Network-based artificial intelligence in mental healthcare: A systematic review of chatbots, artificial...

This systematic review examines how artificial intelligence (AI), including machine learning (ML) models and AI-powered chatbots, contributes to the diagnosis, treatment and ethical governance of mental healthcare. It explores how AI-driven systems form interconnected healthcare networks that enhance accessibility, personalization and resilience of mental health services, aligning with the United Nations Sustainable Development Goal 3: Good Health and Well-Being.

A comprehensive search across PubMed, IEEE Xplore and Google Scholar (2017-2024) was conducted using Boolean combinations of "AI," "machine learning," "chatbots" and "mental health." Screening followed Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) 2020 guidelines, yielding 37 high-quality studies for qualitative synthesis. Extracted data were categorized into three domains: (1) AI- and ML-based diagnostic models, (2) chatbot-enabled mental health support systems and (3) ethical and privacy considerations. Analytical dimensions included algorithmic performance, clinical outcomes, data governance and equity of access.

AI-driven interventions improved accessibility, diagnostic accuracy and therapeutic personalization. Chatbots such as Woebot, Wysa and Tess effectively reduced symptoms of depression and anxiety, increased user engagement and provided scalable support, particularly during the COVID-19 pandemic. ML models, including MentalBERT, MentalRoBERTa and SR-BERT, achieved F1 scores of 68-93% in mental health classification tasks. However, limitations included dataset bias, lack of longitudinal evidence and limited cross-cultural generalizability. Ethical analyses revealed persistent challenges concerning privacy, informed consent, algorithmic bias and accountability.

AI technologies, when integrated with human oversight, offer transformative potential for global mental health systems by creating interconnected and adaptive care networks. These technologies can enhance efficiency, reduce barriers to care and support data-driven public health strategies. However, successful deployment depends on clear ethical frameworks that promote transparency, respect cultural contexts and preserve human oversight. Future research should prioritize longitudinal studies, inclusive datasets and ethical frameworks that maintain human-centered values in AI-enabled mental health systems.