2.1. Concept definition

2.2. Literature review

3.1. Sampling and data collection

This study uses data from two waves of a longitudinal survey of third- and fourth-grade students in rural primary schools in western China. The data were collected by the Center for Experimental Economics in Education at Shaanxi Normal University. This study was conducted in accordance with the principles of the Declaration of Helsinki and was approved by the Institutional Review Board of the School of Economics and Management at Tsinghua University (protocol code 2019-TSEM-002, approval date: 15 November 2019). Data collection was carried out in December 2019 using written questionnaires, with approval obtained from the University. Written informed consent for participation and data use, including consent for publication of anonymized results, was obtained from schools, parents or guardians of participating students. All participants were fully informed that their anonymity would be preserved, the purpose of the research, the intended use of the data, and any potential risks associated with participation. No vulnerable individuals (e.g., patients, refugees) were included in the study. At the beginning of 2019, the research team conducted a preliminary survey in three cities in western China. In each city, seven counties were identified as research sites. Finally, 59 town-central primary schools across 21 counties were randomly selected as sample schools. In each selected school, one third-grade class and one fourth-grade class were randomly selected as the sample classes. The baseline survey was conducted in December 2019, and the follow-up survey was conducted in December 2020. Each survey included a student questionnaire, a standardized English test, and non-cognitive ability assessments. After excluding students who were lost to follow-up, the final sample consists of 3,235 students from 59 primary schools across 21 counties in western China, including only those students observed in both the baseline and follow-up surveys. Attrition between the two waves was minimal, suggesting that panel loss is unlikely to be a major source of bias in this study. Because this paper is based on a quasi-experimental design rather than randomized assignment, our main concern is the non-random selection into IT adoption. For this reason, the empirical analysis emphasizes balance between treatment and control groups after propensity score matching, rather than attrition-balance diagnostics commonly used in randomized trials.

In this study, the outcome variables include students’ cognitive and non-cognitive abilities. Cognitive ability is measured using their standardized English test scores, which reflect students’ academic development within the school setting. To evaluate non-cognitive ability, we apply a range of widely recognized psychological assessment tools. These include the Big Five Personality Test [58], the Grit Scale [59], the Internal-External Control Scale [60], the Mental Health Test (MHT), the Uncertainty Orientation Scale [61], and the Academic and Social Resources Self-Efficacy Scale [62].

Mental health is measured using the Mental Health Test (MHT), a widely used instrument for assessing children’s emotional and psychological difficulties in school settings. In our study, the MHT is treated as one dimension of students’ non-cognitive development because it captures emotional adjustment and self-regulation, which are closely related to children’s social-emotional functioning. The scale includes eight subdimensions: academic anxiety, people anxiety, loneliness tendency, self-blame tendency, hyper-sensitivity, health symptoms, terror tendency, and impulse tendency. The total MHT score ranges from 0 to 90, with higher scores indicating more severe psychological distress and poorer mental health. Accordingly, a positive coefficient on the MHT outcome should be interpreted as a deterioration in mental health rather than an improvement.

The key independent variable is students’ use of information technology, categorized into three domains: learning, entertainment, and socializing. This variable is defined by three survey questions:

1. (1). Have you used a learning machine, a computer, a mobile phone, or the Internet to study English this semester?
2. (2). Do you use a computer or mobile phone to access the Internet for entertainment (e.g., playing games, listening to music, watching videos)?
3. (3). Do you use a computer or mobile phone for social networking or chatting with friends?

Each question offers three possible responses: “often,” “sometimes,” or “never.” These response categories capture the frequency with which students participated in each type of IT activity, but they do not provide exact measures of duration (e.g., hours per day or week) or the amount of effort invested in the activity. Given the young age of the respondents and the need to maintain a simple and comparable questionnaire across rural schools, the survey was designed to measure domain-specific participation rather than precise usage intensity. In the main analysis, we therefore code “often” and “sometimes” as 1 (indicating participation in that domain of IT use) and “never” as 0 (indicating no participation). Accordingly, our treatment variables should be interpreted as indicators of whether students engaged in learning-, entertainment-, or social-oriented IT use, rather than as exact time-use or effort measures.

