Gender‐Specific Networks in the Relationship Between Time Perspective and Short‐Form Video Addiction Among Chinese Adolescents

ABSTRACT

Short‐form video addiction (SFVA) is increasingly prevalent among adolescents and closely related to time perspective, yet their associations at the symptom and dimensional levels, particularly gender differences, remain underexplored. This study applied network analysis to examine cross‐sectional and longitudinal associations between time perspective dimensions and SFVA symptoms in Chinese adolescents, with attention to gender‐specific patterns. A total of 1, 580 Chinese adolescents (55.1% female; M _age = 15.86 ± 0.80 at T1) completed measures on two occasions at 6‐month intervals. Cross‐sectional network analysis showed that ‘future’ and ‘sleep interference’ were the central nodes in both male and female networks. In addition, ‘present‐fatalistic,’ ‘attentional disruption,’ and ‘social concern’ were identified as bridge nodes linking time perspective and SFVA. Cross‐lagged panel network analysis further highlighted the central roles of ‘future’. Furthermore, gender‐specific pathways also emerged. Among males, ‘loss of control’ functioned more as a driving symptom, whereas among females it was more strongly influenced by other symptoms. In girls, ‘social impairment’ and ‘attentional disruption’ were the main predictors of time‐related dimensions, whereas in boys, ‘sleep interference’ and ‘withdrawal’ symptoms played a more prominent role. These findings clarify the complex mechanisms linking time perspective and SFVA and highlight potential gender‐specific intervention targets.

1. Introduction

With the rapid development of digital media, short‐form video platforms such as TikTok (Douyin in China) have gained widespread popularity among adolescents, while concerns about excessive use and addiction risks have also grown. In China, over 10% of adolescents engage with short‐form videos for more than 2 h every weekday. Short‐form video addiction (SFVA) refers to the behavioural state where individuals spend excessive time on short‐form video apps despite facing negative consequences. Unlike behavioural addictions that require sustained goal‐oriented engagement (e.g., online gaming), short‐form video platforms rely on brief content, algorithmic recommendations, and an ‘endless scrolling’ structure. These features create a repetitive cycle of ‘viewing–instant gratification–continued browsing,’ which can disrupt individuals' sense of time (Yang et al. 2024). Accordingly, SFVA may be driven less by goal pursuit, as emphasised in traditional addiction frameworks (Brand et al. 2016), and more by frequent stimulation that captures attention and distorts time perception. Therefore, findings from gaming or general internet addiction may not fully explain SFVA, especially in terms of time cognition (Yang et al. 2024). Furthermore, the relationship between SFVA and time‐related cognitive structures may be bidirectional. Time cognition may shape individuals' vulnerability to short‐video use, while addictive use may also alter time‐related cognitive structures (Mao et al. 2026). Clarifying this dynamic relationship may therefore provide important insights into the cognitive mechanisms underlying SFVA.

1.1. Time Perspective and Short‐Form Video Addiction

According to the I‐PACE model (Brand et al. 2016), personality traits influence addictive behaviour by shaping individuals' cognitive and emotional responses to online applications. As an important cognitive personality trait (Zimbardo and Boyd 1999), time perspective may play a key role in SFVA. Different dimensions of time perspective may show distinct associations with SFVA. Present‐hedonistic orientation emphasises immediate pleasure and sensation‐seeking, which closely matches the instant rewards and continuous stimulation provided by short videos (Xia et al. 2023). This may strengthen the tendency to keep scrolling. Present‐fatalistic orientation reflects feelings of helplessness and low perceived control, making it difficult for individuals to reduce use even when they recognise negative consequences (Chittaro and Vianello 2013). Future‐oriented individuals focus more on long‐term goals and delayed rewards, making them more likely to resist the immediate temptations of short videos (He et al. 2024). Past‐negative orientation is linked to negative emotions, which may increase the likelihood of using short videos as an escape strategy (Chittaro and Vianello 2013). Past‐positive orientation is associated with positive emotions and psychological resources, which may buffer against addiction (Zimbardo and Boyd 1999).

The I‐PACE model also highlights feedback processes. Addictive behaviours may reshape cognitive structures over time (Brand et al. 2016). Although time perspective is relatively stable, it can change across development (Mohammed and Marhefka 2020). Adolescence is a period of high neurocognitive plasticity. Therefore, SFVA may influence the development of time perspective (Mao et al. 2026). This is supported by a recently longitudinal study (Xiong et al. 2026), which found that SFVA predicts time perspective 6 months later. Short‐form videos platforms provide rapid content switching, immediate feedback, and immersive experiences. Repeated exposure to such stimulation may strengthen present‐oriented tendencies while weakening future planning and self‐control (e.g., Nong et al. 2023). Furthermore, the sense of loss of control resulting from addictive behaviour may reinforce a sense of hopelessness about the future, thereby strengthening present‐fatalistic orientation. Compared with present and future orientations, past‐positive and past‐negative orientations may change more slowly because they are shaped more strongly by accumulated life experiences (Zimbardo and Boyd 1999). Therefore, within a 6‐month period, SFVA may have stronger effects on present‐hedonistic, present‐fatalistic, and future orientations than on past‐related dimensions.

