Correction: Development and validation of the conversational AI dependence scale for Chinese college students

Yuanyuan Chen, Mengyun Wang, Shujuan Yuan, Shuicheng YanFrontiers in Psychology 2025 · DOI · OpenAlex

Introduction

As conversational AI tools like ChatGPT, Claude, and Ernie Bot become ubiquitous in academic and personal spheres, understanding their psychological impact is paramount. Researchers have identified a growing phenomenon termed "AI dependence"—a behavioral pattern where users rely on these tools to an extent that may lead to negative psychological or social outcomes. To quantify this, the Conversational AI Dependence Scale (CAIDS) was developed and validated for Chinese college students. This report highlights a critical correction to the original publication, specifically addressing a layout error in the factor analysis results that is vital for the scale's accurate application.

The Problem and the Correction

The original study established the CAIDS as a robust psychometric instrument. However, the initial publication contained a column-alignment error in its primary data table. The headers for the psychometric factors were misordered, potentially leading researchers to misattribute survey items to the wrong behavioral traits. This correction ensures that the four dimensions of AI dependence—negative impacts, mood modification, withdrawal symptoms, and uncontrollability—are correctly mapped to their respective data points.

Methodology and Validation

The CAIDS was validated using a sample of $N = 547$ Chinese college students. The researchers employed Exploratory Factor Analysis (EFA) to uncover the underlying structure of the scale. The analysis confirmed a four-factor model that effectively captures the nuances of AI dependence.

The "Communality" values in the corrected results range from 0.59 to 0.84. In psychometrics, these values indicate the proportion of each item's variance explained by the extracted factors; the observed range suggests a high-quality fit for the model.

Key Dimensions of AI Dependence

The CAIDS identifies four distinct dimensions that constitute AI dependence:

  1. Negative Impacts: Interference with daily responsibilities, academic performance, or social health.
  2. Mood Modification: The use of AI as a coping mechanism to alter emotional states, such as escaping boredom or anxiety.
  3. Withdrawal Symptoms: The experience of distress, irritability, or restlessness when access to conversational AI is restricted.
  4. Uncontrollability: An inability to regulate or cease AI usage despite a conscious desire to do so.

Corrected Factor Analysis Results

The corrected data reveals that the four-factor model explains 74.41% of the total variance, a remarkably high figure in social science research that underscores the scale's explanatory power. Furthermore, the contribution rates for each factor are balanced, ranging from 17.67% to 19.75%, indicating that each dimension is equally significant in defining the construct of AI dependence.

Table 1: Corrected factor analysis results for the Conversational AI Dependence Scale (CAIDS), showing factor loadings, eigenvalues, and contribution rates.

As shown in the table above, the factor loadings for the items are consistently strong (ranging from 0.59 to 0.82), well above the standard 0.50 threshold for significance. This confirms that the survey items are highly representative of their theoretical constructs.

Implications and Future Research

With the publication of this correction, the CAIDS stands as a validated and reliable tool for researchers. While currently focused on Chinese college students, the scale provides a foundation for broader cross-cultural studies. Future research may involve Confirmatory Factor Analysis (CFA) in diverse populations or longitudinal studies to track how AI dependence evolves alongside the rapid advancement of generative AI technologies.

Summary updated

Abstract

[This corrects the article DOI: 10.3389/fpsyg.2025.1621540.].
BibTeX
@inproceedings{chen2025correction,
  author = {Yuanyuan Chen and Mengyun Wang and Shujuan Yuan and Shuicheng Yan},
  title = {{Correction: Development and validation of the conversational AI dependence scale for Chinese college students}},
  booktitle = {Frontiers in Psychology},
  year = {2025},
  doi = {10.3389/fpsyg.2025.1700281},
}
Figure