Artificial intelligence and machine learning in assessing and promoting health and well‐being: Integrating human insight with computational intelligence

The generation of Big Data (characterized by 4V: volume, variety, velocity, and veracity) and the advanced analytics techniques including artificial intelligence (AI) and machine learning (ML) are profoundly shaping health behavior research by enabling new forms of assessment, prediction, and intervention that extend beyond the capabilities of traditional psychological methods (Tariq et al., 2021). Historically, applied psychology has relied heavily on self-report instruments, theory-driven constructs, and linear analytic approaches to understand health behaviors and well-being. While these approaches have yielded foundational insights, they face inherent challenges in modeling nonlinear relationships, capturing dynamic trajectories, and scaling interventions across populations (Box-Steffensmeier et al., 2022). AI/ML approaches can combine data to evaluate the complex interplay between individual, contextual, and structural factors influencing people's behaviors and health outcomes (Li et al., 2022). In addition, the AI/ML approaches may reveal events that are latent or transient in traditional analysis of datasets by identifying unexpected associations through analyses of diverse data (Jing et al., 2024; Piwek et al., 2016; Ruths Topol, 2019). The goal of this special issue of Applied Psychology: Health and Well-Being is to bring together a set of the latest articles on health behavior and health assessment and intervention using AI/ML approaches showcasing multidisciplinary collaborations across epidemiology, bioinformatics, mathematical modeling, statistics, social and behavioral sciences, and bioethics. The 16 contributions (2 literature reviews, 14 empirical studies) in this special issue have exemplified how multidisciplinary teams are using AI/ML to advance understanding of health-related behaviors and psychological well-being. The studies span populations from children to adults, from typically developing individuals to those with special educational needs and apply a range of computational techniques, from large language models to interpretable machine learning and ensemble algorithms. Collectively, these contributions signal an emerging paradigm: one where human-centered AI/ML not only predicts outcomes but also enhances interpretive understanding of psychological processes that underlie health and flourishing. Across the contributions, three complementary roles of AI/ML emerge clearly. First, AI/ML functions as advanced analytic tools that enhance the assessment and prediction of health behaviors and well-being. Second, AI systems increasingly operate as active agents of behavior change, interacting directly with individuals through chatbots, robotic companions, and generative language models. Third, AI/ML brings methodology innovations and ethical challenges in health behavior research. A substantial subset of papers in this special issue demonstrates how ML can improve the prediction of health-related outcomes by uncovering complex, nonlinear patterns that are difficult to detect using conventional statistical techniques. For example, Lilleholt et al. (2024) applied lasso regression and longitudinal panel models to identify key predictors of adherence to health-protective behaviors during the COVID-19 pandemic, revealing that empathy toward vulnerable others and perceived moral costs of nonadherence were among the most robust predictors across time. Similarly, Chen et al. (2024) employed random forests and mixture modeling to predict distinct life satisfaction trajectories among older adults, highlighting the central role of emotional experiences such as happiness and loneliness. Importantly, several contributions emphasize population heterogeneity and personalized pathways to well-being, a key advantage of ML approaches. Lin et al. (2025) applied six ML algorithms to examine how physical literacy predicts adolescents' flourishing across physical, mental, and academic outcomes. They identified “self-control” and “self-expression” as key contributors, emphasizing that motivation and interpersonal competence jointly underpin youth well-being. These findings underscore how ML can support more tailored interventions by identifying subpopulations with distinct risk and resilience profiles. Similarly, based on two-wave data from military personnel, Chen et al. (2025) examined the predictive power of different dimensions of empowerment (personal, interpersonal, and socio-political) on new recruits' health perception using five ML classifiers and further compared the model performance across subgroups. Meng et al. (2025) analyzed data from over 60,000 students across nine countries to identify social–emotional skills that best predict well-being and physical health. Optimism, energy, and stress resistance emerged as the strongest cross-cultural predictors, reaffirming resilience as a universal resource. Tan et al. (2025) turned to an often-overlooked population: children with special educational needs (SEN). Using longitudinal ML models (nonlinear classifiers and clustering techniques), they discovered heterogeneous pathways to subjective well-being, characterized by clusters emphasizing interpersonal, academic, or mixed predictors. By uncovering this heterogeneity, the study demonstrates how ML can reveal individualized trajectories of health and development, advancing the precision and inclusivity of well-being research. Their study showcases ML's ability to handle complex, high-dimensional, and multinational data—yielding