1.2. Limitations of Existing Research

Despite increasing attention to SFVA and time perspective, many studies treat both constructs as unitary variables (e.g., He et al. 2024; Mao et al. 2026). This assumes that all time perspective dimensions function similarly, which is problematic. According to Zimbardo and Boyd (1999), the five dimensions are distinct. For example, present‐hedonistic and future orientations may involve different cognitive and emotional processes, which could contribute to different patterns of SFVA symptoms. Analyses at the total‐score level cannot capture these differences. Furthermore, SFVA also differs from gaming addiction. Gaming involves long‐term goals, whereas short‐video use is characterised by continuous browsing and instant rewards (Yang et al. 2024). This suggests that SFVA may be more closely related to present‐oriented dimensions. This hypothesis needs to be tested at the dimension–symptom level.

Additionally, most existing research is based on a ‘unidirectional causality’ assumption (e.g., ‘time perspective → addiction’) or cross‐sectional data, lacking longitudinal evidence to reveal the temporal associations between SFVA and time perspective. More recently, the Cross‐Lagged Panel Network (CLPN) combines network analysis with cross‐lagged panel models to further elucidate the longitudinal dynamics of psychological structures over time, thereby addressing this gap. Clarifying the longitudinal relationships between SFVA symptoms and time perspective dimensions over time can help uncover the underlying mechanisms of addiction and identify more effective intervention targets.

1.3. Gender Differences

A comprehensive understanding of these dynamics also requires consideration of gender differences. Previous research has revealed gender‐specific patterns in short‐video usage, yet findings remain mixed—some studies have found that males have a higher frequency of use and addiction (e.g., Liu et al. 2025), while other studies have reported that females are more prominent in problem usage behaviour (e.g., Su et al. 2020). Regarding time perspective, females tend to be more future‐oriented, while males are generally more present‐oriented (Li et al. 2024). However, conclusions regarding past‐negative orientation are inconsistent (e.g., Laghi et al. 2015). Together, these findings suggest that gender differences may exist in both the overall levels and specific dimensions of time perspective and SFVA.

In addition, the relationships between variables may differ by gender. Drawing on findings from the broader field of internet addiction, risk pathways that enhance addictive tendencies may be more pronounced in males, while protective pathways that inhibit such tendencies may be more prominent in females (Li et al. 2010). At the same time, negative psychological outcomes of addiction may be more severe among females (Ko et al. 2014). These findings suggest that the links between time perspective and SFVA may vary by gender. However, direct evidence is still limited. Therefore, gender differences were explored in the present study in an exploratory manner rather than through specific directional hypotheses.

1.4. The Present Study

In summary, although existing research indicates a close relationship between SFVA and time perspective, there remains a lack of longitudinal empirical evidence regarding their cross‐time associations at the symptom‐dimension level and the associated gender differences. To address this, the present study used a two‐wave longitudinal design, combining cross‐sectional networks and CLPN analyses, aiming to systematically explore the bidirectional relationship between time perspective and SFVA and the moderating role of gender. The specific objectives are: (a) to map the cross‐sectional association network between time perspective and SFVA; (b) to explore their longitudinal interactions; and (c) to examine the moderating role of gender in this relationship.

2. Methods

2.1. Participants and Procedure

The present study adopted a convenience sampling approach to recruit adolescents from a senior secondary school in Shenzhen, China. Data were collected at two time points 6 months apart: the first wave (T1) in October 2024 (N = 1978 valid responses) and the second wave (T2) in March 2025 (N = 1784 valid responses). Adolescents who did not complete both waves were excluded, resulting in a final matched sample of 1580 students (55.1% girls; M _age = 15.86 years, SD_age = 0.80). Attrition analyses showed no significant differences between retained participants and dropouts in baseline SFVA or addictive behaviour (t = −0.85, p = 0.40), nor in any dimensions of time perspective (ps = 0.24–0.77), suggesting that attrition was unlikely to substantially bias the sample.

Prior to participation, all students were informed of the study's objectives, confidentiality agreements, and the voluntary nature of participation, including the right to withdraw at any time without penalty. Written informed consent was obtained from both students and their legal guardians. The data collection process comprised three main steps: (1) the electronic questionnaire was designed and uploaded via the professional survey platform ‘Wenjuanxing’; (2) the link was distributed uniformly by school mental health teachers; and (3) students completed the survey during computer classes under the supervision of trained research assistants and school psychologists. Ethical approval for this study was granted by the Institutional Review Board of the first author's university. To ensure the accuracy of longitudinal data matching, each participant was assigned a unique identification code. Additional details about the sampling procedure are provided in Section S1.

