Abstract: Crowdsourced online mental health services have emerged as an important complement to traditional one-to-one counseling by allowing individuals with psychological concerns to post questions and receive responses from multiple service providers. This model helps meet patients’ diverse needs, improves response efficiency, and partly alleviates the imbalance between the rising demand for mental health support and the limited supply of professional services. However, because the entry threshold for service providers is relatively low, the quality of psychological advice on such platforms varies greatly. Low-quality responses may mislead patients, delay appropriate treatment, and undermine platform development. Existing studies have mainly examined online health services from the patient perspective or focused on one-to-one service settings, while limited attention has been paid to how service quality is identified, influenced, and improved in crowdsourced online mental health services from the service-provider perspective. To address this gap, this study develops an integrated framework based on the logic of “identification-influence-intervention.”
First, this study proposes an automated approach to identifying service quality in crowdsourced online mental health services. Drawing on counseling theory, it conceptualizes service quality as a multidimensional construct. High-quality psychological responses should help patients clarify their problems, understand the causes and consequences of their distress, and develop feasible coping strategies. In online text-based interactions, responses also include opening and closing components that establish rapport and complete communication. Accordingly, this study develops a five-dimensional evaluation framework for expert assessment. Compared with prior studies that rely on fragmented or inconsistent expert criteria, this framework offers a basis for measuring psychological service quality. Building on this framework, the study further proposes a multi-task learning model to predict service quality automatically. In addition to the main task of predicting response quality, the model incorporates emotional matching and informational matching as auxiliary tasks. Emotional matching captures the extent to which a response fits the help-seeker’s emotional needs, while informational matching reflects how accurately and comprehensively the response addresses the patient’s problem and context. Jointly modeling these related tasks can improve both predictive accuracy and theoretical interpretability.
Second, this study investigates how the professional level of the first service provider affects the participation behavior of subsequent providers. In crowdsourced online mental health services, the first respondent serves as an observable peer whose status and response quality provide important signals to later participants. Based on peer effects theory and expectancy-value theory, this study argues that the first provider’s professional level may generate two competing mechanisms. On the one hand, a highly professional first provider may create a competition-suppression effect. Because patients face substantial information asymmetry and may rely on professional signals when selecting or rewarding answers, later providers may perceive a lower probability of success and become less willing to participate. On the other hand, a highly professional first provider may generate a social-influence effect by signaling that the patient’s problem is serious or worthy of attention, thereby encouraging additional providers to contribute. The study further examines the moderating roles of monetary reward and patient psychological distress. Higher monetary rewards may increase the perceived value of participation and weaken the discouraging effect of competition. In contrast, higher psychological distress may strengthen the perceived advantage of the first high-level provider, making subsequent providers more cautious about participation.
Third, this study explores how providers’ own experience and peer experience jointly shape the dynamic improvement of service quality. Based on learning theory, the study argues that providers can improve later responses by reflecting on prior service experiences. However, the benefits of self-experience may exhibit diminishing marginal returns as additional experience provides less new information and may lead to routine response patterns. The study also distinguishes between successful and failed experiences, proposing that successful experiences may have a stronger positive effect because they reinforce effective practices and enhance self-efficacy. Meanwhile, the public visibility of responses enables providers to learn from peers. By observing others’ high-quality responses, providers may absorb useful knowledge and improve their own service quality. Nevertheless, peer experience may also show diminishing returns due to information redundancy and cognitive overload. Furthermore, the study proposes a complementary relationship between self-experience and peer experience: providers with richer personal experience may better integrate knowledge gained from peers, thereby amplifying the positive effect of peer learning.
This study contributes to the literature in three ways. It develops a theory-based framework for identifying online mental health service quality, enriches research on service-provider behavior in crowdsourced health platforms, and extends the understanding of service quality improvement by treating quality as a dynamic process shaped by individual and peer learning. Practically, the findings can help platforms design automated quality evaluation tools, optimize participation incentives, manage peer influence, and develop targeted strategies to improve the quality of crowdsourced online mental health services.

Key words: crowdsourced online mental health services, service quality, participation behavior, influencing mechanisms