We control for a range of student characteristics at the individual, family, and school levels. At the individual level, these controls include gender (male = 1, female = 0), age, ethnicity (Han = 1, non-Han = 0), boarding situation (boarding = 1, not boarding = 0), and health situation (healthy = 1, unhealthy = 0). At the family level, these controls include whether the student has siblings (one or more siblings = 1; none = 0) and the parents’ education levels (mother’s and father’s education each coded as above junior high school = 1; junior high or below=0). Additionally, we include whether the mother works outside (yes = 1, no = 0), whether the father works outside (yes = 1, no = 0), and the family assets. At the school level, we include school fixed effects in the model to address potential endogeneity arising from differences across schools.

Table 1 provides a summary of the cognitive and non-cognitive development of all students in the sample, along with their individual and family background characteristics. On the whole, students show relatively low levels of cognitive development, with large differences between individuals. For non-cognitive traits, many students score well in areas like extraversion, agreeableness, conscientiousness, and openness. They also tend to show strong grit and are actively involved in school life. At the same time, some problems are noticeable, such as self-blame, hypersensitivity, and academic anxiety. The students in the sample vary widely in age. Around 40 percent are from ethnic minority backgrounds, and fewer than 10 percent are the only child in their families. Most of their parents have relatively lower levels of education. More than one-third of fathers work as migrant laborers, and the overall level of family assets is still low.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Sample descriptives.

https://doi.org/10.1371/journal.pone.0349438.t001

3.2. Treatment description

This study examines three domain-specific “treatments” that capture students’ adoption of information technology (IT) between the baseline survey (December 2019) and the follow-up (December 2020). At each wave, the student questionnaire asked whether the student (i) used a learning machine/computer/mobile phone/Internet for learning (e.g., online classes, homework, searching educational resources), (ii) used IT for entertainment (e.g., games, videos, music), and (iii) used IT for socializing (e.g., chatting/messaging, social networking). Response options were often/sometimes/never; we code often or sometimes as 1 (user) and never as 0 (non-user). Because our survey instrument identifies whether students adopted IT use in each domain but does not record exact time spent, the treatment captures domain-specific adoption between waves rather than the intensive margin of use.

For each domain, we construct a domain-specific panel indicator and restrict the sample to students who were non-users at baseline in that domain. For clarity, the treatment and control groups are defined separately for each IT-use domain. A student is classified as treated in a given domain if they were a non-user at baseline (0) and reported use at follow-up (1); students who remained non-users (0 → 0) serve as the control group for that domain. Because these domains reflect different purposes of IT, treatments are not mutually exclusive across domains (e.g., a student may adopt IT for learning but not for entertainment); therefore, we estimate three separate PSM-DID models (see 3.4), one per domain. Accordingly, the estimated effects should be interpreted as domain-specific adoption effects, rather than as mutually exclusive comparisons across all types of IT use. In other words, the coefficient for one domain captures the effect of adopting that specific type of IT use relative to remaining a non-user in that same domain, but does not imply that students in the treatment group do not also use IT for other purposes. Adoption is non-random; to proxy the selection process we estimate domain-specific propensity scores using rich baseline covariates: gender, age, ethnicity, boarding status, health status, only-child status/siblings, parental education, parental migration (mother/father works outside), family assets, baseline outcomes, and school fixed effects, and then match treated students to comparable controls prior to DID estimation.

3.3. Statistical models

3.3.1. Ordinary least squares model.

The use of information technology (regarded as intervention) is hypothesized to improve students’ cognitive and non-cognitive abilities. This hypothesis is tested using an Ordinary Least Squares (OLS) regression model as follows:

(1)

In equation (1), the dependent variable represents cognitive and non-cognitive abilities of student i of school j at time t. The key independent variable, , is a binary indicator denoting whether student i uses information technology for learning, entertainment, and socializing (1 if used, 0 if not used). The coefficient captures the impact of information technology use on students’ cognitive and non-cognitive abilities. The term encompasses individual, family, and school-level control variables, while denotes the residual error term accounting for unobserved factors.

3.3.2. PSM-DID model.

Students who did not use IT during baseline survey were divided into treatment and control groups. The treatment group received access to IT. During the follow-up survey, students in the treatment group had started using IT, while those in the control group still had no access.

However, the decision to use IT was not made randomly. It is likely that students’ personal traits and family background influenced whether they adopted IT or not. If we do not control for this selection bias, it becomes difficult to tell whether any observed differences in cognitive and non-cognitive outcomes are actually caused by IT use or by other unrelated factors. This limitation makes it hard to draw reliable conclusions about the effect of IT use on student performance.

