Furor over use of ChatGPT in mental healthcare

SAN FRANCISCO – A digital mental health company called Koko is creating an ethical flap for using GPT-3 technology without informing users. ChatGPT is a variant of GPT-3, which creates human-like text based on prompts, both created by OpenAI. Rob Morris (pictured) – co-founder of Koko, a free mental health service and nonprofit that partners with online communities to find and treat at-risk individuals – wrote in a Twitter thread on Friday that his company used GPT-3 chatbots to help develop responses to 4,000 users.

Morris said in the thread that the company tested a “co-pilot approach with humans supervising the AI as needed” in messages sent via Koko peer support, a platform he described in an accompanying video as “a place where you can get help from our network or help someone else.”

“We make it very easy to help other people and with GPT-3 we’re making it even easier to be more efficient and effective as a help provider,” Morris said in the video.

Koko users were not initially informed the responses were developed by a bot, and “once people learned the messages were co-created by a machine, it didn’t work,” Morris wrote on Friday.

“Simulated empathy feels weird, empty. Machines don’t have lived, human experience so when they say, ‘that sounds hard’ or ‘I understand’, it sounds inauthentic,” Morris wrote in the thread. “A chatbot response that’s generated in 3 seconds, no matter how elegant, feels cheap somehow.”

However, on Saturday, Morris tweeted “some important clarification.”

“We were not pairing people up to chat with GPT-3, without their knowledge. (in retrospect, I could have worded my first tweet to better reflect this),” the tweet said.

“This feature was opt-in. Everyone knew about the feature when it was live for a few days.”

Morris said Friday that Koko “pulled this from our platform pretty quickly.” He noted that AI-based messages were “rated significantly higher than those written by humans on their own,” and that response times decreased by 50% thanks to the technology.

Nevertheless, the experiment led to outcry on Twitter, with some public health and tech professionals calling out the company on claims it violated informed consent law, a federal policy which mandates that human subjects provide consent before involvement in research purposes.

“This is profoundly unethical,” media strategist and author Eric Seufert tweeted on Saturday.

“Wow I would not admit this publicly,” Christian Hesketh, who describes himself on Twitter as a clinical scientist, tweeted Friday. “The participants should have given informed consent and this should have passed through an IRB [institutional review board].”

In a statement to Insider on Saturday, Morris said the company was “not pairing people up to chat with GPT-3” and said the option to use the technology was removed after realizing it “felt like an inauthentic experience.”

“Rather, we were offering our peer supporters the opportunity to use GPT-3 to help them compose better responses,” he said. “They were getting suggestions to help them write more supportive responses more quickly.”

Morris told Insider that Koko’s study is “exempt” from informed consent law, and cited previous published research by the company that was also exempt.

“Every individual has to provide consent to use the service,” Morris said. “If this were a university study (which it’s not, it was just a product feature explored), this would fall under an ‘exempt’ category of research.”

He continued: “This imposed no further risk to users, no deception, and we don’t collect any personally identifiable information or personal health information (no email, phone number, ip, username, etc).”

Still, the experiment is raising questions about ethics and the gray areas surrounding the use of AI chatbots in healthcare overall, after already prompting unrest in academia.

Arthur Caplan, professor of bioethics at New York University’s Grossman School of Medicine, wrote in an email to Business Insider, an online publication, that using AI technology without informing users is “grossly unethical.”

“The ChatGPT intervention is not standard of care,” Caplan told Insider. “No psychiatric or psychological group has verified its efficacy or laid out potential risks.”

He added that people with mental illness “require special sensitivity in any experiment,” including “close review by a research ethics committee or institutional review board prior to, during, and after the intervention.”

Caplan said use of GPT-3 technology in such ways could impact its future in the healthcare industry more broadly. “ChatGPT may have a future as do many AI programs such as robotic surgery,” he said. “But what happened here can only delay and complicate that future.”

Morris told Insider his intention was to “emphasize the importance of the human in the human-AI discussion.”

“I hope that doesn’t get lost here,” he said.