The topological structure of adolescents’ internet adaptation: A longitudinal tracking study

doi:10.3724/SP.J.1041.2025.0415

Abstract

Abstract:

As the saying goes, “Survival of the fittest”. Nowadays, the Internet has become a critical channel for information acquisition, social interaction, and educational learning. Adolescents’ internet adaptation capabilities must be continuously improved to adapt to this rapidly developing information age. Internet adaptation is inherently a “multidimensional system” encompassing various stages and dimensions. However, there remains a gap in the research exploring the internal topological characteristics and functional mechanisms of internet adaptation. Consequently, this study aims to employ network analysis techniques to elucidate the core characteristics, internal structure, dynamic evolution, and relationships with external variables of adolescents’ internet adaptation through network analysis. This approach will offer a comprehensive framework for understanding adolescents’ successful adaptation in the digital age and provide scientific insights for preventing and intervening in adolescent internet addiction.

This study collected all data through paper-and-pencil questionnaires. At Time 1, valid data were obtained from 5783 participants (Males for 37.4%, M_age = 17.20 years, SD = 2.62). Five months later, data from 1235 of these participants were tracked (Males for 38%, M_age = 14.98 years, SD = 1.66). Based on the research objectives, we conducted cross-sectional network analysis, network comparison, and cross-lagged network analysis. All cross-sectional and cross-lagged network analyses were primarily conducted using R (V.4.3.2). Network visualizations were created with the qgraph package (version 1.9.5). The accuracy of edge estimates was assessed by performing 1000 bootstrap iterations to construct 95% non-parametric bootstrap confidence intervals for each edge.

In the cross-sectional network of internet adaptation, “Internet curiosity” is the node with the highest strength (1.18). Network comparison results indicate no significant difference in the overall strength between the T1 (3.52) and the T2 network (3.79) (p = 0.120), although the network invariance test result is significant (p < 0.001). The cross-lagged network analysis shows that “Internet self-efficacy” has the strongest out-expected influence (0.60), “Internet learning ability” and “Internet information searching” has the strongest in-expected influence (0.31 & 0.30). Additionally, the cross-lagged network analysis of internet adaptation and internet addiction reveals that “Internet information protection capability” exhibits the strongest outgoing predictive ability.

The main conclusions are as follows: (1) Adolescent internet adaptation is characterized by its dynamic and staged nature; (2) Adolescents’ internet curiosity plays a multifaceted role in their internet adaptation process: insufficient curiosity can lead to low internet self-efficacy, while excessive curiosity can result in poor internet self-control; (3) Internet self-efficacy has the most significant impact on the overall development of internet adaptation, serving as the “primary driving force”; (4) Internet learning ability and internet information search receive the most internal influence, constituting the main “landing point” of adolescents’ internet adaptation. (5) Internet information protection is the strongest predictor of cross-cluster outgrowth of internet addiction networks, acting as a “guardian” of adolescents’ internet adaptation.

Key words: Adolescents, Internet Adaptation, Internet Addiction, Topological Structure, Network Analysis

DONG Wanghao, ZHANG Jie, MENG Sujie, JIA Min, WANG Weijun. (2025). The topological structure of adolescents’ internet adaptation: A longitudinal tracking study. Acta Psychologica Sinica, 57(3), 415-427.

Figures/Tables 17

Figure 1. Schematic diagram of the dynamic process of internet adaptation.

Table 1 Descriptive Statistics of Internet Adaptation and Internet Addiction Dimensions