To address this issue, this study adopts the Propensity Score Matching (PSM) method, drawing on the approaches proposed by Zheng, Ye and Chen [36,63,64]. First, we estimate each student’s likelihood of using IT through logistic regression, using a range of background variables. Because IT adoption was not random, students’ likelihood of using IT was influenced by baseline characteristics such as gender, age, ethnicity, health status, boarding status, parental education, parental migration status, and family assets. These covariates were included in the propensity score model to proxy the non-random selection mechanism. The regression model is specified as Eq. (2):

(2)

where is a dummy variable representing whether the respondent received IT in the follow-up survey (1 = received; 0 = not received). is a set of multidimensional indicators based on students’ individual and family levels. and represent periods before and after IT adoption.

Then, students from the treatment and control groups are matched based on their predicted probabilities. This matching process helps make the samples more similar, which greatly reduces the impact of selection bias.

Subsequently, a Difference-in-Differences (DID) approach is applied to compare changes in outcomes between matched treatment and control groups. The estimated impact is obtained by subtracting the differences observed in the control group from those observed in the treatment group. This method helps eliminate the influence of unobserved variables, thereby providing an accurate estimate of IT’s actual effect. The PSM-DID estimator is formalized as:

(3)

In equation (3), T and C represent treatment and control groups (D = 1 for treatment, D = 0 for control), respectively. and represent periods before and after IT adoption. Variables and indicate the cognitive and non-cognitive ability scores of students before and after the change in information technology use (namely treatment), while represents control variables. The results of the balance test between the treatment group and the control group before matching are presented in Supporting information (S1–S3 Table), and Supporting information (S4–S6 Table) display the differences between the two groups after matching. As can be observed, the groups exhibit no statistically significant differences after matching.

To address concerns regarding the plausibility of the parallel trends assumption with only two survey waves, we provide several supporting arguments. First, the balance tests (Supporting information, S1–S3 Table) indicate that, after propensity score matching, the treatment and control groups are statistically comparable in both baseline outcomes and observable characteristics. This suggests that the two groups started from a similar position before IT adoption. Second, the short time span between the baseline (2019) and follow-up (2020) surveys reduces the likelihood of unobserved shocks or major policy changes that could have systematically affected one group but not the other. Third, our model includes rich sets of individual, family, and school-level controls as well as school fixed effects, which further mitigate potential confounding. Finally, we conducted robustness checks using alternative matching strategies and outcome measures, and the results remain consistent. Taken together, these features enhance the credibility of the identifying assumption even though the formal pre-trend test cannot be implemented with only two periods.

3.3.3. Heterogeneity model.

The impact of information technology (IT) use may differ across student groups with distinct characteristics. To examine this heterogeneity, individual and family characteristics, such as gender, age, ethnicity, boarding situation, health situation, siblings, parental education level, parental work outside, and family assets, are considered in this study. Specifically, the analysis investigates whether IT use has varying effects depending on students’ gender, age, maternal education level, and family assets.

The heterogeneity model employed is defined as follows:

(4)

In equation (4), the interaction term represents the interaction between IT use () and specific student characteristics (). The coefficient captures the heterogeneous effects of IT use associated with these characteristics. The term represents the residual error capturing unobserved individual-level factors.

3.4. Research hypotheses

Based on the defined variables, we propose the following hypotheses, which will be tested through the subsequent model specification:

• Hypothesis 1: The use of IT in education can enhance students’ cognitive and non-cognitive abilities. Such use may broaden their horizons and stimulate their innovative thinking.
• Hypothesis 2: IT use in entertainment can hamper students’ cognition and non-cognition. IT use in entertainment distracts students to excessive visual/graphical content which impacts their developmental process.
• Hypothesis 3: Using IT for social interaction might also impair students’ cognitive and non-cognitive abilities. IT in socializing compromises students’ privacy and reduces opportunities for real-life interaction and cooperation.
• Hypothesis 4: The impact of IT use is more significant among Han nationality students. Due to differences in knowledge backgrounds and growth environments, Han students are better equipped to understand and utilize IT tools compared with non-Han students.
• Hypothesis 5: IT utilization will have a much bigger effect on older students, who are more capable of utilizing IT.
• Hypothesis 6: The effect of IT use may be greater for students whose mothers have lower levels of education. This is due to mothers with low education being less effective in affecting their children’s family education and, hence, making these students more vulnerable to the impact of IT use.
• Hypothesis 7: The impact of IT use is more significant for students from families with fewer assets. This is because limited family assets constrain educational opportunities, thereby amplifying the role of information technology in their learning.