The six‐month interval was selected based on recommendations from the optimal time‐lag framework (Dormann and Griffin 2015), which suggests that appropriate intervals should match the temporal stability and developmental pace of the target constructs. Given the relatively stable nature of time perspective, a 6‐month interval was considered suitable for capturing meaningful psychological changes during adolescence. This choice was also consistent with previous longitudinal studies on time perspective and smartphone addiction/SFVA among Chinese adolescents that adopted 6‐month intervals (e.g., Mao et al. 2026; Xiong et al. 2026). In addition, the interval aligns with the academic semester cycle of Chinese senior secondary school students.

2.2. Measures

2.2.1. Time Perspective

The Chinese version of the Zimbardo Time Perspective Inventory (Z TPI‐C; Zimbardo and Boyd 1999; Li et al. 2022) was used to assess adolescents' time perspective. Previous studies have demonstrated that the Z TPI‐C exhibits good reliability and is suitable for application within the Chinese cultural context (Li et al. 2022). The ZTPI‐C consists of 25 items encompassing five dimensions: Past Negative (6 items), Present Hedonistic (4 items), Future (5 items), Past Positive (7 items), and Present Fatalistic (3 items). Responses are rated on a 5‐point Likert scale, ranging from 1 (strongly disagree) to 5 (strongly agree). In the present study, the Cronbach's α for the five subscales at T1 and T2 ranged from 0.790 to 0.904.

2.2.2. Short‐Form Video Addiction

Adolescents' short‐form video addiction was measured using the Short‐Form Video Addiction Scale (SFVAS), an instrument adapted from the Social Network Service Addiction Scale developed by Choi and Lim (2016). The SFVAS consists of six items rated on a 5‐point Likert scale (1 = ‘strongly disagree’ to 5 = ‘strongly agree’). Average item scores were calculated, with higher values indicating more severe addiction tendencies. Prior studies have confirmed the SFVAS as a psychometrically sound tool for use with Chinese adolescents (Yang et al. 2025). In the current sample, the Cronbach's α for the scale was 0.863 and 0.896 at T1 and T2, respectively.

2.3. Data Analysis

Descriptive analyses, cross‐sectional Glasso network, and cross‐lagged panel network analyses were conducted in R (version 4.4.3). Before the main analyses, longitudinal and gender‐based measurement invariance tests were conducted for SFVA and time perspective. The results supported scalar invariance across time and gender. Detailed results are reported in Tables S1 and S2. In addition, the final sample size (N = 1580) was considered adequate for the present network analyses with 11 nodes and supported stable edge estimation and interpretable centrality indices (Epskamp et al. 2018).

2.3.1. Descriptive Analyses

Prior to conducting network analyses, independent samples t‐tests were performed to examine gender differences in all key variables included in the study. This preliminary step aimed to determine whether boys and girls significantly differed in levels of SFVA and time perspective dimensions. Due to group differences in age, age was controlled as a covariate in all subsequent analyses.

2.3.2. Cross‐Sectional Symptom Network Analysis

To investigate the associations between SFVA and time perspective among girls and boys, we estimated two cross‐sectional symptom networks (CSSNs) using first‐wave data (T1). Network estimation and visualisation were conducted using the R package qgraph (version 1.9.8), applying the Extended Bayesian Information Criterion graphical least absolute shrinkage and selection operator (EBICglasso). In each network, nodes represented dimensions of SFVA and time perspective, while edges corresponded to pairwise partial correlations between nodes. The thickness of each edge reflects the strength of the association, with green edges indicating positive correlations and red edges indicating negative correlations.

To identify the most central nodes within each network, we calculated node centrality using the Expected Influence (EI) metric. In addition, to detect bridge nodes that serve as key connectors between communities, we estimated Bridge Expected Influence (BEI) using the bridge function from the NetworkTools package (version 1.4.0). In line with prior research, the top three nodes ranked by EI were interpreted as the most central, and the top three nodes ranked by BEI were regarded as key bridge symptoms.

2.3.3. Cross‐Lagged Panel Network Analysis

To investigate the longitudinal dynamics between SFVA and time perspective, two cross‐lagged panel networks (CLPNs) were estimated from baseline (T1) to follow‐up (T2) separately for girls and boys using the R package glmnet (version 4.1.8). The analyses employed least absolute shrinkage and selection operator (Lasso) regularisation with 10‐fold cross‐validation to estimate both autoregressive and cross‐lagged effects.

