With the rapid advancement of artificial intelligence (AI) technology, an increasing number of university students are integrating AI into their academic studies and daily life. Based on literature review, the present empirical research examining the heterogeneity of motivations for AI use and its causal relationship with mental health is relatively limited, resulting in an inadequate understanding of the “empowering” potentials and “risk” concerns associated with AI use among university students. Moreover, existing research on the relationship between AI use motivations and psychological health is scarce, predominantly cross-sectional, and largely focused on negative indicators, restricting a comprehensive understanding of their mutual relationship. According to the dual-factor model of mental health, it is crucial to consider both positive and negative psychological health indicators when evaluating the association of AI use motivations and mental health, providing a more comprehensive and objective reflection of individual psychological well-being. Therefore, this study examines the relationship between four types of AI use motivations and mental health among university students by selecting flourishing and meaning in life as positive psychological indicators, and anxiety and depression as negative psychological indicators, to better understand the “empowering” and “risk” roles of AI use among university students.
This study used both cross-sectional and longitudinal designs to investigate the heterogeneity and influencing factors of university students’ motivations for AI use and their relationships with mental health. Study 1 was a cross-sectional survey of 5,760 university students from cities including Beijing, Shanghai, Zhengzhou, and other cities in China. All participants completed a questionnaire regarding university students’ AI use and mental health. Latent profile analysis identified four subtypes of AI use motivations: the tool-oriented type (28.09%), the high tool-entertainment motivation type (30.57%), the moderate multi-motivation type (36.02%), and the globally high motivation type (5.31%). Gender, grades, place of birth, and major effectively predicted these subtypes of AI use motivations among university students. Moreover, significant differences in mental health levels were observed among the identified latent classes of AI use motivations, with the efficiency high tool-entertainment motivation group scoring the highest, followed sequentially by the tool-oriented, moderate multi-motivation, and globally high motivation groups, which scored the lowest.
In Study 2, 221 university students from a university in Henan Province were selected to participate in a six-month longitudinal survey exploring the bidirectional relationship between AI use motivations and mental health. Cross-lagged analysis indicated that anxiety significantly predicted escapism motivation and social motivation, depression significantly predicted escapism motivation, and a higher sense of meaning significantly predicted instrumental motivation, while AI use motivation did not significantly predict mental health indicators.
The findings reveal the heterogeneity of university students’ motivations for using AI and their relationship with mental health. These results provide empirical evidence for personalized interventions and psychological health promotion strategies. However, there are some limitations need further consideration. First, the longitudinal survey included only two waves of data collection over a six-month interval, with a relatively small and homogeneous sample. Future research could increase the number of survey waves (e.g., three or more), expand the sample size, and enhance sample heterogeneity to more comprehensively and accurately reveal the long-term relationship patterns between AI use motivations and mental health. Second, the survey in this study was conducted from 2024 to 2025, at which time the overall intensity of university students’ motivations for AI usage was not yet prominent. With the rapid popularization and deeper integration of generative AI technologies among university students, the intensity and structure of motivations for AI use within this group may change. Therefore, future research should continuously observe and dynamically track the evolving trends of university students’ AI use motivations, capturing their actual conditions and patterns of change more accurately in the current context.

Cognitive computational modeling quantifies human mental processes using mathematical frameworks, thereby translating cognitive theories into testable hypotheses. Modern cognitive modeling involves four interconnected stages: defining models by formalizing symbolic theories into generative computational frameworks, collecting data through hypothesis-driven experiments, inferring parameters to quantify cognitive processes, and evaluating or comparing models. Parameter inference, a critical step that facilitates the integration of models and data, has traditionally relied on maximum likelihood estimation (MLE) and Bayesian methods like Markov Chain Monte Carlo (MCMC). These approaches depend on explicit likelihood functions, which become computationally intractable for complex models—such as those with nonlinear parameters (e.g., learning dynamics) or hierarchical/multimodal data structures.
To address these challenges, simulation-based inference (SBI) emerged, leveraging parameter-data mappings via simulations to bypass likelihood calculations. Early SBI methods, however, faced computational redundancy and scalability limitations. Recent advances in neural simulation-based inference (NSBI), or neural amortized inference (NAI), harness neural networks to pretrain parameter-data relationships, enabling rapid posterior estimation.
Despite its advantages, NSBI remains underutilized in psychology due to technical complexity. This work focuses on neural posterior estimation, one of three NSBI approaches alongside neural likelihood estimation and neural model comparison. Neural posterior estimation operates in two phases: training and inference. During the training phase, parameters are sampled from prior distributions, and synthetic data are generated using the model; a neural network is then trained to approximate the true posterior from these training pairs. In the inference stage, real data are input to the trained network to generate parameter samples. The BayesFlow framework enhances neural posterior estimation by integrating normalizing flows—flexible density estimators—and summary statistic networks, enabling variable-length data handling and unsupervised posterior approximation. Its GPU-accelerated implementation further boosts efficiency.
Neural posterior estimation has expanded the scope of evidence accumulation models (EAMs), one of the most widely used framework in cognitive modeling. First, it enables large-scale behavioral analyses, as demonstrated by von Krause et al. (2022), who applied neural posterior estimation to drift-diffusion models (DDMs) for 1.2 million implicit association test participants. By modeling condition-dependent drift rates and decision thresholds, they revealed age-related nonlinear cognitive speed changes, peaking at age 30 and declining post-60. Neural posterior estimation completed inference in 24 hours versus MCMC’s 50+ hours for a small subset, demonstrating its scalability.
Second, neural posterior estimation supports dynamic decision-making frameworks, exemplified by Schumacher et al. (2023), who combined high-level dynamics with low-level mechanisms using recurrent neural networks (RNNs). Their simultaneous estimation of hierarchical parameters achieved over 0.9 recovery correlations and superior predictive accuracy compared to static models.
Finally, neural posterior estimation facilitates neurocognitive integration, as shown by Ghaderi-Kangavari et al. (2023), who linked single-trial EEG components (e.g., CPP slope) to behavior via shared latent variables like drift rate. This approach circumvented intractable likelihoods and revealed associations between CPP slope and non-decision time.
NSBI enhances cognitive modeling by enabling efficient analysis of complex, high-dimensional datasets. Its key limitations include model validity risks (biased estimates from incorrect generative assumptions), overfitting concerns (overconfident posteriors on novel data), and upfront training costs for amortized methods. Future work should refine validity checks—such as detecting model misspecification—and develop hybrid inference techniques. NSBI’s potential extends to computational psychiatry and educational psychology, promising deeper insights into cognition across domains. By addressing complexity barriers, NSBI could democratize advanced modeling for interdisciplinary research, advancing our understanding of human cognition through scalable, data-driven frameworks.