4.1. Effects of information technology use on cognitive and non-cognitive abilities

S1–S3 Tables in supporting information present the covariate differences between treatment and control groups before matching, and S4–S6 Tables in supporting information show the corresponding results after matching. In all three IT-use domains, observable differences were effectively eliminated after matching. Figs 1–3 further illustrate that standardized mean differences were substantially reduced and remained well below the 10% threshold, indicating that covariate balance was achieved.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Standardized mean differences before and after matching for covariates: IT use in learning‌‌.

https://doi.org/10.1371/journal.pone.0349438.g001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Standardized mean differences before and after matching for covariates: IT use in entertainment.

https://doi.org/10.1371/journal.pone.0349438.g002

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Standardized mean differences before and after matching for covariates: IT use in socializing.

https://doi.org/10.1371/journal.pone.0349438.g003

To ensure comparability, we used a kernel density matching method based on propensity scores to match students in the treatment and control groups. After matching, the results from the logistic regression showed no significant differences between the two groups across all observed characteristics, which indicates that the groups were sufficiently balanced. We then used the Propensity Score Matching Difference-in-Differences (PSM-DID) approach to estimate Equation (3). Table 2 presents the effects of IT use in three different domains, learning, entertainment, and socializing on students’ cognitive and non-cognitive outcomes.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Effects of IT use on cognitive and non-cognitive abilities.

https://doi.org/10.1371/journal.pone.0349438.t002

The regression results show that using IT in any of the three domains does not lead to a clear improvement in students’ cognitive abilities. The estimated coefficients for cognitive outcomes are not statistically significant across learning (−0.043), entertainment (0.016), and socializing (−0.005). There may be two reasons for this. First, enhancing cognitive ability usually takes time and sustained effort. A one-year follow-up might not be long enough to observe meaningful changes. This is in line with Jaeggi et al. [65], who found similar results in a study conducted in the United States. Second, although IT can support learning, the potential benefits might be canceled out by its downsides. For example, frequent use of IT may reduce focus, lower study efficiency, and delay cognitive progress. This explanation is supported by previous studies, such as those by Cheema and Zhang and Falck et al. [66,67].

In contrast, IT use has more noticeable and varied effects on non-cognitive abilities. When used for learning, IT appears to improve students’ openness (0.079**), sense of control (0.377**), and liking for school (0.257**). However, using IT for entertainment shows negative effects. It lowers conscientiousness (−0.108***), reduces grit (−0.116***), and increases levels of anxiety (0.499**), impulsiveness (0.537***), and emotional sensitivity (0.477***). Similarly, IT use in social contexts lowers conscientiousness (−0.081***), raises anxiety (0.219*), and increases impulsivity (0.584***). These mixed outcomes suggest that IT may help develop open-mindedness and personal engagement, but at the same time, it can also bring about mental stress and behavioral problems. There are two likely explanations for these results. First, students at this age are still developing physically and mentally, and they may have difficulty understanding complex or harmful online content. This can negatively affect their personality development. Second, overuse of IT can reduce face-to-face communication with family and peers, which may lower their sense of responsibility and increase feelings of guilt. A plausible explanation is that learning-oriented IT differs from entertainment and social use in that it is embedded in goal-directed academic activities. Through access to new knowledge, interactive tasks, and feedback-rich learning environments, students may become more curious and more willing to engage with unfamiliar content, which is consistent with higher openness. At the same time, because such use is more closely connected to schoolwork and teacher-guided activities, it may reinforce students’ sense of connection to school and increase their positive feelings toward the learning environment.

These findings align with studies such as the randomized intervention by Banerjee et al. [50] in India and research by Ko et al. [57] in Taiwan, China. More broadly, this pattern is also consistent with recent research suggesting that the effects of children’s digital technology use are often more visible in engagement, well-being, and other non-cognitive dimensions than in short-term academic outcomes. Recent evidence further indicates that the consequences of digital use vary substantially by purpose and context, which may help explain why the estimated cognitive effects remain limited while the non-cognitive effects appear more mixed in our study. However, conflicting conclusions exist in the literature, including studies by Wittwer and Senkbeil [68] in Germany and Fiorini [55], which reported no significant relationship between IT use and non-cognitive abilities. Such inconsistencies likely result from differences in sample selection criteria, analytical methods, and definitions of IT usage. This study uniquely contributes by utilizing comprehensive survey data from rural western China, differentiating among various dimensions of IT use, and addressing specific aspects like duration and resource availability that have been overlooked in parts of the existing literature.

