Artificial intelligence is becoming part of everyday life, especially in areas that were surrounded by shame and stigma. AI chatbots, unlike friends, family, and therapists, won’t judge. 37% of the adults in the UK already used AI bots for therapeutic purposes at least once [1].
It’s easy to see why. AI is available around the clock and responds instantly. But the qualities that make AI appealing can also make it risky. Emerging research suggests that excessive reliance on AI may have more risks than benefits, particularly among vulnerable users.
This article looks at both perspectives through an evidence-based psychological lens: what good and bad AI can bring to mental health.
AI Effects on Mental Well-Being
The facts below are based on research on the effects of AI on mental health. However, it’s important to note that the evidence at the moment is correlational, not causal, meaning scientists see a connection but don’t yet know how to prove it. It can also be a coincidence.
AI therapy is the most common among vulnerable populations like older adults or men, who are ashamed to seek help for their mental struggles [1]. Considering this, it’s better to assess your current well-being state with an online EQ assessment before relying on AI to navigate difficult emotions. As seen from the section below, the health risks associated with AI are undeniable.
Negative Effects of Artificial Intelligence
- AI use for therapy is highest among middle- and older-aged people: 64% of 25-34-year-olds and 15% of those aged 55+ reported using AI for mental health [1].
- Men (42%) were more likely to use chatbots than women (33%), which is probably connected to the fact that men are less likely to seek help when facing emotional difficulties [1].
- 24% of the people in the UK who use AI for mental support do so as a last resort because of long waiting lines for mental health services [1].
- A third (33%) of AI therapy users reported worsening of mental health symptoms, such as psychosis, suicidal inclinations, and feeling more depressed/anxious [1].
- Due to extensive reliance on AI for mental health, a new phenomenon in psychology appeared called “ChatGPT-induced psychosis,” which is characterized by delusional thinking, dependency, social withdrawal, and suicidal ideations [2].
- Up to 25% of adolescents who use AI develop dependency on chatting with bots [2].
- ChatGPT-3 endorsed or reinforced false statements in 26% of requests,including ones about mental health[2].
- 28% of Americans have had at least one romantic encounter with AI chatbots, which is negatively associated with mental health and relationships with real people due to compulsive overuse, social isolation, and reinforcement of negative relationship patterns [3].
- Adults aged 45 to 65 who used AI daily had 50% greater odds of reporting at least moderate depression than less frequent users. [4].
- In one experiment, AI chatbots trained for supporting mental health have failed to recognize suicidal intentions [5].
- 89% of psychologists worry AI could contribute to and promote self-harm, based on their experiences and stories when AI failed to recognize a mental health crisis [6].
Can AI Help With Mental Health? AI Benefits
Despite the concerns surrounding AI and mental health, the evidence isn’t entirely negative. A growing body of questionnaires and user feedback suggests that AI chatbots can provide meaningful support for some people.
Of course, the AI effects will depend on the current mental well-being of a person, their coping skills, and whether they received professional mental health care.
Here are some statistical findings that point to AI benefits for mental health:
- Researchers found that Therabot, an AI chatbot designed for mental health treatments, helped people with diagnosed mental health conditions to reduce 51% of depression symptoms, 31% of generalized anxiety disorder, and 19% of eating disorder symptoms [7].
- Users themselves report positive experiences. Among people who had used AI chatbots for mental health or well-being, 66% said they found them beneficial [1].
- According to users’ self-assessment [1]:
- More than a quarter (27%) reported feeling less alone after talking to a chatbot.
- 24% said AI helped them manage difficult emotions.
- Another 20% believed that interacting with a chatbot helped them avoid a potential mental health crisis.
- 21% reported that AI provided useful information about managing suicidal thoughts, including directing them toward crisis helplines and other support services.
- 68% of therapists with clients using AI chatbots report that these interactions made them feel validated and provided immediate support they wouldn’t have otherwise [6].
- Nearly half of psychologists (49%) reported that their patients’ communication with AI usually has a positive impact [6].
- 47% of psychologists believe AI therapy has the potential to improve mental health services overall [6].
- 40% of psychologists are optimistic that AI can help those patients who don’t have access to mental health services because of long waiting lists, financial barriers, geographic location, or discomfort speaking to another person [6].
The Future of AI for Mental Health
The strongest conclusion that can be drawn from the findings above is that AI in mental health will be a collaborative tool, at least in the near future. Considering all the risks for the mental well-being, mostly because of a lack of awareness about AI, it’s not ready to be a stand-alone solution for mental health.
On the other hand, the contribution and potential of AI bots to mental health cannot be ignored. They already expand access to emotional support, especially for people facing long waiting lists and financial barriers.
At the current stage of AI therapy development, users have to follow a few important rules to minimize risks:
- Don’t rely on AI during a mental health crisis.
- Verify important advice.
- Maintain real human relationships.
- Protect your privacy.
- Choose AI tools designed for mental health.
- Combine using AI mental health tools with professional mental support, when possible.