insights that transcend cultural boundaries and inform globally relevant interventions. Beyond predictive accuracy, a defining theme of the special issue is the pursuit of psychological interpretability in AI-based assessment. Traditional psychometric tools often struggle to capture the fluidity and complexity of human well-being in real-world contexts. Han et al. (2024) addressed this challenge by developing a machine learning–based instrument for subjective well-being using Weibo social media data. Their model achieved strong criterion validity and reliability while maintaining interpretability through SHAP-based feature analysis, identifying linguistic indicators linked to cultural values, morality, and emotional tone. The study demonstrates how ML can reveal psychologically meaningful constructs that transcend self-report bias. Rather than treating ML models as opaque “black boxes,” several studies integrate algorithmic performance with theory-informed explanations Liu et al. (2025) introduced a psychologically interpretable AI framework (emoLDAnet) for screening loneliness, depression, and anxiety using multimodal multi-modal signals (facial expression and physiology), explicitly mapping model outputs onto established psychological constructs. The model integrates the OCC–PAD–LDA psychological transformation framework, translating emotional cues into meaningful psychological data. Complementing this approach, Wang et al. (2025) proposed ScaleLLM, a large language model–based framework that aligns psychological scale data with domain knowledge to enhance both predictive accuracy and interpretability in health assessments. Together, these studies demonstrate that interpretability is not a limitation of AI, but a design choice that can be addressed through thoughtful methodological integration. The special issue also highlights AI's expanding role as a tool for promoting health behaviors through direct interaction with individuals. Chan et al.'s (2024) systematic review and meta-analysis of chatbot-based vaccination interventions provides compelling evidence that AI-driven conversational agents can improve vaccination attitudes and, in some cases, uptake, particularly when interventions are tailored and interactive. At a more micro-interactional level, Hu et al. (2025) showed that AI-assisted venting can effectively reduce high-arousal negative emotions such as anger and frustration, demonstrating AI's potential to support emotional regulation processes traditionally studied in clinical and counseling psychology. Guo and Wan (2025) demonstrated that a robotic companion's food choices can influence individuals' own meal selections, particularly when the robot is perceived as highly anthropomorphic. This work extends classic theories of social modeling and norm adherence into human–robot contexts, illustrating how AI agents can become socially meaningful actors in health-related decision making. Vandelanotte et al. (2024) complemented this line of inquiry through focus groups exploring user perceptions of AI-driven digital assistants for physical activity. Participants valued personalization, adaptability, and feedback, but also emphasized privacy, trust, and autonomy. These perspectives highlight the psychological prerequisites for successful digital interventions: users must perceive AI not as surveillance but as support. Another prominent contribution of the special issue lies in its treatment of generative AI and natural language processing (NLP) as innovations to reshaping the research practice. Qiao et al. (2025) demonstrated how large language models can dramatically reduce the time required for qualitative thematic analysis while maintaining substantial agreement with human coding, highlighting the potential of generative AI to augment, not replace, qualitative psychological research. Wu and Qiu (2024) utilized NLP techniques and panel vector autoregression methodology to explore perceived attitudes of social media users towards the digital transformation of agriculture and assessed its impact on total agricultural output and agricultural science and technology inputs. At a broader level, Vanhée et al.'s (2025) bibliometric analysis of AI for cognitive behavioral therapy (AI4CBT) situates these empirical advances within a rapidly expanding interdisciplinary field, identifying key trends, methods, and challenges that will shape future research. Looking forward, several priorities emerge from this body of work. Future research should continue to integrate AI methods with established psychological theories, ensuring that predictive gains translate into conceptual understanding. Greater emphasis is needed on explainability and fairness, especially for applications involving vulnerable populations or clinical decision making. Finally, interdisciplinary collaboration among psychologists, data scientists, clinicians, and ethicists will be essential to ensure that AI-driven innovations advance health and well-being in ways that are scientifically sound and socially responsible. As we celebrate predictive breakthroughs, we must continue to ask what these models mean in human terms. The ultimate goal of AI in applied psychology is not automation but augmentation—enhancing empathy, insight, and decision-making through technology that learns from and with us. We envision a future where AI serves as both mirror and facilitator: reflecting the complexity of human behaviors while collaborating with individuals and communities toward healthier, more flourishing lives. None of the authors have a conflict of interest to disclose. Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.