By Jane Wakefield
Technology reporter
Would you share your deepest anxiety with Alexa? Or maybe ask Siri for some emotional support after a particularly stressful day?
We are increasingly turning to chatbots on smart speakers or websites and apps to answer questions.
And as these systems, powered by artificial intelligence (AI) software, become ever more sophisticated, they are starting to provide pretty decent, detailed answers.
But will such chatbots ever be human-like enough to become effective therapists?
Computer programmer Eugenia Kuyda is the founder of Replika, a US chatbot app that says it offers users an “AI companion who cares, always here to listen and talk, always on your side”.
Launched in 2017, it now has more than two million active users. Each has a chatbot or “replika” unique to them, as the AI learns from their conversations. Users can also design their own cartoon avatar for their chatbot.
Ms Kuyda says that people using the app range from autistic children who turn to it as a way to “warm up before human interactions”, to adults who are simply lonely and need a friend.
Others are said to use Replika to practise for job interviews, to talk about politics, or even as a marriage counsellor.
And while the app is designed primarily to be a friend or companion, it also claims it can help benefit your mental health, such as by enabling users to “build better habits and reduce anxiety”.
Around the world there are almost one billion people with a mental disorder, according to the World Health Organization (WHO). That is more than one person out of every 10.
The WHO adds that “just a small fraction of people in need have access to effective, affordable and quality mental health care”.
And while anyone with a concern for either his or herself, or a relative, should go to a medical professional in the first place, the growth of chatbot mental health therapists may offer a great many people some welcome support.
Dr Paul Marsden, a member of the British Psychological Society, says apps that aim to improve your mental wellbeing can help, but only if you find the right one, and then only in a limited way.
“When I looked, there were 300 apps just for anxiety… so how are you supposed to know which one to use?
“They should only be seen as a supplement to in-person therapy. The consensus is that apps don’t replace human therapy.”
Yet at the same time, Dr Marsden says he is excited about the power of AI to make therapeutic chatbots more effective. “Mental health support is based on talking therapy, and talking is what chatbots do,” he says.
Dr Marsden highlights the fact that leading AI chatbot firms, such as OpenAI, the company behind the recent headline-grabbing ChatGPT, are opening up their technology to others.
He says this is enabling mental health apps to use the best AI “with its vast knowledge, increasing reasoning ability, and proficient communication skills” to power their chatbots. Replika is one such provider that already uses OpenAI’s technology.
New Tech Economy is a series exploring how technological innovation is set to shape the new emerging economic landscape.
But what if a person’s relationship with their chatbot therapist becomes unhealthy? Replika made headlines in February when it was revealed that some users had been having explicit conversations with their chatbot.
The news stories appeared after Luka, the firm behind Replika, updated its AI system to prevent such sexual exchanges.
Not all users are happy at the change. One wrote on Reddit: “People who found a refuge from loneliness, healing through intimacy, suddenly found it was artificial not because it was an AI but because it was controlled by people.”
Luka’s move may be related to the fact that also in February, Italy’s data protection agency banned it from using the personal data of Italians.
The Italian watchdog claimed that the app was used by under-18s who were getting “replies which are absolutely inappropriate for their age. It added that the app could also “increase the risks for individuals still in a developmental stage or in a state of emotional fragility”.
The move may limit the use of Replika in Italy, and Luka could be fined. It says it is “working closely with Italian regulators and the conversations are progressing positively”.
UK online privacy campaigner Jen Persson says there needs to be more global regulation of chatbot therapists.
“AI companies that make product claims about identifying or supporting mental health, or that are designed to influence your emotional state, or mental well-being, should be classified as health products, and subject to quality and safety standards accordingly,” she says.
Ms Kuyda argues that Replika is a companion, like owning a pet, rather than a mental health tool. She adds that it shouldn’t be seen as a replacement for help from a human therapist.
“Real-life therapy provides incredible insights into the human psyche which is not just through text or words, but by seeing you in person and seeing the body language, your emotional responses as well as an incredible knowledge of your history,” she says.
Other apps in the mental health sector are far more cautious about using AI in the first place. One of those is meditation app Headspace, which has more than 30 million users, and in the UK is approved by the NHS.
“Our core belief and entire business model at Headspace Health is anchored in human-led and human-focused care – the connection our members have via live conversations with coaches and therapists through chat, video or in-person is irreplaceable,” says Headspace’s chief executive Russell Glass.
He adds that while Headspace does use some AI, it does it “highly selectively”, and while maintaining “a depth of human involvement”. The firm does not use AI to chat to users, instead Mr Glass says it only utilises it for things like providing users with personalised content recommendations, or assisting human care providers to write their notes.
Yet Dr Marsden says that AI-powered therapy chatbots will only continue to get better. “New AI chatbot technology appears to be evolving skills for effective mental health support, including empathy and understanding of how the human mind works,” he says.
His comments come after a recent study by Cornell University in New York State, which put ChatGPT through a number of tests that look at how well people can understand that others might think differently. The AI’s scores were equivalent to those of a nine-year-old child.
Previously this sort of cognitive empathy had been regarded as uniquely human.