual experiential avoidance can lead to a persistent reliance on the internet, ultimately fostering addiction.⁷⁹ For instance, a Spanish study found positive correlations between experiential avoidance and internet addiction, mobile phone addiction, and video game addiction.⁸⁰ A study of 500 undergraduate students by Chou et al⁸¹ found that experiential avoidance positively predicted internet addiction. Another survey of 2,731 respondents showed that experiential avoidance could not only directly lead to internet addiction but also indirectly contribute to it through the mediation of depression.⁸² Given this evidence, we propose:

H3: Experiential avoidance mediates the relationship between parental emotional neglect and internet addiction among left-behind children.

The Moderating Role of Growth Mindset

Growth mindset is contrasted with a fixed mindset. A fixed mindset assumes that personal qualities, such as intelligence, are static and unchangeable, leading individuals to avoid challenges and effort.³⁴ In contrast, a growth mindset is a flexible belief system positing that abilities can be developed through dedication, learning, and practice.³⁴ Individuals with a growth mindset tend to embrace challenges and demonstrate greater resilience in face of setbacks.⁸³ This study proposes that growth mindset plays a crucial protective role against internet addiction among LBCs.

First, a growth mindset encourages individuals to find meaning from challenges and set long-term growth goals.³⁵ Children with a strong growth mindset firmly believe that personal traits and abilities are malleable.⁸⁴ When they perceive emotional neglect, they might proactively attempt to build supportive relationships with teachers, grandparents, or other guardians, or choose alternative coping strategies like keeping a diary or doing some sports to regulate their emotions. Consequently, when facing emotional neglect, LBCs with a higher growth mindset are more likely to direct their attention toward support-seeking, skill development, or interest diversion rather than immersing themselves excessively in cyberspace;Second, individuals with a strong growth mindset are more willing to accept and confront negative events and emotions, viewing them as opportunities for growth rather than threats.⁸⁵ This emotional acceptance capacity may be helpful to mitigate the driving effect of experiential avoidance on internet addiction; Third, even if currently rejected by normative peers, LBCs with a growth mindset believe their social circles can change through effort, preventing them from feeling permanently stuck in deviant peer groups. Therefore, when facing pressure from deviant peers, those with a high growth mindset might remain more focused on long-term goals like academic achievement rather than passively succumb to the influence of deviant peers’ internet use. Furthermore, growth mindset can encourage individuals to adopt proactive problem-solving strategies.⁸⁶ Thus, when interacting with deviant peers, especially when faced with online invitations from deviant peers, LBCs with a strong growth mindset might be more inclined to seek advice from teachers or guardians, thereby resisting the negative influence. Conversely, individuals with a lower growth mindset have lower self-efficacy and less confidence in their ability to control their behavior and environment,⁸⁷ making them more susceptible to participating in online activities invited by deviant peers, ultimately leading to internet addiction. Based on this analysis, we propose the following hypotheses:

H4a: Growth mindset moderates the direct relationship between parental emotional neglect and internet addiction among left-behind children.

H4b: Growth mindset moderates the relationship between deviant peer affiliation and internet addiction among left-behind children.

H4c: Growth mindset moderates the relationship between experiential avoidance and internet addiction among left-behind children.

Based on the above hypotheses and research questions, the research framework is illustrated in Figure 1.

Figure 1.

Research hypothesis model proposed in this study.

Methods

Participants and Procedure

Definition of Left-Behind Children

The conceptualization of left-behind children in Chinese academic discourse has generally centered on three core elements:⁸⁸ (a) the number of migrant parents, (b) the duration of parental absence, and (c) the child’s age. Regarding the first element, the vast majority of studies agree that a child with at least one parent working away from home qualifies as a left-behind child.⁸⁹ Regarding the second element, although a minority of studies have adopted a threshold of one year or more,⁹⁰ the prevailing consensus—and the standard adopted by most empirical investigations—is a duration of six months or more.^88,91 Regarding the third element, the academic literature in China generally defines left-behind children as minors under the age of 18.⁹² Synthesizing these criteria, the present study operationalizes “left-behind children” as rural children under 18 years of age whose at least one parent has migrated to another city for work for a period of six months or more.

Study Timeframe and Sampling Source

Data were collected in September 2025. Due to constraints in research funding, personnel, and geographic accessibility, a multi-stage convenience sampling strategy was employed. First, we selected three provinces in Western China—Sichuan, Guizhou, and Gansu—based on their status as major labor-exporting regions with high concentrations of rural left-behind children. Considering that rural left-behind children are primarily enrolled in schools within county-level administrative regions of China,⁸⁸ Within each province, we selected county-level administrative districts with large-scale labor emigration and high proportions of left-behind children: Zhongjiang County, Renshou County, Jintang County, Shehong County and Pingshan County in Sichuan Province; Zhijin County and Xixiu District in Guizhou Province; and Jingyuan County in Gansu Province. These counties are all recognized as rural labor-exporting areas within their respective provinces. Second, within each county, we selected 1–2 primary or secondary schools, and administered the survey to intact classes after selecting specific grade levels. In total, 12 schools and 58 classes participated in the study. Third, to avoid stigmatizing participants with the “left-behind” label, we distributed questionnaires to all students in the selected classes and subsequently identified the left-behind children subsample based on their residency status—specifically, whether at least one parent had been working away from home for more than six months.

It is important to note that although students were nested within schools, the present study included only 12 schools. With such a small number of clusters at the school level, multilevel modeling would be statistically underpowered and could produce unstable estimates.^93,94 Simulation studies have demonstrated that multilevel models require a minimum of 30 to 50 clusters at the higher level to yield reliable estimates of random effects and standard errors.⁹⁵ Given the limited number of schools in this study, we adopted a design-based approach rather than a model-based approach to address potential clustering effects. Specifically, we employed a multi-stage sampling strategy that deliberately sampled schools from diverse counties and geographic regions within each province to enhance between-school variability. This approach may help reduce the risk that observed associations are driven by school-specific idiosyncrasies. However, we acknowledge that without directly modeling intraclass correlation, school-level dependence cannot be empirically ruled out. Additionally, the large sample size (N = 1,526) and the dispersion of participants across 12 schools with an average of approximately 128 students per school contribute to the robustness of the estimates. Nevertheless, we acknowledge the limitation of not formally modeling school-level clustering and have addressed this in the Limitations section.