Dimension	N = 5783			N = 1235 (T1)			N = 1235 (T2)
Dimension	Mean (SD)	Skewness	Kurtosis	Mean (SD)	Skewness	Kurtosis	Mean (SD)	Skewness	Kurtosis
Self-Control	3.95(1.07)	0.11	?0.15	4.13(1.06)	0.03	?0.43	4.49(0.93)	?0.25	?0.28
Interpersonal Interaction	3.87 (1.06)	?0.04	?0.08	3.93(1.07)	?0.17	?0.13	4.44(0.92)	?0.19	?0.27
Information search	4.07(1.15)	?0.21	?0.13	4.07(1.17)	?0.31	?0.11	4.36(1.06)	?0.41	0.00
Information Protection	4.39(1.09)	?0.26	?0.44	4.50(1.06)	?0.38	?0.42	4.53(0.96)	?0.40	?0.10
Proactive Coping	4.54(1.05)	?0.32	?0.31	4.72(0.99)	?0.38	?0.52	4.65(0.93)	?0.43	0.25
Learning Ability	4.06(1.04)	?0.06	0.13	4.05(1.06)	?0.13	0.06	4.49(0.95)	?0.29	?0.04
Self-Efficacy	3.97(1.11)	0.01	?0.10	3.94(1.15)	?0.08	?0.20	4.48(0.98)	?0.31	?0.08
Curiosity	4.11(1.02)	?0.08	0.12	4.13(1.03)	?0.18	0.10	4.42(0.93)	?0.18	?0.11
Craving	-	-	-	2.41(1.00)	0.41	?0.11	2.72(1.08)	0.14	?0.49
Tolerance	-	-	-	2.28(1.04)	0.44	?0.38	2.57(1.13)	0.20	?0.66
Behavioral Dyscontrol	-	-	-	2.24(1.07)	0.52	?0.36	2.48(1.15)	0.32	?0.66
Withdrawal	-	-	-	2.00(1.08)	0.84	?0.08	2.25(1.13)	0.53	?0.55
Time Extension	-	-	-	2.57(1.16)	0.31	?0.70	2.66(1.13)	0.16	?0.71
Negative Consequences	-	-	-	1.62(0.93)	1.36	1.03	2.01(1.12)	0.71	?0.57
Deception	-	-	-	1.94(1.06)	0.91	0.11	2.22(1.15)	0.48	?0.73
Escape	-	-	-	2.24(1.18)	0.62	?0.45	2.52(1.28)	0.30	?0.94

Figure 2. Regularized partial correlation network of internal dimensions of internet adaptation.
Note. The thicker the edge, the stronger the conditional correlation between nodes. Blue edges indicate positive relationships, and red edges indicate negative relationships. For a color version, please refer to the electronic version.

Figure 3. Node strength and expected impact of internet adaptation.

Figure 4. Cross-sectional network at T1 and T2.
Note. For visual comparison, the network is presented in a circular layout.

Figure 5. Cross-lagged network of adolescent internet adaptation (Autoregressive paths omitted).

Figure 6. Outgoing and incoming predicted effects in the cross- lagged network of internet adaptation.

Figure 7. Cross-lagged network model of T1 internet adaptation and T2 internet addiction.
Note. The model represents the temporal interactions between internet adaptation and internet addiction over two time points.

Figure 8. Cross-cluster outgoing predictability (left) and incoming predictability (right) in the cross-lagged network of internet adaptation and internet addiction.

Figure S1. Nonparametric bootstrap method for testing network adaptation to the strength difference of individual nodes of the transect network.
Note. Black squares in the figure indicate that the 95% confidence interval corresponding to the difference in the strengths of the two nodes derived from the bootstrap method does not contain zero, while gray squares indicate that the 95% confidence interval corresponding to the difference in the strengths of the two nodes contains zero. The diagonal lines indicate the magnitude of the individual node intensities (below).

Figure S2. Nonparametric bootstrap method for testing network adaptation to transect differences in the expected impact of individual nodes of the network.

Figure S3. Stability test results of the Case-droping method for testing network adaptation to intensity and expected impacts in the transect network
Note: The X-axis indicates the percentage of remaining after randomly dropping samples on the original sample; the Y-axis indicates the correlation of the centrality metrics between the remaining samples and the original sample. Points on the lines in the figure indicate the average correlation (from 90% to 10%) between the intensity estimated in the full sample and the intensity estimated on a random subsample that retains only a percentage of cases. Shaded areas indicate 95% bootstrap confidence intervals for the correlation estimates. Higher values indicate better stability of the centrality estimates. (below)

Figure S4. Nonparametric bootstrap method for testing network adaptation to edge differences in transect networks.

Figure S5. Nonparametric bootstrap method to test the accuracy of marginals estimation in network-adapted transect networks.
Note. The red and black dots on each horizontal line indicate the sample values and the estimated bootstrap means of the marginals, respectively. The shaded portion of each horizontal line is the 95% confidence interval of the bootstrap estimate of the marginals, and the narrower gray shading means the higher accuracy of the marginals. (Same below)

Figure S6. Case-droping method to test the stability of network adaptation to in-expected and out-expected impacts in cross-lagged networks.

Figure S7. Nonparametric bootstrap method to test the accuracy of margin estimation in network-adapted cross-lag networks.

Figure S8. Nonparametric bootstrap method for testing network adaptation to borderline differences in cross-lagged networks.