4.2. Robustness check

To ensure the accuracy and reliability of the main regression findings, this study conducts robustness tests by modifying the independent variable construction, altering the matching strategy used in propensity score matching, and employing alternative measures of cognitive abilities.

First, the robustness of findings concerning independent variables has been assessed by integrating IT usage across the three dimensions—learning, entertainment, and socializing—into a single composite measure using principal component analysis. Table 3 illustrates that the composite variable still yields an insignificant impact on cognitive ability, consistent with results in Table 2. Concerning non-cognitive outcomes, the composite measure shows significant effects, including decreases in conscientiousness (−0.037*) and increased neuroticism (0.043***), openness (0.041**), locus of control level (0.190**), mental health (1.479***), people anxiety (0.210**), hyper-sensitivity (0.264***), impulse tendency (0.312***), class like (−0.128*), and teacher like (0.182**). These results reinforce previous findings regarding the mixed non-cognitive effects of IT use.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Robustness check 1&2: principal component analysis of IT use dimensions and alternative matching strategies.

https://doi.org/10.1371/journal.pone.0349438.t003

Second, to check the robustness of the results, we replace the kernel density matching approach with the nearest neighbor matching approach (also in Table 3), which lead to the same result that do not show a strong improvement in any cognitive outcome in line with the previous conclusion. In terms of non-cognitive outcomes, the results show improvement in terms of openness (0.079**), locus of control level (0.375**), mental health (1.331*), hypersensitivity (0.458**), impulse tendency (0.372**), like for school (0.254**), and fear-related tendency (0.241*) retained strong relations, supporting the stability of the main results.

In the final check, shown in Table 4, we tested robustness by changing the way cognitive ability was measured. Instead of using standardized English test scores, we used students’ self-reported cognitive rankings in their class. Once again, IT use in learning (−0.031), entertainment (−0.052), and socializing (−0.021) does not show any significant effects on cognitive outcomes. This further supports the earlier conclusion that short-term IT use may not have a strong impact on cognitive development.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Robustness check 3: alternative outcome using students’ self-reported cognitive ability.

https://doi.org/10.1371/journal.pone.0349438.t004

Taken together, the robustness checks confirm the above conclusions. While IT use does not appear to improve cognitive skills in the short term, it continues to show mixed and complicated effects on several non-cognitive dimensions.

4.3. Heterogeneity analysis

To explore whether the effects of information technology use in learning, entertainment, and socializing vary across different student subgroups, we estimate Equations (3) and (4) by including interaction terms. Table 5 reports the results across four key categories: ethnicity, mother’s education level, student age, and family asset level.

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Heterogeneity in IT use for different types of students.

https://doi.org/10.1371/journal.pone.0349438.t005

4.4. Discussion

Our findings on the educational impact of IT use are broadly consistent with trends reported in international research. When it comes to cognitive outcomes, we do not find any significant improvement in students’ test scores related to IT usage. This result aligns with many previous studies in both developed and developing countries, which also report either weak or no effects of computer access on student achievement [6,67,69]. For example, large-scale programs that provided students with home computers or internet access have often failed to show clear short-term gains in academic performance [68,70]. In some cases, the introduction of digital tools has even produced negative academic results. One well-known example is a Romanian voucher study, where students who received home computers saw a drop in grades, possibly due to distractions, even though their computer skills improved [69]. At the same time, our finding of no significant effect does not contradict earlier evidence that well-designed educational technology programs can be effective. Structured computer-based learning interventions have shown clear improvements in math and language scores in various settings, including urban India and Latin America [50,71]. The lack of a strong effect in our case may be due to how IT is used in the study setting. General exposure to technology, especially when used for multiple purposes, may not directly improve test performance. Even when students have access to valuable information or educational software, the benefits may be offset by time spent on non-academic digital activities, as discussed in earlier work [55].

Compared with the limited effects on cognitive outcomes, the influence of IT use on non-cognitive outcomes in this study is more obvious and shows a mix of both positive and negative results. Students who used IT more often became more open to new experiences and expressed a stronger liking for school, which suggests higher levels of engagement and emotional connection to their learning environment. At the same time, these students also reported higher levels of anxiety, self-blame, and other negative emotional reactions than their peers. This combination of positive and negative outcomes is similar to findings from international research on the relationship between technology use and students’ social and emotional development. Past studies have not reached a consistent conclusion about how IT affects non-cognitive abilities [66]. While some work has shown that using computers can help students build self-esteem and develop a stronger sense of control over their learning [72], other studies have raised concerns. For example, when students spend a lot of time on screens without clear guidance, they may interact less with others in real life, which can increase the risk of psychological problems [21,73]. The higher levels of anxiety observed in our study may reflect these risks and contribute to the wider concern about how unstructured technology use might affect children’s mental health. This pattern is also consistent with recent findings. Emerging evidence suggests that the developmental consequences of children’s digital technology use are highly heterogeneous across purpose and context, and that non-cognitive domains such as engagement, well-being, and self-regulation may respond more quickly than short-term academic performance. In this sense, the limited cognitive effects and the more mixed non-cognitive effects found in our study are not unusual, but instead reflect the broader complexity of children’s digital experiences.

Although this study focuses mainly on the estimated effects of IT use, it is also useful to consider the possible mechanisms behind these results. One reason why we do not find a significant effect on cognitive ability could be the “substitution effect”. In rural settings, where access to structured digital platforms is limited and parental supervision is often lacking, students may use IT more for entertainment than for learning. As a result, the time they spend on devices might take the place of activities like reading or doing homework [55].

Another explanation for the lack of clear cognitive benefits is the “masking effect”. This occurs when positive and negative impacts appear at the same time and cancel each other out. On the one hand, using IT for learning can provide useful content, interactive formats, and personalized learning paths supported by digital tools and data-based systems. These features may help with cognitive growth. On the other hand, without proper guidance, the advantages of IT can be outweighed by problems linked to entertainment or social media use. These include distractions and difficulty focusing. For younger students who are still building their basic thinking and emotional skills, the attraction of digital content may shift their attention away from schoolwork, leading to slower or more hidden negative effects over time.

Third, it is also possible that the short time frame of the study contributes to the results. Cognitive development usually happens gradually and requires long-term effort. Even though IT has the potential to support learning by giving better access to information or by encouraging critical thinking, such benefits might take time to show. The lack of change over one year may reflect a timing issue rather than an actual lack of impact. Studies that follow students for a longer period could offer better insight into these long-term effects.

In contrast, the mixed effects we see on non-cognitive outcomes may be linked to psychological or behavioral factors. A more specific explanation is needed for why learning-oriented IT, rather than entertainment- or social-oriented IT, is associated with improvements in some non-cognitive dimensions. One likely mechanism is that learning-oriented IT exposes students to a broader range of knowledge, problem-solving tasks, and unfamiliar ideas in a structured way. Such exposure may stimulate curiosity and exploratory thinking, which is closely related to openness to experience. A second mechanism is that learning-oriented IT is typically embedded in school-related activities, such as completing assignments, accessing educational materials, or interacting with teacher-directed content. Because of this close connection with formal learning, students may feel more engaged in classroom activities and more emotionally connected to school, which helps explain the positive effects on school liking and teacher liking. At the same time, the freedom that comes with using IT might boost students’ confidence, especially among those who feel comfortable with digital technology [74]. A third mechanism is that digital learning tools often provide immediate feedback, opportunities for self-paced learning, and a greater sense of autonomy in task completion. These features may strengthen students’ perceived competence and sense of control, which are important components of non-cognitive development. This interpretation is broadly consistent with the literature emphasizing the role of school belonging, self-efficacy, and supportive learning environments in shaping children’s social-emotional outcomes. However, when IT use becomes excessive or lacks structure, especially for entertainment, it can also create emotional pressure. Prior research shows that too much exposure to the internet is linked with higher anxiety levels and weaker emotional control [75].

Variability between subgroup effects in this scenario might capture differences in the degree of parental engagement, and thus digital literacy. To succeed in online learning environments with autonomy, students from a lower socioeconomic status face more difficulties by themselves, and consequently, the productive IT use on their behalf can be also less significant. These can thus tell us more about the role of an enabling context in moderating the influence of IT in fostering students’ learning.