Consistent with the cross‐sectional Glasso networks, nodes represent variables, and the thickness and colour of edges indicate the strength and direction (positive or negative) of associations. However, directed arrows were included to illustrate the directionality of cross‐lagged effects. Moreover, centrality metrics that namely cross‐lagged ‘in’ Expected Influence (IEI) and ‘out’ Expected Influence (OEI) were calculated. IEI quantifies the extent to which a node at follow‐up is predicted by other nodes at baseline, while OEI reflects how strongly a baseline node predicts other nodes at follow‐up. In this study, nodes ranking in the top three for OEI and IEI were identified as the most central within the network.

2.3.4. Network Stability and Accuracy

The R package bootnet (version 1.4.3) was utilised to assess the stability and accuracy of both the cross‐sectional and cross‐lagged panel networks. First, nonparametric bootstrapping was performed to generate confidence intervals (CIs) for edge weights, where narrower CIs indicate greater estimation precision and stability. Second, the case‐dropping bootstrap procedure was applied to calculate the correlation stability coefficient (CS‐C) for centrality metrics, including Expected Influence (EI), In Expected Influence (IEI), Out Expected Influence (OEI), and Bridge Expected Influence (BEI). A CS‐C value exceeding 0.25 was considered indicative of acceptable stability.

2.3.5. Network Comparison

To compare the CSSNs between girls and boys, we conducted three tests using the R package NetworkComparisonTest (version 2.2.2): the network structure invariance test, the global strength invariance test, and the edge strength invariance test. The network structure invariance test assessed differences in the overall distribution of edge weights between groups. The global strength invariance test evaluated group differences based on the absolute sum of all edge weights. Lastly, the edge invariance test examined differences at the level of individual edges within the networks.

3. Results

3.1. Descriptive Statistics

Table 1 outlines the detailed descriptions of each item along with their associated node labels. Summary statistics and results from between‐group analyses for all study variables assessed at T1 are displayed in Table 2.

TABLE 1.

Network nodes of the study.

Cluster	Node	Item	Abbreviation
Short video addiction	SVA1	Watching short videos makes it hard for me to concentrate on studying.	S‐Attentional Disruption
	SVA2	Watching short videos takes up a lot of my sleep time and makes it hard for me to sleep well.	S‐Sleep Interference
	SVA3	Watching short videos has negatively affected my social activities.	S‐Social Impairment
	SVA4	My family and friends think I spend too much time watching short videos.	S‐Social Concern
	SVA5	I feel uncomfortable if I can't watch short videos.	S‐Withdrawal
	SVA6	I try to spend less time watching short videos, but I can't do it.	S‐Loss of Control
Time Perspective	TP1	e.g., Painful past experiences keep replaying in my mind.	T‐Past‐Negative
	TP2	e.g., Familiar childhood scenes, sounds, and smells often evoke fond memories for me.	T‐Past‐Positive
	TP3	e.g., I act impulsively.	T‐Present‐Hedonistic
	TP4	e.g., Because current events are so unpredictable, it's impossible to plan for the future.	T‐Present‐Fatalistic
	TP5	e.g., When I want to accomplish something, I set goals and think about specific ways to achieve them.	T‐Future

Abbreviations: S, short video addiction; T, denotes time perspective.

TABLE 2.

Descriptive statistics (mean ± SD) by gender with t‐test.

Variable	Overall	Boys	Girls (N = 870)	t
	(N = 1580)	(N = 710)
Age	15.86 ± 0.80	15.95 ± 0.81	15.79 ± 0.79	3.91***
SVA1	3.12 ± 1.09	3.12 ± 1.15	3.12 ± 1.05	−0.09
SVA2	2.44 ± 1.11	2.54 ± 1.17	2.36 ± 1.04	3.14**
SVA3	2.03 ± 1.03	2.18 ± 1.14	1.90 ± 0.91	5.19***
SVA4	2.47 ± 1.15	2.55 ± 1.22	2.40 ± 1.09	2.66**
SVA5	2.26 ± 1.12	2.31 ± 1.19	2.22 ± 1.06	1.42
SVA6	2.32 ± 1.12	2.37 ± 1.18	2.27 ± 1.07	1.87
TP1	23.42 ± 4.76	23.31 ± 5.26	23.51 ± 4.30	−0.79
TP2	22.86 ± 5.35	22.41 ± 5.71	23.23 ± 5.02	−3.01**
TP3	9.22 ± 2.62	9.07 ± 2.80	9.34 ± 2.46	−2.06*
TP4	11.97 ± 3.16	11.59 ± 3.36	12.29 ± 2.94	−4.34***
TP5	16.99 ± 3.52	16.91 ± 3.68	17.05 ± 3.39	−0.77

^* p < 0.05.

^** p < 0.01.

^*** p < 0.001.