Age Range and Inclusion/Exclusion Criteria

Given that younger children may have difficulty in comprehending questionnaire items, our survey in primary schools was restricted to fifth and sixth graders (approximately age 11 and above). Additionally, we excluded senior high school seniors (Grade 12) because they were under intense pressure from the National College Entrance Examination (Gaokao) and could not be reasonably expected to participate in a voluntary survey. Thus, the actual survey population spanned from fifth-grade primary school students (age 11) to second-year senior high school students (age 17).

Regarding informed consent procedures, because all participants were minors under the age of 18, we obtained both parental permission and child assent prior to data collection. Parental or guardian written informed consent was obtained through consent forms distributed to families via the participating schools. For the children themselves, assent was sought through an age-appropriate explanation of the study purpose, the voluntary nature of participation, and the right to withdraw at any time without consequence.⁹⁶ School administrators were not involved in participant recruitment. Their role was limited to providing logistical support, such as scheduling survey administration times and facilitating communication with families for consent form distribution. All recruitment and consent procedures were conducted by the research team in accordance with the ethical approval granted by the Institutional Review Board of the authors’ institution (IRB-SWJTU-SH-034).

Data Collection and Missing Data Handling

The survey was administered using paper-and-pencil questionnaires in a classroom setting. Prior to data collection, the research team provided remote training to coordinating teachers at each school to ensure familiarity with the questionnaire structure and standardized administration instructions. During administration, teachers distributed the questionnaires in class, read the standardized instructions aloud, and remained available to answer questions while students completed the survey independently. All completed questionnaires were collected on-site immediately upon completion. A total of 3,560 questionnaires were collected. We then applied three exclusion criteria to identify invalid responses: (a) questionnaires with incomplete responses (ie, missing more than 10% of items); (b) questionnaires exhibiting invariant response patterns (eg, selecting the same option for a large number of consecutive items); and (c) questionnaires containing logical inconsistencies among reverse-coded items. After applying these criteria, 3,102 valid questionnaires were retained, including 1,526 from left-behind children and 1,576 from non-left-behind children. The present analysis focuses exclusively on the 1,526 valid questionnaires completed by left-behind children.

It should be noted that given that the questionnaire included items about childhood emotional neglect, which may evoke distress in some participants, we implemented a distressed respondent protocol in accordance with established ethical guidelines for trauma-related research with minors.⁹⁷ Specifically, prior to survey administration, coordinating teachers were trained to recognize signs of participant distress, such as visible upset, reluctance to continue, or emotional withdrawal. Participants were informed before the survey that they could skip any question they found uncomfortable and could withdraw from the survey at any time without penalty. The cover page of the questionnaire included contact information for school counselors who were available to provide support if needed. In cases where a participant exhibited signs of significant distress, the teacher was instructed to pause the survey and offer the participant the option to take a break or to speak privately with the school counselor. The results afterwards prove that no participants requested withdrawal or required crisis intervention during the data collection process.

Sample Size Power Analysis

To determine whether the sample size was adequate for the planned statistical analyses, we conducted a post-hoc power analysis using G*Power 3.1.⁹⁸ Following the conventions established by Cohen,⁹⁹ we specified a small to medium effect size (f² = 0.05) for the multiple regression analyses, which is conservative and appropriate given that the expected effects in social science research often fall within this range. With a significance level of α = 0.05 and a desired statistical power of 0.80, we calculated the minimum required sample size based on the actual number of predictors in our full moderated mediation model. Specifically, the model includes: (a) the independent variable (emotional neglect), (b) two mediators (deviant peer affiliation and experiential avoidance), (c) three interaction terms (Emotional Neglect × Growth Mindset, Deviant Peer Affiliation × Growth Mindset, and Experiential Avoidance × Growth Mindset), and (d) seven control variables (gender, grade, socioeconomic status, frequency of parent-child communication et al) This yields a total of 13 predictors. Using G*Power with these parameters (f² = 0.05, α = 0.05, power = 0.80, number of predictors = 13), the minimum required sample size was calculated as 305 participants.^98,99 Our final sample of 1,526 participants exceeds this requirement by approximately fivefold, indicating that the study is sufficiently powered to detect the hypothesized effects, including the moderation effects.

Furthermore, for the mediation analyses, we followed the recommendations of Fritz and MacKinnon,¹⁰⁰ who conducted a simulation study to determine the sample sizes required to achieve 0.80 power for detecting indirect effects using bootstrapping. According to their guidelines, for a small effect size in both the a-path (predictor to mediator) and b-path (mediator to outcome) and using a bias-corrected bootstrap test with 5,000 resamples, the recommended sample size is 462 participants for a simple mediation model. For a parallel mediation model with two mediators, the required sample size is comparable or even smaller given the increased information in the model.^100,101 Our sample of 1,526 participants substantially exceeds this threshold, ensuring adequate power for detecting the indirect effects through both deviant peer affiliation and experiential avoidance.

Measures

Emotional Neglect

This study used the emotional neglect subscale in the Childhood Trauma Questionnaire (CTQ-SF) developed by Bernstein et al⁴³ to measure the emotional neglect experiences of left behind children. Items were rated on a 5-point Likert scale. Higher scores indicate more severe experiences of emotional neglect. The reliability and validity of this scale have been tested by Chinese scholars.¹⁰² In this study, the Cronbach’s α for this subscale was 0.854.

Experiential Avoidance

This study used the Acceptance and Action Questionnaire Second Edition (AAQ-II) developed by Bond et al¹⁰³ to measure participants’ experiential avoidance. The questionnaire consists of 7 items. Items were rated on a 5-point Likert scale, and a total score was calculated, with higher scores reflecting greater experiential avoidance. Sample items include: “Painful memories make it difficult for me to live a life that I would value” and “It seems that most people are handling their lives better than I am”. Cronbach’s α was 0.773 in this study.

Deviant Peer Affiliation

This variable was measured using an 8-item scale adapted from Li et al¹⁹ employing a 5-point Likert scale. Participants reported how many of their good friends engaged in various deviant behaviors (eg, smoking, drinking). Cronbach’s α was 0.843 in this study.

Growth Mindset

We used a 6-item scale adapted from Dweck’s Implicit Theories of Intelligence Scale¹⁰⁴ to assess the growth mindset of these LBCs. Items were rated on a 5-point Likert scale from 1 (Strongly Disagree) to 5 (Strongly Agree). Higher scores indicate a stronger growth mindset. A sample item is: “No matter what your current intelligence level is, you can always change it quite a bit”. Cronbach’s α was 0.801.

