Goldie Jessie, Dennis Simon, Hipgrave Lyndsey, Coleman Amanda
Melbourne School of Psychological Sciences, Medicine, Dentistry and Health Sciences, University of Melbourne, Grattan Street, Parkville, Melbourne, 3010, Australia, 61 467607835.
Monash School of Psychological Sciences, Faculty of Medicine, Nursing and Health Sciences, Monash University, Melbourne, Australia.
JMIR Hum Factors. 2025 Sep 16;12:e71065. doi: 10.2196/71065.
Generative artificial intelligence (AI) chatbots have the potential to improve mental health care for practitioners and clients. Evidence demonstrates that AI chatbots can assist with tasks such as documentation, research, counseling, and therapeutic exercises. However, research examining practitioners' perspectives is limited.
This mixed-methods study investigates: (1) practitioners' perspectives on different uses of generative AI chatbots; (2) their likelihood of recommending chatbots to clients; and (3) whether recommendation likelihood increases after viewing a demonstration.
Participants were 23 mental health practitioners, including 17 females and 6 males, with a mean age of 39.39 (SD 16.20) years. In 45-minute interviews, participants selected their 3 most helpful uses of chatbots from 11 options and rated their likelihood of recommending chatbots to clients on a Likert scale before and after an 11-minute chatbot demonstration.
Binomial tests found that Generating case notes was selected at greater-than-chance levels ( 15/23, 65%; P=.001), while Support with session planning (P=.86) and Identifying and suggesting literature (P=.10) were not. Although 55% (12/23) were likely to recommend chatbots to clients, a binomial test found no significant difference from the 50% threshold (P=.74). A paired samples t test found that recommendation likelihood increased significantly (19/23, 83%; P=.002) from predemonstration to postdemonstration.
Findings suggest practitioners favor administrative uses of generative AI and are more likely to recommend chatbots to clients after exposure. This study highlights a need for practitioner education and guidelines to support safe and effective AI integration in mental health care.
生成式人工智能（AI）聊天机器人有潜力改善从业者和客户的心理健康护理。有证据表明，AI聊天机器人可以协助完成诸如文档记录、研究、咨询和治疗练习等任务。然而，研究从业者观点的研究有限。
这项混合方法研究调查：（1）从业者对生成式AI聊天机器人不同用途的看法；（2）他们向客户推荐聊天机器人的可能性；（3）观看演示后推荐可能性是否增加。
参与者为23名心理健康从业者，包括17名女性和6名男性，平均年龄为39.39岁（标准差16.20）。在45分钟的访谈中，参与者从11个选项中选出他们认为聊天机器人最有用的3种用途，并在11分钟的聊天机器人演示前后，用李克特量表对他们向客户推荐聊天机器人的可能性进行评分。
二项式检验发现，生成病例记录的选择率高于随机水平（15/23，65%；P = 0.001），而会话计划支持（P = 0.86）和识别并推荐文献（P = 0.10）则未达到。虽然55%（12/23）的人可能会向客户推荐聊天机器人，但二项式检验发现与50%的阈值没有显著差异（P = 0.74）。配对样本t检验发现，从演示前到演示后，推荐可能性显著增加（19/23，83%；P = 0.002）。
研究结果表明，从业者倾向于将生成式AI用于管理用途，并且在接触后更有可能向客户推荐聊天机器人。这项研究强调需要对从业者进行教育并制定指导方针，以支持在心理健康护理中安全有效地整合AI。