3.2. Cross‐Sectional Symptom Network of SFVA and Time Perspective Among Boys and Girls

3.2.1. Network Estimation and Inference

To examine the associations between SFVA and time perspective in girls and boys, we constructed two CSSNs using first‐wave data (T1). Each network comprised 55 edges in total, with 37 and 35 having nonzero weights for girls and boys, respectively (see Tables S3 and S4), and networks with bridge nodes are shown in Figure 1.

FIGURE 1.

Cross‐sectional symptom networks (CSSNs) structures. (A) Girls. (B) Boys.

Figure 2A,B shows the node centrality, measured by Expected Influence (EI), within the symptom networks for both girls and boys. Across both groups, ‘T‐Future’ (TP5) emerged as the most central node (EI = −2.54 in girls, −2.00 in boys). ‘S‐Sleep Interference’ (SVA2) was also identified as a key central node in both networks, ranking second in the boys' network (EI = 1.58) and third in the girls' network (EI = 1.11). For girls, the third most central node was ‘S‐Social Concern’ (SVA4; EI = 1.01), whereas for boys, it was ‘T‐Past‐Positive’ (TP2; EI = −1.13).

FIGURE 2.

Centrality indices (expected influence, EI; A, B) and bridge centrality indices (bridge expected influence, BEI; C, D). (A, C) Girls. (B, D) Boys.

Figure 2C,D presents the bridge centrality, defined as Bridge Expected Influence (BEI), of each node within the SFVA and time perspective CSSNs for both genders. Both groups shared two key bridge nodes: ‘T‐Present‐Fatalistic’ (TP4; BEI = 0.17 for girls, 0.23 for boys) and ‘S‐Attentional Disruption’ (SVA1; BEI = 0.14 for girls, 0.18 for boys). And ‘T‐Future’ (TP5; BEI = −0.15) was the second most bridge node for girls, while ‘S‐Social Concern’ (SVA4; BEI = 0.10) emerged as the third most bridge node for boys.

3.2.2. Network Accuracy and Stability

Results from the nonparametric bootstrapping analyses demonstrated that the confidence intervals (CIs) around all edge weights were relatively narrow (Figure S1), indicating robust estimates. In addition, the majority of pairwise comparisons of edge weights (Figure S2) and node centrality measures (Figure S3) reached statistical significance in both girls and boys, supporting the reliability of the findings. Case‐drop bootstrapping further validated the stability of Expected Influence and Bridge Expected Influence, with CS‐coefficients of 0.67 and 0.75 for girls, and 0.59 and 0.75 for boys, respectively (Figure S4).

3.2.3. Network Comparison

We compared the network structures of girls and boys. The overall network strength was almost the same in both groups (girls = 4.75, boys = 4.76; S = 0.012, p = 0.97; see Figure S5A). However, the overall network structure showed a significant difference (M = 0.23, p < 0.001; Figure S5B). The results of the edge invariance tests, which examine differences in specific connections between nodes, are shown in Table S5. Focusing on the bridging edges between the two communities, the connection between ‘S‐Sleep Interference’ (SVA2) and ‘T‐Present‐Fatalistic’ (TP4) was significantly stronger for girls (p = 0.03). In contrast, the edge between ‘S‐Loss of Control’ (SVA6) and ‘T‐Present‐Fatalistic’ (TP4) was significantly stronger in the boys' network (p = 0.03).

3.3. Cross‐Lagged Panel Networks of SFVA and Time Perspective Among Boys and Girls

3.3.1. Network Estimation and Inference

To investigate the longitudinal associations between SFVA and time perspective in girls and boys, we estimated two cross‐lagged panel networks (CLPNs) from baseline to follow‐up. Each network contained 121 possible edges, with 92 being nonzero in the girls' network and 66 in the boys' network. The resulting network structures are presented in Figure 3A,C.

FIGURE 3.

Cross‐lagged panel networks (CLPNs) structures and corresponding centrality indices. (A) CLPN structure for girls. (B) Centrality indices (out‐expected influence, OEI; in‐expected influence, IEI) for girls. (C) CLPN structure for boys. (D) Centrality indices (OEI and IEI) for boys.

Figure 3B,D present the centrality indices, specifically, Out Expected Influence (OEI) and In Expected Influence (IEI)—for each node in the cross‐lagged panel networks (CLPNs) of SFVA and time perspective among girls and boys. Regarding OEI, ‘T‐Future’ (TP5; OEI = −1.83 for girls, −1.76 for boys), and ‘T‐Present‐Hedonistic’ (TP3; OEI = −1.21 for girls, −1.38 for boys) exhibited the highest out‐centrality across both groups. Additionally, ‘S‐Sleep Interference’ (SVA2; OEI = 1.42 for girls, 1.09 for boys) exhibited greater out‐centrality in girls, while ‘S‐Loss of Control’ (SVA6; OEI = 1.52 for boys, 0.69 for girls) was more prominent in the boys' network. Concerning IEI, ‘T‐Future’ (TP5; IEI = −2.21 for girls, −1.56 for boys) and ‘T‐Past‐Positive’ (TP2; IEI = −1.56 for girls, −1.81 for boys) were the most influential nodes regarding incoming effects in both genders. Notably, ‘S‐Loss of Control’ (SVA6; IEI = 1.04) exhibited greater in‐centrality in girls, while ‘S‐Attentional Disruption’ (SVA1; IEI = 0.99) was more influential in the boys' network.