Internet Addiction

Internet addiction was measured using an adapted version of the 9-item Problematic Internet Use Questionnaire.^105,106 Items were rated on a 5-point Likert scale (1=Never, 5=Always), with higher scores indicating more severe internet addiction. Sample items include: “Do you neglect your duties (schoolwork) because of the internet?” and “Do people close to you complain about you spending too much time online?”Cronbach’s α was 0.923.

Data Analysis

We first used SPSS 26.0 to compute descriptive statistics for all variables and the bivariate correlations between them. Next, we tested the hypothesized mediation model using Model 4 of PROCESS.¹⁰⁷ To examine whether growth mindset moderated the direct path from emotional neglect to internet addiction, as well as the indirect paths through the two mediators, we employed PROCESS Model 15. This model tests moderation on both the direct path and the indirect paths, allowing us to distinguish between moderated mediation and moderation of the direct effect.

Results

Common Method Bias Test

In this study, common method bias was controlled through both procedural and statistical approaches.¹⁰⁸ Procedurally, two main measures were implemented. First, the survey cover page explicitly informed participants of full anonymity, stated no identifying personal information would be collected, clarified all responses served solely academic research purposes, and emphasized there were no “correct or incorrect” answers to reduce social desirability bias and response defensiveness. Second, variables such as emotional neglect, internet addiction, experiential avoidance, deviant peer affiliation and growth mindset were distributed separately throughout the questionnaire rather than clustered together, thereby disrupting respondents’ response inertia and attenuating recency effects across scales.

Statistically, Harman’s single-factor test was conducted using an unrotated principal component factor analysis on all items. The analysis extracted 6 factors with eigenvalues greater than 1. The first factor accounted for 24.034% of the total variance, well below the critical threshold of 40%, indicating that common method bias was not a serious issue in this study.¹⁰⁹

To further test for common method bias, we employed a partial correlation method controlled by the first principal component as an additional statistical procedure.¹¹⁰ Specifically, we first extracted the first principal component from all measurement items and used it as a proxy variable for the common method factor. Subsequently, this first principal component was entered as a covariate in the regression model to re-examine the mediation effects. The key criterion was to observe whether the path coefficients from the independent variable to the dependent variable, as well as the significance and direction of the mediation effects, underwent any substantial changes after controlling for the common method factor.¹¹⁰ Following this procedure, after including the first principal component of all items as a covariate in the regression model, we found that the direct path coefficient from parental emotional neglect to internet addiction decreased from 0.4947 to 0.0330, but the direction and significance level remained unchanged (still p < 0.001). The path coefficient for the mediation pathway through deviant peer affiliation decreased from 0.2426 to 0.0101, but the direction and significance level remained unchanged (still p < 0.001). The path coefficient for the mediation pathway through experiential avoidance decreased from 0.4174 to 0.0094, but the direction and significance level remained unchanged (still p < 0.001). These results indicate that common method bias did not substantially affect the mediation model estimates in this study.

Descriptive Statistics and Correlations

Means, standard deviations, and correlations among the key variables are presented in Table 1. Preliminary correlation analysis revealed that parental emotional neglect was positively correlated with deviant peer affiliation, experiential avoidance, and internet addiction among LBCs. Growth mindset was negatively correlated with emotional neglect, deviant peer affiliation, experiential avoidance and internet addiction. Furthermore, both deviant peer affiliation and experiential avoidance were positively correlated with internet addiction.

Table 1.

Means, Standard Deviations, and Correlation Matrix for Study Variables

Variables	M	SD	Emotional Neglect	Deviant Peer Affiliation	Experiential Avoidance	Growth Mindset	Internet Addiction
Emotional Neglect	13.7024	4.2095	1
Deviant Peer Affiliation	13.6992	3.7774	0.0929***	1
Experiential Avoidance	16.1933	4.4508	0.3844***	0.1302***	1
Growth Mindset	21.3401	3.3187	−0.1398***	−0.3132***	−0.0768***	1
Internet Addiction	21.0603	6.3212	0.4794***	0.2077***	0.4442***	−0.1454***	1

Note: ***P<0.001.

Parallel Mediation Analysis

The authors used Hayes’ PROCESS Model 4 for SPSS, with emotional neglect as the independent variable, internet addiction as the dependent variable, and deviant peer affiliation and experiential avoidance as parallel mediators. Prior research has demonstrated that gender,¹¹¹ grade level,¹¹² family socioeconomic status,¹¹³ parental educational attainment,¹¹⁴ and the frequency of parent-child communication¹¹⁵ are closely associated with internet addiction among left-behind children and adolescents. By controlling for these variables, we aimed to more accurately estimate the net effect of parental emotional neglect on internet addiction among left-behind children.

It is worth noting that in the sociological literature, socioeconomic status (SES) is conventionally measured using three core indicators: family income, parental education, and occupational status.¹¹⁶ Occupational status, however, involves complex procedures such as classification according to the Chinese National Standard Classification of Occupations, conversion to the International Standard Classification of Occupations (ISCO), and transformation into the International Socio-Economic Index of occupational status (ISEI). Given these operational complexities, the present study did not include occupational status in the initial survey design. Importantly, empirical research in the Chinese context has demonstrated that SES indices estimated solely on the basis of education and income are largely consistent with subjective evaluations of occupational prestige.¹¹⁷ Li’s¹¹⁷ national survey study found that the primary determinants of prestige position in contemporary Chinese society are education, income, power, and work unit type—with education and income being the most influential factors. Accordingly, we controlled for annual household income, paternal education, and maternal education (coded as follows: primary school or below = 6 years; middle school = 9 years; secondary technical school or general high school = 12 years; associate degree = 15 years; bachelor’s degree or above = 16 years) as proxies for family SES.

The results (see Table 2) showed that parental emotional neglect positively predicted internet addiction (β= 0.7004, p <0.001). After including the two mediators in the model, the direct predictive effect of emotional neglect on internet addiction remained significant, though attenuated (β= 0.4947, p <0.001). It indicates that research hypothesis 1 is supported. Parental emotional neglect also significantly and positively predicted both deviant peer affiliation (β= 0.1222 p <0.001) and experiential avoidance (β= 0.4219, p <0.001). Furthermore, both deviant peer affiliation (β=0.2426, p <0.001) and experiential avoidance (β= 0.4174, p <0.001) significantly predicted internet addiction, supporting the parallel mediation model H1,H2,H3.

