Thinking machines, on the couch

When a relative first showed David Lopis an AI tool, it struck him as exciting and faintly ridiculous at the same time.

“It looked really interesting,” the psychologist said.

“But what on earth could I do with it in my actual work? It seemed like a gimmick.”

He went looking for answers, contacting AI companies and quizzing colleagues.

Mostly he found confusion, so he set about filling the gap himself.

He now works as an AI consultant and educator for clinicians, alongside his role as a board-approved supervisor, and has written a book, AI for Mental Health Clinicians.

The timing is awkward as AI has had a bad run in mental health headlines, from chatbots accused of nudging users toward suicide to people taking up with AI girlfriends.

Lopis doesn’t wave any of it away.

“I get all the doom and gloom stories and the horror stories,” he said.

“They really suggest the need for clinician involvement and oversight and risk management, which I think is absolutely critical.”

What they don’t justify, in his view, is walking away.

The most popular tool among clinicians is the AI scribe, which drafts notes and reports from a session for the clinician to check and finalise.

Lopis said paperwork that once took hours can be done in 10 to 20 minutes, leaving more time for clients.

“It’s become a bit of – can we say, as Jews – our holy grail of sorts,” he said.

His own favourite trick is less obvious.

A clinician can tell ChatGPT or Claude that a client loves Harry Potter, basketball and trains, outline the problems they’re dealing with while leaving out anything identifying, and ask for therapeutic strategies built around those interests.

“It’s on their level,” he said, “through their interests and through how they see the world.”

Consent is where the profession is most divided.

Recording sessions for an AI scribe is contentious, and Lopis cited a psychiatrist who told patients he wouldn’t see them unless they agreed to it.

It cuts the other way too, with some clients bringing their own recording devices so they can review a session afterwards.

Lopis won’t hand down rulings.

“As a Jew, I like to answer questions with more questions,” he said.

His job, as he sees it, is to lay out the issues so the profession can decide.

He understands why AI frightens people, pointing to military applications, the sense of an arms race, and algorithms making decisions in aged care and the NDIS.

The bigger problem, he argues, is that for many clinicians a vendor’s sales event is the full extent of their AI education.

“A lot of people don’t know what they don’t know,” he said.

“They read a couple of articles and then get scared off.”

He believes there is a trust deficit to close before clients will have faith in how their therapists use the technology, and that one workshop won’t keep anyone across a field moving at “breakneck speed”.

AI for Mental Health Clinicians is available at: amazon.com.au/dp/1764599403