Regarding the longitudinal associations between SFVA and time perspective, several predictive pathways emerged. Among girls, ‘S‐Attentional Disruption’ significantly predicted ‘T‐Present‐Fatalistic’ (SVA1 → TP4, β = 0.12), while ‘S‐Social Impairment’ was a strong predictor of both ‘T‐Past‐Negative’ (SVA3 → TP1, β = 0.09) and ‘T‐Present‐Hedonistic’ (SVA3 → TP3, β = 0.09). In boys, ‘S‐Sleep Interference’ negatively predicted ‘T‐Past‐Positive’ (SVA2 → TP2, β = −0.20), whereas ‘S‐Withdrawal’ showed significant negative associations with both ‘T‐Future’ (SVA5 → TP5, β = −0.19) and ‘T‐Past‐Positive’ (SVA5 → TP2, β = −0.14).

3.3.2. Network Accuracy and Stability

Nonparametric bootstrapping revealed that the confidence intervals (CIs) for all edge weights were narrow (Figure S6), suggesting precise estimations. Additionally, the majority of pairwise comparisons of edge weights (Figure S7) and node centrality indices (Figure S8) reached statistical significance in both girls and boys. Case‐drop bootstrapping further demonstrated good stability for both In Expected Influence (IEI) and Out Expected Influence (OEI), with CS‐coefficients of 0.67 across groups (Figure S9).

4. Discussion

This study is the first to use a two‐wave longitudinal network approach to examine the longitudinal associations between time perspective and short‐form video addiction (SFVA) among Chinese adolescents, with a specific focus on gender differences. By integrating cross‐sectional and cross‐lagged panel networks, we identified both core symptoms and key temporal pathways linking time perspective to SFVA. These findings provide new insights into the mechanisms of SFVA and highlight potential gender‐specific intervention targets.

4.1. Core Symptoms and Bridge Nodes in the Cross‐Sectional Network

The cross‐sectional network analysis revealed the static association patterns within the system. ‘Future’ showed the strongest centrality in both males and females, suggesting that it may occupy a pivotal position within the interplay between time perspective and SFVA. Individuals with high future orientation typically focus more on long‐term goals and behavioural consequences, and thus are less likely to indulge in the fleeting gratification provided by instant entertainment (Zimbardo and Boyd 1999). Previous studies have similarly found that future orientation is closely associated with lower levels of problematic internet use and addiction risk (e.g., Kim et al. 2017; Przepiorka and Blachnio 2016), suggesting it may serve as an important protective factor against short‐video addiction. In addition, ‘sleep interference’ also exhibited relatively high centrality, indicating that it is not only a significant manifestation of short‐video addiction but may also play a central role in maintaining the entire symptom network. Previous research has indicated that sleep problems further impair an individual's cognitive control, potentially creating a vicious cycle of ‘impaired sleep—decreased self‐control—excessive use,’ which may explain its central position in the network (Chen et al. 2019; Wu and Yao 2026).

In addition, ‘present‐fatalistic’ and ‘attentional disruption’ served as bridge nodes linking time perspective and SFVA in both gender networks. ‘Present‐Fatalistic’ reflects feelings of helplessness and fatalistic beliefs about the present, which may motivate adolescents to use short videos as a form of emotional escape (Przepiorka and Blachnio 2016). ‘Attentional disruption’ may weaken adolescents' ability to maintain long‐term goals, thereby influencing their time perspective (Liu et al. 2022). Therefore, interventions targeting these bridge symptoms may help reduce the connections between time perspective and SFVA. However, cross‐sectional networks cannot capture directional changes between symptoms over time. To address this limitation, the present study further employed cross‐lagged panel network analysis.

4.2. The Longitudinal Associations Between Time Perspective and SFVA

The cross‐lagged panel network (CLPN) provided empirical support for the longitudinal relationship between time perspective and SFVA. In the pathway from SFVA to time perspective, ‘sleep interference’ emerged as the core symptom driving changes across the system. In the Chinese context, adolescents are already highly sensitive to sleep deprivation (Short et al. 2013), and high academic pressure combined with early school start times (typically 7:30 a.m.) leads to chronic sleep restriction. Against this background, excessive short‐video use may further reduce sleep time, particularly through delayed bedtime behaviours. Over time, this may impair executive control and emotional regulation, gradually weakening individuals' time‐related cognitive structures (Xiao et al. 2026). The finding suggests that sleep‐related problems may play an important role in the longitudinal maintenance of SFVA. In addition, ‘future’ and ‘past‐positive’ were the nodes most strongly influenced by other variables. This suggests that adaptive temporal orientations may be more likely to fluctuate under the influence of persistent SFVA‐related experiences.