Table 2.

Parallel Mediation Model Analysis Results

	Internet Addiction	Deviant Peer Affiliation	Experiential Avoidance	Internet Addiction
Gender	0.4431 (0.2989)	0.0711*** (0.1952)	0.2054*** (0.2191)	−0.5625 (0.2917)
Grade	0.0533 (0.1594)	0.5101*** (0.1041)	−0.2888* (0.1169)	0.0501 (0.1507)
Annual household income	−0.2612 (0.7939)	0.3059*** (0.0347)	−0.0955 (0.0172)	−0.5143 (0.6071)
Father’s educational attainment	0.0632 (0.1224)	−0.0318 (0.0799)	−0.0073 (0.0897)	0.0740 (0.1145)
Mother’s educational attainment	0.3967** (0.1324)	0.1387 (0.0865)	0.0966 (0.0971)	0.3228** (0.1240)
Frequency of communication with father	−0.0210 (0.0392)	−0.0155 (0.0256)	−0.0391 (0.0287)	−0.0009 (0.0366)
Frequency of communication with mother	−0.0659 (0.0418)	0.0158 (0.0273)	0.0062 (0.0306)	−0.0723 (0.0391)
Emotional neglect	0.7004*** (0.0349)	0.1222*** (0.0228)	0.4219*** (0.0256)	0.4947*** (0.0356)
Deviant peer affiliation				0.2426*** (0.0369)
Experiential avoidance				0.4174*** (0.0329)
R2	0.2410	0.0935	0.1770	0.3369
F	60.1949	19.5639	40.7865	76.9634

Notes: All regression coefficients in the table are standardized regression coefficients. The figures in brackets behind the regression coefficients are the standard errors. *P<0.05, **P<0.01, ***P<0.001.

Bootstrap confidence interval analysis (5,000 samples) further confirmed these effects (see Table 3). The 95% bias-corrected confidence intervals for the direct effect of emotional neglect on internet addiction and for the specific indirect effects through both mediators did not include zero. This indicates that emotional neglect not only directly predicts LBCs’ internet addiction but also does so indirectly through both deviant peer affiliation and experiential avoidance. Thus, research hypotheses 2 and 3 are supported. In this model, the direct effect (0.4947), the indirect effect via deviant peer affiliation (0.0297), and the indirect effect via experiential avoidance (0.1761) accounted for 70.63%, 4.24%, and 25.13% of the total effect, respectively.

Table 3.

Decomposition of Total, Direct, and Indirect Effects

Effect Type	Effect Value	Boot SE	LLCI	ULCI	Relative Effect (%)
Total Effect	0.7004	0.0349	0.6320	0.7689
Direct Effect	0.4947	0.0356	0.4248	0.5645	70.63%
Total Indirect Effect	0.2058	0.0233	0.1618	0.2532	29.37%
Indirect via Deviant Peers	0.0297	0.0081	0.0155	0.0471	4.24%
Indirect via Experiential Avoidance	0.1761	0.0230	0.1333	0.2230	25.13%

A Moderated Analysis

To examine whether growth mindset moderated the direct path from emotional neglect to internet addiction, as well as the indirect paths through the two mediators, we employed PROCESS Model 15. Contrary to the protective moderation anticipated in the moderating role of growth mindset, the observed direction of the significant interaction was positive rather than negative. As shown in Table 4, when emotional neglect, the two mediators, growth mindset, and their interaction terms were included in the regression model predicting internet addiction, the results showed that emotional neglect, deviant peer affiliation, and experiential avoidance remained positive predictors. However, among the interaction terms, only the Emotional Neglect × Growth Mindset interaction was significant (β=0.0238, p <0.05). Hypothesis 4a is supported, indicating that growth mindset significantly moderates the direct association between emotional neglect and internet addiction. However, hypotheses 4b and 4c are not supported, as growth mindset did not significantly moderate the indirect pathways through deviant peer affiliation or experiential avoidance. Thus, the overall pattern is best characterized as a moderation of the direct effect rather than a moderated mediation model.

Table 4.

Moderated Mediation Analysis Results

Regression Equation		Goodness of Fit		Significance of Regression Coefficient
Outcome Variable	Predictor	R2	F	β	t
Internet Addiction		0.3399	55.5688
	Gender			−0.5243	−1.7944
	Grade			0.0321	0.2126
	Annual household income			−0.0134	−0.6250
	Father’s educational attainment			−0.0042	−0.1147
	Mother’s educational attainment			−0.0715	−1.8259
	Frequency of communication with father			−0.0585	−0.5098
	Frequency of communication with mother			−0.3321	−2.6733**
	Emotional Neglect			0.4877	13.5452***
	Deviant Peer Affiliation			0.2264	5.8338***
	Experiential Avoidance			0.4083	12.3449***
	Growth Mindset			−0.0651	−1.4547
	Emotional Neglect × Growth Mindset			0.0238	2.2344*
	Deviant Peer Affiliation× Growth Mindset			−0.0026	−0.2606
	Experiential Avoidance × Growth Mindset			−0.0023	−0.2496

Notes: *P<0.05, **P<0.01, ***P<0.001.

To probe the significant interaction, a simple slope analysis was conducted, dividing growth mindset into high (+1 SD) and low (−1 SD) groups. The specific moderating effect is depicted in Figure 2 and detailed in Table 5. The analysis showed that the positive association between parental emotional neglect and internet addiction was weaker for LBCs with a lower growth mindset (simple slope = 0.4088, p <0.001) and stronger for those with a higher growth mindset (simple slope = 0.5509, p <0.001).

Figure 2.

The Moderating Effect of Growth Mindset on the Relationship Between Emotional Neglect and Internet Addiction.

Table 5.

