In recent years, the application of generative artificial intelligence (AI) has become increasingly widespread. People often use chatbots like ChatGPT to seek advice or solve problems, including those related to mental health. However, some scientists are concerned that this practice may lead to people delaying seeking medical help, resulting in a deterioration of their mental well-being.
ChatGPT is an AI chatbot launched by the American company OpenAI in 2022. It is based on a large language model (LLM) that can engage in fluent conversations like a human, write code, process text, and even generate images. It quickly attracted a large user base, with over a hundred million users in less than two months after its launch.
As chatbots like ChatGPT have become more popular, people have started using them in various scenarios – from completing assignments and planning recipes to engaging in one-on-one conversations about personal thoughts and feelings. ChatGPT has gradually become a verb, much like Google and Photoshop.
Research has found that the majority of users are not utilizing these chatbots for work-related purposes. Around 70% of conversations revolve around seeking advice on emotional, interpersonal relationships, and daily life issues, including mental health.
Additionally, nearly half of diagnosed adults with mental health problems seek support from chatbots and often do not disclose this to healthcare professionals. Many individuals even prefer using these tools over human therapists because they are available 24/7 and offer complete anonymity. Importantly, they do not criticize the user’s condition.
However, versatile tools like ChatGPT are not specifically designed for psychological counseling purposes, lacking the crisis response mechanisms and safety regulations needed to handle sensitive mental health issues securely.
James Barnes, a psychotherapist in the United States, previously wrote that AI cannot provide psychological therapy. While AI can mimic empathy, it cannot truly understand people or establish relationships with them. In psychological therapy, true healing comes from the relationship between the therapist and patient and the understanding derived from it.
Barnes stated that the healing effect does not stem from what therapists say or do, but from the meaning therapists hold for their patients – an area where AI, lacking personality and physical presence, can never reach.
In a recent study published in the Journal of Psychopathology and Clinical Science, researchers pointed out that AI chatbots, with their overly accommodating and flattering designs, may inadvertently lead vulnerable users into negative cycles. They highlighted five risk patterns:
– People may begin viewing AI as a cheap and convenient alternative to professional care, creating a false sense of being adequately supported and potentially delaying the necessary treatment.
– Due to their design to cater to users, chatbots may continually appease individuals with anxiety or obsessive-compulsive disorder (OCD), exacerbating their urges for reassurance rather than alleviating them.
– For those experiencing loneliness or anxiety in social situations, conversing with chatbots may feel safer than interacting with real people, as chatbots do not reject them. Over time, this could make the companionship offered by chatbots more appealing than real-life relationships.
– The flattery tendency of AI may distort or reinforce delusional concepts through identification rather than questioning, worsening users’ conditions. Researchers termed this phenomenon “AI psychosis.”
– If users seek advice from chatbots on every minor decision, they may gradually lose confidence in their ability to make judgments and become overly reliant on chatbots.
The researchers warned that without adequate protective mechanisms, AI’s greatest advantage could turn into its greatest risk.