In the pathway from time perspective to SFVA, ‘future’ and ‘present‐hedonism’ emerged as key predictive dimensions in both genders. A stronger ‘future’ orientation generally reflects a greater ability to balance long‐term goals against immediate temptations. As a result, adolescents with higher future orientation may be less likely to rely on the instant gratification provided by short‐video platforms. In contrast, a stronger ‘present‐hedonistic’ orientation reflects a preference for immediate rewards and emotional pleasure. This tendency may increase dependence on the rapid feedback mechanisms embedded in short‐video platforms. These findings suggest that adolescent SFVA is not merely a problem of media use itself, but is also closely related to individual differences in time‐related cognitive styles.

4.3. Gender Differences

Furthermore, the results also revealed gender differences in the longitudinal dynamics between SFVA and time perspective. First, ‘loss of control’ exhibited a significant gender‐specific role. Among males, it had a higher outward expected influence, meaning it could predict other nodes over time. In contrast, among females, it showed higher in‐expected influence, indicating that its changes were more strongly predicted by other nodes. This pattern is partly consistent with previous findings suggesting that males tend to exhibit stronger impulsivity and externalising tendencies in addictive behaviours (e.g., Cross et al. 2011). As a result, ‘loss of control’ may function more as a driving symptom in males. In contrast, females are more likely to display internalising patterns, in which psychological experiences are more closely interconnected (McLean and Anderson 2009). Therefore, ‘loss of control’ in females may be more embedded within an interconnected symptom system.

Second, among girls, ‘social impairment’ and ‘attentional disruption’ were the primary predictors of time‐related dimensions. This is consistent with previous research indicating an interaction effect between social relationships and time perspective in females (Yeung et al. 2007). When SFVA leads to social impairment, individuals may perceive a disruption in their real‐world social support systems. Stable social connections serve as the psychological foundation for maintaining ‘future continuity’, which may influence females' perception of time. Furthermore, ‘imagining the future’ and ‘processing the present’ share a set of cognitive resources. Frequent attention interruption may consume cognitive resources, thereby affecting females' sense of time.

Among boys, ‘sleep interference’ and ‘withdrawal’ symptoms were the primary predictors of time‐related dimensions. This finding aligns with Cross et al. (2011)'s observation that males may be more prone to impulse control deficits. When males excessively engage with short videos, their diminished self‐regulation makes it harder for them to interrupt the behaviour, leading them to sacrifice sleep for immediate gratification and thereby disrupting their perception of time. At the same time, the intense cravings associated with withdrawal further reinforce a preference for immediate feedback, making males more likely to become trapped in the immediate reward cycle of short videos. However, it should be noted that these interpretations are primarily based on existing theoretical frameworks. The specific mechanisms underlying these gender differences still require further investigation.

Despite differences in specific pathways, network comparison analyses showed no significant gender difference in overall network strength. This suggests that the overall association between time perspective and SFVA is similar across genders. Gender differences may lie more in how these relationships operate rather than in their overall magnitude.

4.4. Implications

Theoretically, this study demonstrates the value of network analysis in capturing longitudinal associations between time perspective and addiction symptoms, refining addiction models. The observed gender differences further highlight the importance of integrating social gender and developmental factors into theoretical frameworks. Practically, two intervention strategies are recommended. First, interventions targeting maladaptive temporal cognition—particularly by strengthening future orientation—may reduce adolescents' vulnerability to SFVA. Second, targeting central and bridge symptoms, such as sleep interference, attention control, and social functioning, may weaken the cross‐time connections between time perspective and SFVA. Moreover, gender‐specific strategies should be prioritised: for boys, interventions should focus on improving sleep quality and impulsive behaviour (e.g., self‐regulation interventions for sleep and mindfulness training for SFVA); for girls, emphasis should be placed on attention training, social skills development, and adjusting fatalistic mindsets. Developing personalised, gender‐informed interventions can significantly improve their effectiveness.

4.5. Limitations and Future Directions

Although this study makes important contributions, several limitations should be noted. First, only two waves of data with a six‐month interval were used, limiting inferences about long‐term dynamics and causal stability. Future studies should include more measurement points and longer follow‐ups. Second, participants were recruited from a single school using convenience sampling, which may limit the generalizability of the findings. Future studies should recruit more diverse and representative samples. Third, the sample consisted only of Chinese adolescents, and the findings have not been validated in other cultural contexts (e.g., Western adolescents), where cultural background and platform use may differ. Future research should replicate these findings across cultures and age groups. Third, all data were self‐reported, which may introduce recall and social desirability biases. Future studies should incorporate multi‐method approaches, such as behavioural measures, physiological indicators (e.g., sleep monitoring), and experimental paradigms.