Conditional Direct Effects of Emotional Neglect on Internet Addiction at Different Levels of Growth Mindset

Level of Growth Mindset	Effect	Boot SE	LLCI	ULCI
M-1SD	0.4088	0.0509	0.3090	0.5086
M	0.4877	0.0360	0.4171	0.5583
M+1SD	0.5509	0.0454	0.4619	0.6399

Discussion

The family occupies a central position in child development, exerting a lasting influence on the growth of children and adolescents.¹¹⁸ The “Risky families model” proposed by Repetti et al¹¹⁹ posits that growing up in a conflictual, cold, neglectful, and unsupportive family environment can impair the development of children’s emotional and social competencies, thereby predisposing them to risk behaviors. Empirical research indicates that emotional neglect is one of the most common forms of maltreatment experienced among adolescents.⁴⁵ Compared to physical or sexual abuse, neglect is more prevalent and equally detrimental to children’s short-term and long-term development.¹²⁰ Berman et al¹²¹ categorized adverse childhood experiences into “threatening experiences” (eg, domestic violence) and “deprivation experiences” (eg, neglect and abuse), arguing that both types have enduring effects on children’s physical and mental health.

Although preliminary evidence links emotional neglect to adolescent smartphone addiction,¹²² the specific psychological mechanisms—particularly whether they involve singular or multiple pathways—remain inadequately explored. To address this gap, our study employed a parallel mediation model and examined whether growth mindset moderated the direct effect of emotional neglect on internet addiction, as well as the indirect effects through the two mediators, situating the issue within an ecological framework.^123,124 This allows for a systematic examination of how familial risk factors, peer influences, and individual characteristics converge to predict internet addiction among left-behind children. The key findings and their implications are discussed below.

Missing Love and Virtual Reliance: The Direct Association of Emotional Neglect

Our results confirm a direct, positive relationship between parental emotional neglect and internet addiction in LBCs. This aligns with existing literature, such as Wang et al’s¹³ work on emotional neglect and social media addiction. The underlying mechanism appears to be compensatory. When emotional needs are chronically unmet in the real world due to parental neglect, LBCs often turn to the online world as an alternative realm for psychological solace and fulfillment.¹²⁵ This behavior is consistent with Caplan’s¹²⁶ Theory of Problematic Internet Use and Kardefelt-Winther’s²⁵ Compensatory internet use model, which posit that individuals may use online environments to alleviate psychological distress. This finding serves as a critical reminder to migrant parents: beyond material provision and academic oversight, proactive and consistent emotional communication—through video calls, voice messages, and other available means—is essential for their children’s psychological well-being and serves as a protective factor against digital overuse.

Bad Company Corrupts Good Morals: Emotional Neglect is Linked to Internet Addiction via Deviant Peers

Our analysis reveals that deviant peer affiliation serves as a significant but relatively modest mediator between emotional neglect and internet addiction, accounting for approximately 4.24% of the total effect. This pathway suggests that when LBCs experience emotional neglect—characterized by insufficient parental warmth and monitoring—they may turn to deviant peer groups as alternative sources of belonging.^28,64 Once affiliated with such groups, peer pressure and social learning processes^68,70 may facilitate the adoption of excessive internet use behaviors through a “peer contagion” effect.⁷¹ This finding aligns with previous research demonstrating that deviant peer affiliation mediates the relationship between family risk factors and adolescent problematic behaviors.^127,128 However, given the limited magnitude of this indirect effect, the practical significance of this pathway should be interpreted with caution. Nevertheless, it offers a valuable insight for intervention: strengthening family connectedness and remote monitoring may help reduce LBCs’ reliance on deviant peer groups for emotional compensation.

Escaping: The Bridging Role of Experiential Avoidance Between Emotional Neglect and Internet Addiction

Our findings further establish experiential avoidance as a pivotal mediator in the link between emotional neglect and internet addiction, accounting for approximately 25.13% of the total effect, which is substantially larger than the deviant peer pathway. The model suggests that the persistent emotional pain associated with parental neglect fosters a tendency to avoid and suppress these distressing inner experiences. The internet then becomes a readily accessible and highly effective tool for this purpose, offering immersive games, videos, and social feeds that provide temporary escape from psychological pain.¹²⁹ This is supported by multiple theoretical explanations. First, experiential avoidance is a known correlate of compulsive and addictive behaviors,^130,131 making individuals prone to using the internet for emotional avoidance.¹³² Second, in the I-PACE model of addiction, experiential avoidance is conceptualized as a cognitive bias that fuels craving and addictive cycles. The substantial effect size of this pathway highlights the potential value of addressing experiential avoidance in prevention and intervention efforts. Consequently, this “escape-addiction” cycle becomes self-perpetuating, highlighting the need to build internal resilience in LBCs—helping them develop a robust psychological capacity to navigate emotional distress rather than seeking to escape it.

Why is experiential avoidance a substantially stronger mediator than deviant peer affiliation? The marked difference in effect sizes between these two pathways warrants careful theoretical consideration. We propose three interconnected explanations.

First, experiential avoidance represents a more proximal and automatic response to emotional neglect than deviant peer affiliation. When LBCs experience emotional neglect, the immediate psychological consequence is the generation of painful internal states such as loneliness, sadness, anxiety, and feelings of worthlessness.¹⁶ Experiential avoidance—the tendency to escape or suppress these aversive private experiences³⁰—is a reflexive, almost instinctive coping response that does not require external social opportunities to be activated. In contrast, deviant peer affiliation requires the presence of accessible deviant peer groups, the social opportunity to form such affiliations, and the individual’s social competence to navigate these relationships.²⁸ For many LBCs, particularly those in remote rural areas, the availability and accessibility of deviant peer groups may be limited, whereas the internal experience of emotional distress is universal and inescapable.

Second, the reinforcement schedule of experiential avoidance is more immediate and consistent than that of deviant peer affiliation. The relief obtained from escaping negative emotions through internet use is instantaneous, such as opening a game, scrolling through videos, or entering a virtual social space provides almost immediate distraction from psychological pain.^25,129 This immediate negative reinforcement (ie, removal of an aversive state) is considered a powerful correlate of addictive behavior.⁵² By contrast, the benefits of deviant peer affiliation are contingent upon successful social interaction, which requires time, effort, and reciprocation from peers.²⁶ The delayed and uncertain nature of these social rewards makes the peer pathway less potent in explaining addiction.

