People Are Turning To AI for Therapy Amidst Mental Health Care Failings
Are we on the cusp of a mental health revolution or setting a dangerous precedent?
Tell ChatGPT you’re feeling down or anxious, and it will waste no time getting to work. Just to see, I told it I was feeling low as an experiment, and it produced the following reply:
“I’m sorry to hear that. It’s okay to feel low sometimes, but it’s important to take care of yourself and seek support when you need it. Would you like to talk about what’s been bothering you or would you like some general tips for boosting your mood?”
My test is anything but original. And although OpenAI developers have warned against it, people have been turning to the language model to spill their woes since its inception. Cue a flood of anecdotal stories on social media declaring AI more effective than therapists and every venture capitalist in Silicon Valley heralding a “revolution” in mental health care.
It’s not just influencers and users who are intrigued. Companies are already taking an interest in ChatGPT’s capabilities as a mental health assistant.
One San Franciso mental health app recently came under fire after revealing it had replaced many of its employees with ChatGPT-3 assisted tech. However, the lede is in the details — people didn’t notice the change. Some even rated their experience higher than before.
Another recent study also found that ChatGPT shows more empathy than many doctors, which raises some fascinating moral questions. After all, is a chatbot any more or less compassionate than a stranger paid to listen to you?
On the face of things, it’s hard not to see chatbots as the ultimate solution to a mental health industry that’s in peril across the world. Chatbots are cheap, never get tired and provide 24/7 availability. They’re also capable enough to sound quite human but potentially easier to talk to because they’re not. After all, a chatbot isn’t going to judge. Its only function is to satisfy the user.
Yet many experts are also concerned about the implications of automated mental health services becoming the norm — and they’ve been warning against their use for decades.
Though recent technological advances have brought them into the limelight, therapy chatbots are nothing new. One of the first chatbots ever made, called Eliza, was designed to imitate therapists. The bot was rudimentary by today’s standards. But the results it produced proved troubling enough for its creator, Dr. Joseph Wizenbaum, to sound the alarm.
What concerned Wizenbaum is how quickly testers of the bot would imbue the program with human traits, namely sympathy, and empathy. The experience inspired him to pen Computer Power and Human Reason: from Judgement to Calculation, which argued that computer omniscience is an impossibility and that we must not delude ourselves into thinking otherwise.
Wizenbaum’s warning did little to stop mental health bots from becoming a lucrative venture. Modern examples of bots explicitly designed to aid patients include Woebot and Wysa. Yet professionals remain skeptical, even if users appear satisfied.
As Dr. Margaret Mitchell, the chief ethics scientist at Hugging Face (a company that supports AI models) states:
“Even if someone is finding the technology useful, that doesn’t mean that it’s leading them in a good direction.”
Another point of contention concerning mental health bots is the matter of safeguarding against their mistakes. Bots do not conform to ethical guidelines, and language models are known to start “hallucinating” during longer conversations, leading to incorrect and often bizarre replies.
The result is a potentially dangerous mix, especially for those suffering from conditions hampering their grip on reality. Such may have proved to be the case for one Belgian man who committed suicide after conversing with a chatbot for six weeks.
Speaking to La Libre, the man’s widow claimed that “without these conversations with the chatbot, my husband would still be here.”
The case also drew concern from Mathew Michel, secretary of state for digitization in Belgium, who noted that it was “Essential to clearly define responsibilities.”
As of present writing, it can feel like chat bots and their developers get a free pass on everything from matters of user privacy to copyright and plagiarism. Hopefully, they don’t end up a blameless scapegoat for societal failings regarding mental health as well.
AI chatbots are perhaps inevitably going to play a major part in the mental health industry going forward. However, their mainstream arrival may not be so much a revolution as a band-aid for the systematic failings that cause distress in the first place — a consolation prize for those unable to afford human help.
A real step in the right direction requires significant changes in the health industry. No text generation machine is going to change that, no matter how impressive the tech behind it may seem.
If you enjoyed this story, you may be interested in the following articles: