🧬 biology

A general factor underlies the mental-health risk of seven digital technologies

A study of over 10,000 adults reveals that the mental health risks associated with seven different digital technologies are driven by a single general factor of problematic use rather than platform-specific mechanisms, suggesting that apparent differences in harm reflect a shared user vulnerability.

Original authors: Zhikai Yu, Jing Li

Published 2026-09-09

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Zhikai Yu, Jing Li

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Public debate often treats digital life as a collection of distinct dangers, where one app might be a poison and another merely a nuisance. We hear that social media fuels anxiety while video calls simply tire us out, leading to a landscape of regulations that target specific platforms as if they were unique villains. This view assumes that the harm comes from the technology itself, shaped by its specific features like endless scrolling or video feeds. However, a different possibility exists: that the risk might not belong to the screen at all, but to the person holding it. In the study of mental health, researchers have long observed that different problems often cluster together, suggesting a shared underlying vulnerability rather than isolated causes. If this pattern holds true for digital life, then the intense focus on banning or restricting individual apps might be missing the real source of distress.

A team of researchers set out to test this idea by looking at seven different types of digital technology used by adults: social media, AI companions, short-form videos, sleep trackers, smartphones, dating apps, and videoconferencing. They recruited more than ten thousand adults and asked them detailed questions about how they used each of these tools and how those experiences affected their mental well-being. The goal was to see if the problems people reported were unique to each technology or if they all pointed to a single, common thread. The researchers built a unified measure of distress to compare the technologies side by side, treating them all with the same rigorous standards to ensure a fair comparison.

The results challenged the common assumption that some platforms are inherently more dangerous than others. When the researchers looked at the raw numbers, social media appeared to have the strongest link to mental health struggles, followed closely by AI companions and dating apps, with sleep trackers showing the weakest link. The differences were small but statistically detectable. However, when the researchers applied a more sophisticated statistical test to see if these differences meant anything real, the picture changed completely. They found that once they accounted for a general tendency to use digital tools in a problematic way, the specific details of which app a person used added no extra explanatory power. In other words, knowing that someone struggled with social media did not tell you anything more about their mental health than knowing they struggled with any other digital tool. The apparent ranking of danger vanished when the shared underlying factor was considered.

This general factor dominated the findings across all seven technologies. The researchers discovered that the scores for all fourteen different measures they used—covering both how people used the tools and the outcomes they experienced—were almost entirely explained by this single, shared vulnerability. While the individual questions within the surveys showed only a modest amount of shared information, the overall scores for each technology were so tightly linked that they behaved as if they were measuring the same thing. The researchers also looked at the data using network analysis and by grouping people into profiles based on their usage patterns. Both methods confirmed the same structure: people fell into groups of low, moderate, or high engagement, but these groups did not differ in which technologies they used. Instead, they differed only in the overall intensity of their engagement across the board.

The study suggests that the distress people feel is not driven by the specific mechanics of a platform, such as an algorithm or a video feed, but by a broader susceptibility to digital engagement. This does not mean that all technologies are identical, but rather that the risk of mental health problems arises from a shared human vulnerability that cuts across different types of digital use. The researchers noted that their findings do not prove exactly what this vulnerability is; it could be a general tendency toward addiction, a negative outlook on life, or simply the way the questions were worded. However, the data strongly indicates that the differences between platforms are far less significant than public debate often assumes.

These findings have important implications for how society approaches digital regulation. Policies that single out specific apps, such as social media, as the primary cause of a mental health crisis may be based on a misunderstanding of the problem. If the risk comes from a shared vulnerability rather than a specific platform, then restricting access to one app might not solve the underlying issue. The study suggests that a more effective approach might be to identify and support individuals who show signs of this broad-spectrum vulnerability, rather than focusing regulatory energy on the tools themselves. The researchers also observed that people who had a history of mental health diagnoses were sometimes less likely to report them, while those who chose not to disclose their history actually scored higher on distress measures, hinting that stigma might be hiding the true extent of the problem.

Ultimately, this large-scale study offers a quieter, more unified view of digital life. It suggests that the anxiety and exhaustion people feel are not unique symptoms of a specific app, but rather signs of a deeper, shared struggle with digital engagement. While the study cannot yet pinpoint the exact nature of this vulnerability, it provides strong evidence that the differences between technologies are small and that the real challenge lies in understanding the human factor that connects them all.

Technical Summary: A General Factor Underlies the Mental-Health Risk of Seven Digital Technologies

Problem Statement
Public discourse and emerging legislation (e.g., Australia's Online Safety Amendment Act) often operate on the assumption that specific digital technologies, particularly social media and short-form video, pose significantly greater mental health risks than others, such as videoconferencing or sleep trackers. However, existing literature consists of seven largely separate domains, each utilizing distinct instruments, samples, and theoretical frameworks. Consequently, no prior study has directly compared the relative mental health risks of different technologies within the same individuals. This gap makes it impossible to determine whether observed differences in risk are due to platform-specific mechanisms or a shared underlying vulnerability.