Third, experiential avoidance is a transdiagnostic vulnerability factor that underlies a wide range of psychopathological conditions and addictive behaviors.^30,76 Research has consistently demonstrated that experiential avoidance is a core process across various forms of addiction, including substance use disorders, gambling, and technological addictions.^130,131 In contrast, deviant peer affiliation is a more specific, context-dependent risk factor that operates primarily through social learning and peer pressure mechanisms.^27,133 The broader, more generalized nature of experiential avoidance as a vulnerability factor may explain why it exerts a more powerful influence on internet addiction than the more circumscribed pathway of deviant peer affiliation. A recent study of high school students found that experiential avoidance was a significant predictor of internet addiction, with a standardized beta coefficient of 0.43,¹³⁴ further supporting the robust association between experiential avoidance and problematic internet use.

These differences have important implications for intervention: while addressing peer influences remains valuable, particularly for LBCs who are already embedded in deviant peer networks, the substantially larger effect of experiential avoidance suggests that cognitive-emotional regulation training—such as mindfulness-based interventions and Acceptance and Commitment Therapy (ACT)—should be prioritized as a core component of prevention and intervention programs for internet addiction among LBCs.³⁰

The Specific Moderating Role of Growth Mindset

By adopting growth mindset as a moderator, we found an unexpected but theoretically informative pattern: the interaction between growth mindset and emotional neglect on internet addiction was positive and significant (β = 0.0238, p < 0.05).This indicates that a stronger growth mindset amplified rather than buffered the positive association between emotional neglect and internet addiction. It should be noted, however, that the magnitude of this interaction effect is relatively modest, and the simple slopes depicted in Figure 2 are only moderately different. This finding runs counter to the conventional assumption that growth mindset is uniformly protective and requires careful theoretical explanation. In this regard, we propose two complementary explanations for this counterintuitive finding, while cautioning against overinterpreting the practical significance of this small effect.

First, growth mindset may lose its protective function or even backfire in structurally constrained environments. For left-behind children, parental absence is a long-term, structural reality. Most migrant parents return home only briefly during the Spring Festival each year, leaving their children with little opportunity to alter this objective circumstance through personal effort.¹³⁵ When individuals with a strong growth mindset attempt to reinterpret their left-behind status as a challenge to be overcome, yet find that the situation remains fundamentally unchanged despite their efforts, they may experience a disconnect between adaptive effort and reality feedback. This disconnect can generate heightened frustration and a sense of helplessness.¹³⁶ Accumulated frustration may paradoxically strengthen avoidance motivation, driving individuals to seek refuge in the online world, where immediate feedback and a sense of control are readily available.¹³⁷ This interpretation aligns with recent theoretical advances in mindset research. Ji and colleagues¹³⁸ proposed that vulnerability and psychological affordances are two crucial boundary conditions for growth mindset effects: when individuals face vulnerability (eg, adverse circumstances) and the environment fails to provide psychological affordances that support growth mindset actions, mindset interventions are unlikely to produce meaningful effects. Similarly, Zhu and Zhao¹³⁹ demonstrated through computational modeling that growth mindset effects vanish when environmental structures disincentivize long-term investment, leading agents to converge on a harvest-only strategy. Empirical evidence from neighborhood inequality research further confirms that the positive effects of growth mindset are significantly weaker in areas characterized by greater disparities.¹⁴⁰ These findings collectively suggest that growth mindset is not a universally effective psychological resource; its benefits depend critically on supportive environmental conditions.¹⁴¹

Second, growth mindset may promote “misguided persistence” when goal direction is misaligned. The essence of growth mindset is the belief that personal attributes can be developed through effort.³⁴ However, this belief does not specify the direction of that effort. A strong growth mindset is associated with high self-efficacy and strong goal-directed behavior,⁸⁷ which are typically considered positive psychological capital. Yet when left-behind children lack external resources such as parental emotional support, this high self-efficacy may paradoxically lead them to over-rely on their own abilities to cope with stress while refusing to seek external help, such as confiding in teachers or communicating with grandparents. This self-reliance tendency, in the absence of an external support network, may push individuals toward the most accessible domain for achieving a sense of mastery: the online world.¹⁴² In the virtual environment, the logic of “effort leads to progress” is highly salient—games offer immediate feedback, level progression, and virtual social rewards. A left-behind child with a strong growth mindset may come to believe that “with more practice, I can perform better in this game”. In other words, a strong growth mindset amplifies the individual’s persistence and investment in online activities. This phenomenon—persistence based on incorrect assumptions, ineffective strategies, or misaligned goals—has been termed “misguided persistence”.^143,144 Yadav¹⁴⁵ identified a related “Growth Mindset Paradox”, arguing that a framework designed to liberate learners from fixed thinking can itself become a form of cognitive entrapment when persistence becomes divorced from effectiveness evaluation. Without structured exit conditions or reflective mechanisms to evaluate when persistence becomes counterproductive, growth mindset may inadvertently reinforce maladaptive behavioral patterns. In the context of internet addiction, high growth mindset individuals may persist in online activities precisely because they believe their efforts will yield improvement, thereby accelerating the development of addiction rather than preventing it.

These two explanations are not mutually exclusive. For left-behind children, the combination of structural constraints (the unchangeable reality of parental absence) and misaligned goal direction (effort directed toward virtual rather than constructive goals) may jointly produce the observed positive moderation effect. This finding does not invalidate the value of growth mindset as a psychological construct; rather, it highlights the critical importance of context and goal orientation in determining whether growth mindset functions as a protective or a risk-amplifying factor.

This finding makes two important theoretical contributions. First, it extends the boundary conditions of growth mindset theory. While a large body of research has demonstrated the protective effects of growth mindset across various domains,^34,35 the present study reveals a critical boundary condition: when individuals face structural, unchangeable social-environmental risks such as long-term parental migration and family structure disruption, the protective function of growth mindset may be reversed. This echoes recent meta-analytic findings that growth mindset effects are weak overall and highly context-dependent,¹⁴¹ and aligns with the Mindset × Context theory, which emphasizes that growth mindset interventions appear most effective in contexts that provide psychological affordances for growth mindset actions.¹³⁸ In the absence of such affordances, growth mindset may fail to deliver its promised benefits or even produce unintended negative consequences.