5. Conclusions

This study is the first to use a network analysis to examine the gender‐specific relationship between time perspective and SFVA among Chinese adolescents. The results indicate that ‘future’ time perspective and ‘sleep interference’ hold central positions in the network, while ‘present‐fatalistic,’ ‘attentional disruption,’ and ‘social concern’ serve as key bridging factors between time perspective and SFVA. Furthermore, the results revealed a longitudinal predictive association between dimensions of time perspective and symptoms of SFVA. Gender differences primarily manifest in variations in the importance of symptoms and predictive pathways. Overall, these findings not only enhance our understanding of the developmental mechanisms linking adolescents' time perspectives and SFVA but also provide empirical evidence for intervention strategies, encompassing both universal objectives (e.g., enhancing future time perspective) and gender‐specific approaches (e.g., social–emotional support and repair for girls, and behaviour and sleep management for boys).

Author Contributions

Yuyin Wang: conceptualization, funding acquisition, project administration, supervision, writing – review and editing. Xiaohui Luo: conceptualization, formal analysis, writing – original draft, writing – review and editing. Yuqin Deng: writing – review and editing. Yunhong Shen: conceptualization, data curation, investigation, methodology, writing – original draft, writing – review and editing. Yuqiong Pang: methodology, writing – original draft, writing – review and editing. Ruiqi Wang: writing – review and editing.

Funding

This study was supported by the Ministry of Education of China Humanities and Social Sciences Research Project (Grant No. 22YJAZH107 & 23YJA190007), the Guangdong Basic and Applied Basic Research Foundation (Grant No. 2023A1515030200). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Ethics Statement

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional review board of the authors' university and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Consent

Informed consent was obtained from all participants included in the present study.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Bootstrapped confidence intervals for all edge weights. The red dots represent the observed values of each edge weight in the network, and the black dots represent the bootstrapped mean estimates, both ordered from the highest to the lowest. The grey area indicates the 95% confidence intervals estimated using a non‐parametric bootstrap procedure. Wider intervals indicate lower stability, while narrower intervals indicate higher stability. (A) CSSN of the girls. (B) CSSN of the boys.

Figure S2: Bootstrapped stability test for edge weights. This figure shows the bootstrapped difference tests (α = 0.05) for edge weights. The colour of the boxes indicates whether the edge weights differ significantly from each other (black) or not (grey). The diagonal line represents the strength of edge weights, ranging from red (negative associations), to white (weaker associations), and to blue (stronger associations). (A) CSSN of the girls. (B) CSSN of the boys.

Figure S3: Bootstrapped stability test for node strength. This figure shows the bootstrapped difference tests (α = 0.05) for node strength. The colour of the boxes indicates whether there is a significant difference between nodes: grey boxes indicate no significant difference, while black boxes indicate a significant difference. The numbers in the white boxes along the diagonal represent the node strength values of the corresponding nodes. (A) CSSN of the girls. (B) CSSN of the boys.

Figure S4: The x‐axis represents the proportion of cases from the original sample retained at each step. The y‐axis represents the average correlation between the centrality indices of the original network and those from networks re‐estimated after progressively excluding larger proportions of the sample. (A) CSSN of the girls. (B) CSSN of the boys. Figure S5. The NCT (Network Comparison Test) results. (A) Global strength invariance between girls and boys in the CSSNs. (B) Structure invariance between girls and boys in the CSSNs.

Figure S6: Bootstrapped confidence intervals for all edge weights in the CLPNs. (A) Girls. (B) Boys.

Figure S7: Bootstrapped stability test for edge‐weight in the CLPNs. (A) Girls. (B) Boys.

Figure S8: Bootstrapped stability test for node strength in the CLPNs. (A, C) Girls. (B, D) Boys.

Figure S9: The x‐axis represents the proportion of cases from the original sample retained at each step. The y‐axis represents the average correlation between the centrality indices of the original network and those of networks re‐estimated after progressively excluding larger proportions of the sample. (A) CLPN for girls. (B) CLPN for boys.

Table S1: Fit Indices for Longitudinal Measurement Invariance Testing Across Time (T1 and T2).

Table S2: Fit Indices for Measurement Invariance Testing Across Gender (Males and Females).

Table S3: Partial correlation coefficients between undirected edges in T1 for girls (N = 810).

Table S4: Partial correlation coefficients between undirected edges in T1 for boys (N = 710).

Table S5: Edge strength invariance test of CSSNs between girls and boys.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.