Methodology
The study recruited a large sample of 10,683 adults via WeChat, resulting in an analytic sample of 10,454 complete cases after applying a conservative, adaptive multi-criterion response-time threshold (1.83 s/item) to filter careless responders.

Measurement: The researchers developed and administered a battery of 14 custom scales covering seven digital technologies: social media, AI companions, short-form video, sleep trackers, smartphones, dating apps, and videoconferencing. For each technology, they measured both "Usage" (problematic use) and "Outcome" (psychological consequences).
Data Quality & Reliability: Cronbach's $\alpha$ for the 14 scales ranged from 0.725 to 0.832. However, exploratory factor analysis on the 25-item social media scale failed to recover the intended five-dimensional structure, suggesting high inter-correlation among items.
Analytic Strategy: The authors employed a deliberately conservative strategy to adjudicate between "platform-specific harm" and a "shared latent factor":
1. Unified Distress Index: Seven outcome scales were standardized and averaged into a single index.
2. Nested Model Comparison: A formal statistical test compared Model A (predicting distress using a single general usage factor extracted from the seven technologies) against Model B (predicting distress using all seven raw usage scores).
3. Psychometric Modeling: Hierarchical reliability ( $\omega_H$ ), item-level Harman tests, and eigenvalue decomposition of residual matrices were used to assess factor structure.
4. Network & Latent Profile Analysis: Partial-correlation networks (graphical lasso) and Gaussian mixture models were used to check for differentiated structures or distinct user profiles.
5. Validity Checks: Known-groups validity was assessed against self-rated mental health status and diagnosis history.

Key Results

Raw Rankings vs. Explanatory Power: Raw correlations between individual technology usage and the unified distress index showed a slight ranking, with social media highest ( $r = 0.916$ ) and sleep trackers lowest ( $r = 0.883$ ). However, the nested model comparison revealed that this ranking was statistically insignificant in terms of explanatory power. Adding specific technology details to a model containing only a general usage factor resulted in a negligible increase in variance explained ( $\Delta R^2 = 0.0001$ , $F(6, 10448) = 1.19$ , $P = 0.309$ ).
Dominance of a General Factor: At the scale-mean level, a single general factor explained nearly all variance in the 14 scales ( $\omega_H = 0.984$ ). While item-level sharing was more modest (36.5% variance explained by the first factor), this is consistent with aggregation effects of ~20 intercorrelated items per scale.
Network and Profile Convergence: Psychometric network analysis showed no identifiable bridge nodes or differentiated structure; all edge weights were confined to a narrow range (0.05–0.09). Latent profile analysis identified three groups (low, moderate, high engagement) that differed only in overall severity, not in which specific technologies were emphasized.
Validity: The unified distress index demonstrated a clean monotonic gradient across self-rated mental health status, supporting its substantive validity. Notably, respondents who declined to disclose diagnosis history scored higher than those who reported a prior diagnosis, suggesting potential underreporting due to stigma.

Key Contributions

Empirical Benchmarking: This is the first study to simultaneously measure problematic use and outcomes across seven distinct digital technologies within a single sample, directly testing the assumption of platform-specific risk hierarchies.
Methodological Framework: The study introduces an adaptive, multi-criterion response-time threshold for large-scale online surveys and a robust nested model comparison strategy to test whether highly correlated predictors offer independent explanatory power.
Measurement Battery: The authors provide a publicly available, diagnostically transparent battery of 14 scales. While the authors note limitations in the scales' internal dimensional structure, the transparency regarding these shortcomings serves as a starting point for future scale development.

Significance and Claims
The paper claims that apparent differences in mental health risk among digital technologies likely reflect a shared, cross-platform vulnerability rather than distinct, platform-specific harm mechanisms. The findings suggest that the "p factor" logic from psychopathology may apply to digital engagement, where a general disposition (e.g., addiction proneness or negative affectivity) drives distress across all technologies.

The authors argue that if replicated, these findings challenge the empirical footing of policies that single out specific platforms (like social media) for regulation while ignoring others. They posit that regulatory attention might be better directed toward identifying and supporting individuals with a "broad-spectrum digital engagement vulnerability" rather than restricting access to specific applications. The study concludes that the observed risk is not a methodological artifact but a substantively meaningful general factor, though the precise psychological nature of this factor (e.g., whether it is a true transdiagnostic disposition or an artifact of item wording) requires further investigation.

Technical Summary: A General Factor Underlies the Mental-Health Risk of Seven Digital Technologies

More like this