Second, this finding enriches the approach/avoidance framework by clarifying the complex role of cognitive beliefs in motivational processes. Within the approach/avoidance framework,²² growth mindset should theoretically strengthen approach-oriented tendencies toward positive goals, such as seeking help or improving academic performance, while weakening avoidance-oriented tendencies such as escaping through internet use. However, the present study demonstrates that under the specific structural risk of parental emotional neglect, the effort-persistence aspect of growth mindset can be misdirected toward the avoidance pathway. This suggests that the directionality of cognitive beliefs depends critically on the individual’s current motivational goal: if the dominant motivation is to escape pain, a growth mindset may paradoxically amplify effort directed toward “find better ways to avoid it”. This insight has important practical implications: growth mindset cultivation must be accompanied by goal clarification—that is, not only fostering the belief that “effort can bring change” but also guiding that effort toward constructive alternatives, such as building new social connections or developing offline interests, rather than allowing it to drift toward online immersion.

Direct Effect Moderation Only: Why Did Growth Mindset Not Moderate the Indirect Pathways?

An important finding of this study is that growth mindset moderated only the direct path from emotional neglect to internet addiction. The moderation of the indirect pathways through deviant peer affiliation and experiential avoidance was not significant. This pattern indicates that the result is a moderation of the direct effect rather than a moderated mediation. Importantly, the significant moderation on the direct path was positive in direction: a stronger growth mindset amplified rather than buffered the association between emotional neglect and internet addiction. However, growth mindset did not significantly moderate the indirect pathways through deviant peer affiliation or experiential avoidance. We offer the following explanation for this selective moderating pattern.

The selective moderation observed here is consistent with a front-end cognitive appraisal mechanism. Growth mindset primarily operates at the level of initial interpretation of adversity. When confronted with emotional neglect, individuals with a strong growth mindset are more likely to engage in cognitive reframing, viewing the situation as a challenge that can be managed. However, for left-behind children facing the structurally unchangeable reality of parental absence, this cognitive reframing may paradoxically intensify frustration when reality fails to respond to effort, thereby strengthening avoidance motivation toward the internet. This front-end cognitive process occurs at the point when the individual first encounters and interprets the stressor. Hence, growth mindset moderates the direct association between emotional neglect and internet addiction, albeit in a positive direction.

However, once maladaptive patterns have been established through behavioral channels such as deviant peer affiliation or emotional channels such as experiential avoidance, the moderating power of growth mindset becomes limited. This limitation can be attributed to at least three factors.

First, the immediacy of social reinforcement. The stimulation and social feedback from deviant peer affiliation are immediate and powerful. When a left-behind child is embedded in a deviant peer group, peer pressure and social learning processes operate in real time.^68,70 In contrast, the long-term growth through effort advocated by a growth mindset involves delayed, uncertain feedback. Given the imbalanced development of the adolescent reward system, the allure of immediate rewards often overpowers the consideration of long-term benefits.¹⁴⁶ Thus, even if a child holds a strong growth mindset, the immediate social rewards from deviant peers may override the cognitive belief system.

Second, the automaticity of experiential avoidance. Growth mindset is a cognitive belief, whereas experiential avoidance is an automated emotional coping habit that often operates at a subconscious or automatic level¹⁴⁷ and is typically characterized by rigidity.¹⁴⁸ The power of cognitive beliefs is often insufficient to break the cycle of deeply ingrained, habitual behaviors. Once experiential avoidance has become a default response to emotional distress, it may persist regardless of the individual’s conscious beliefs about malleability. This is consistent with the broader literature showing that automatic emotion regulation processes are relatively resistant to cognitive interventions.³⁰

Third, the mismatch between coping orientation and coping mechanism. Growth mindset promotes problem-focused coping styles, such as facing challenges directly and persisting through difficulty.⁸⁶ Experiential avoidance, by contrast, employs emotion-focused coping styles centered on escape and suppression. This fundamental difference in coping approaches means that growth mindset and experiential avoidance operate through largely independent psychological mechanisms. The non-significant moderation of the experiential avoidance pathway reflects the limited crossover between cognitive beliefs about malleability and automatic emotional avoidance habits.

In sum, growth mindset functions as a front-end cognitive factor that interacts with the initial interpretation of familial adversity. However, once adversity has given rise to behavioral patterns of deviant peer affiliation or automatic habits of experiential avoidance, the moderating role of growth mindset becomes ineffective. This selective pattern, characterized by moderation of the direct effect only, suggests that growth mindset is most relevant at the point of stressor appraisal, not at the point of established behavioral or emotional coping patterns. This distinction between direct effect moderation and moderated mediation is important for both theoretical interpretation and intervention design. The absence of moderated mediation indicates that the mediating mechanisms themselves are not influenced by growth mindset; rather, growth mindset operates primarily by altering the initial response to the stressor. This distinction has important implications for intervention: cultivating a growth mindset may be most effective as a preemptive preventive measure before maladaptive patterns have solidified. For left-behind children who have already developed deviant peer affiliations or experiential avoidance habits, more targeted interventions are necessary.

Practical Implications

Our findings, viewed through an ecological lens, highlight the interconnectedness of risk and protective factors across family, peer, and individual systems. The following implications should be interpreted as potential directions suggested by the data rather than as confirmed intervention outcomes.

At the family level, the research suggests the importance of parental emotional nourishment and remote collaboration. Parental emotional neglect emerged as the root risk factor associated with internet addiction in left-behind children. This finding suggests that migrant parents may benefit from prioritizing active, high-frequency emotional communication rather than focusing solely on material support and academic inquiries. Parents working away might make full use of modern communication tools for emotional exchanges with their children while simultaneously establishing a parenting alliance with at-home guardians to synchronize information about the child’s situation.

At the school and community level, the mediating role of deviant peer affiliation points to the potential value of constructing peer support systems. Schools and communities could consider serving as a line of defense against negative peer influence. Teachers and community workers might be trained to identify signs of undesirable friendships among left-behind children and to help them build healthy social circles through organized positive group activities and social skills courses. However, the relatively small effect size of this pathway (4.24%) suggests that interventions targeting peer relationships alone may have limited impact.

At the individual level, the findings point to a dual intervention strategy that equally emphasizes cognitive restructuring and emotional acceptance. The unexpected positive moderation effect of growth mindset, however, warrants particular caution. While growth mindset cultivation has been widely promoted as a universal protective factor, the present findings suggest that cultivating growth mindset without attention to environmental constraints and goal direction may be insufficient or even counterproductive for left-behind children facing structural adversity. This does not mean that growth mindset is inherently harmful; rather, it suggests that growth mindset interventions should be implemented with caution and ideally in conjunction with efforts to improve the actual environmental conditions that shape children’s daily experiences. For example, growth mindset training might be combined with structured goal-setting exercises that explicitly guide effort toward constructive alternatives, such as building new social connections, developing offline hobbies, or improving academic skills, rather than allowing effort to drift toward online immersion.

In addition, the key mediating role of experiential avoidance suggests that psychological therapies such as mindfulness training and Acceptance and Commitment Therapy (ACT) might be employed to help left-behind children learn to accept rather than avoid negative emotions, potentially breaking the cycle of distress-avoidance-addiction.^30,77 These approaches target the automatic emotional coping habits that our data suggest are the dominant pathway linking emotional neglect to internet addiction.

A comprehensive, multi-pronged approach is likely to be most effective: one that simultaneously enhances family emotional support, mitigates negative peer influences, fosters adaptive cognitive mindsets with clear goal direction, and builds healthy emotional regulation skills in this vulnerable population.

Limitations and Future Directions

This study has several limitations that should be acknowledged. First, its cross-sectional design precludes definitive causal inferences. Although we have framed the analyses in terms of prediction and association, the temporal ordering among variables cannot be established with certainty. Future longitudinal research is needed to unravel the temporal dynamics and causal directions among emotional neglect, deviant peer affiliation, experiential avoidance, growth mindset, and internet addiction.

Second, while the sample was drawn from major labor-exporting provinces in Western China, its representativeness could be enhanced by future studies employing random sampling and including other geographic regions. The findings may not generalize to left-behind children in other cultural or economic contexts.

Third, despite the procedural and statistical controls for common method bias, the exclusive reliance on self-report data remains a limitation. All variables were measured using self-report questionnaires completed by the same participants at a single time point. The Harman single-factor test and the first-principal-component-controlled partial correlation method provided some evidence that common method bias did not substantially affect the results. However, both approaches are indirect assessments of common method variance, and they cannot completely rule out the possibility that shared method variance may still be present. Self-report measures are also subject to recall bias, social desirability bias, and shared source variance. Therefore, common method bias cannot be fully excluded in this study. Future studies would benefit from incorporating multiple informants, such as parents, teachers, or objective behavioral indicators of internet use.

Fourth, the unexpected positive moderation effect of growth mindset, while theoretically interpretable, requires replication in independent samples before strong conclusions can be drawn. We encourage future research to examine whether similar patterns emerge in other populations facing structural constraints, such as children in poverty, children with chronic illness, or children in institutional care. Additionally, future studies might investigate the specific mechanisms through which growth mindset interacts with structural adversity, possibly using experimental or intervention designs to test the causal pathways proposed in this study.

Fifth, although students were nested within schools, the present study did not employ multilevel modeling to account for potential school-level clustering effects. This decision was primarily motivated by methodological considerations: with only 12 participating schools, the number of clusters at the school level was insufficient for reliable multilevel model estimation. Simulation studies indicate that multilevel models require a minimum of 30 to 50 clusters to produce trustworthy estimates of random effects and standard errors.^93,94 With fewer than 30 clusters, multilevel models are prone to convergence failures and biased standard error estimates.⁹⁵ Given this constraint, we adopted a design-based approach, sampling schools from diverse counties and geographic regions to enhance between-school variability. We acknowledge that it does not empirically eliminate the possibility of school-level dependence. Without directly modeling intraclass correlation, unmodeled clustering may still affect the precision of our estimates. Future research with a larger number of schools should employ multilevel modeling to formally account for school-level clustering effects and to examine whether the observed relationships vary across school contexts. This would allow researchers to partition variance into within-school and between-school components and to test whether school-level factors moderate the relationships examined in this study.

Conclusion

This cross-sectional study examined the associations between parental emotional neglect and internet addiction among Chinese left-behind children, with a focus on the mediating roles of deviant peer affiliation and experiential avoidance, as well as the moderating role of growth mindset. Our findings indicate that parental emotional neglect is positively associated with internet addiction through two distinct pathways: a behavioral approach pathway via deviant peer affiliation and an emotional avoidance pathway via experiential avoidance. Notably, experiential avoidance emerged as the substantially stronger mediator, accounting for approximately 25.13% of the total effect, compared to 4.24% for deviant peer affiliation. This marked difference suggests that the automatic and immediate tendency to escape painful internal states is a more powerful mechanism linking emotional neglect to problematic internet use than the social process of seeking belonging through deviant peers.

Contrary to the conventional assumption that growth mindset uniformly functions as a protective factor, we found that growth mindset significantly moderated the direct association between emotional neglect and internet addiction in a positive direction: a stronger growth mindset amplified rather than weakened this link. This unexpected finding suggests that in structurally constrained environments where parental absence is an unchangeable reality, growth mindset may promote misguided persistence or amplify frustration when effort fails to alter objective circumstances, thereby strengthening rather than reducing the motivation to seek refuge online. However, growth mindset did not moderate the indirect pathways through deviant peer affiliation or experiential avoidance, indicating that its influence is limited to the direct association rather than the mediating mechanisms.

These findings underscore the importance of distinguishing between direct and indirect effects when examining the role of cognitive beliefs in the context of family adversity. The results also highlight that growth mindset is not a universally protective resource; its effects depend critically on contextual factors and goal direction. Given the cross-sectional design, all findings should be interpreted as associations rather than causal mechanisms. Future longitudinal research is needed to establish temporal precedence and to examine whether the patterns observed here replicate across different populations and settings. Despite these limitations, the present study contributes to a more comprehensive understanding of the multiple pathways through which emotional neglect may be associated with internet addiction among left-behind children, and it offers empirically grounded directions for future research and intervention design.

Funding Statement

This study was jointly supported by two projects, as follows: 1. Humanities and Social Science Fund of Ministry of Education:Research on the Occurrence Mechanism and Collaborative Governance of Social Media Addiction Among Left-behind Children in Western China in the Era of Algorithm-Driven Recommendations (22YJC860022). 2. China Youth & Children Research Association Project:A Study on the Underlying Mechanisms, Multidimensional Risks, and Collaborative Governance Framework of AI Addiction Among Adolescents (2026B61).

Ethics Statement

The studies involving humans were approved by Department of Communication, School of Humanities, Southwest Jiaotong University (IRB-SWJTU-SH-034),The studies were conducted in accordance with the local legislation and institutional requirements.The participants checked the informed consent option on the questionnaire.

Written informed consent for participation in this study was provided by the participants’ legal guardians/next of kin.

Disclosure

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.