AI chatbots are being used in healthcare, both in and out of the doctor’s office, more than ever before. Patient and medical providers alike are turning to AI for treatments and diagnoses, but transparency remains a key concern, a new survey by Pew Research Center shows.
A poll of 3,488 US adults conducted between June 22 and June 28 found that a quarter of respondents use chatbots to diagnose symptoms. Those turning to chatbots for health mainly cite convenience and affordability as their motive. Patients also use AI chatbots to find information quickly and at no cost, better understand a doctor’s diagnosis, interpret lab results and decide whether to seek in-person care.
Of these survey participants, almost half (47%) say the health information generated by the chatbot was “extremely” helpful, and 48% said the information was “somewhat” helpful. Only 5% said the results were not too helpful or not at all helpful.
AI chatbots work by simulating a conversation between two parties and responding to your prompts or inputs using the massive amounts of information the models were trained on. Many people appreciate the convenience they offer, as chatbots are available 24/7 and can often be accessed for free and sometimes without an account. Chatbots, however, are known to “hallucinate,” or generate false information in their responses, which can be particularly dangerous or alarming when your personal health is in question. Among other issues, they’re trained to produce a probable answer, not necessarily one grounded in truth, and they’re likely to sound particularly confident about that answer.
But generative artificial intelligence is also showing up at the doctor’s office — and Pew found that Americans have greater concerns about its use there.
Read also: ChatGPT Health Will See You Now, and It’s a Big Step Forward for Medical Queries
When doctors turn to chatbots
While recent numbers show that about one in three US adults have used chatbots for health-related questions, medical providers are also increasingly turning to AI in the office for note-taking, interpreting exams and more.
While views on AI use in doctors’ offices vary, one factor remains of critical concern to patients: transparency.
Of those surveyed, 72% say it is “extremely or very important” for their medical provider to disclose any use of AI. The survey also found that a vast majority of people (eight in 10) want to be told if AI is being used directly for care or medical decisions, such as diagnoses and the analysis of scans, X-rays and lab results. AI is frequently used for triage purposes — and has even been shown to outperform human ER doctors — as well as to complete clinical tasks more quickly. For instance, AI has been used effectively to reduce the window of X-ray interpretations from 11.2 to 2.7 days.
Although Americans want advance notice when AI is used in their healthcare, nearly half (47%) are unsure whether it has ever played a role in their treatment, underscoring limited awareness of when, where and how the technology is used.
Americans are not fully comfortable with AI
The survey shows that Americans may find AI chatbots convenient, but not entirely trustworthy.
Views are mixed on sharing personal health histories and diagnoses with AI. Among those surveyed, 29% say they are extremely or very comfortable sharing health information with chatbots, while 26% say they are uncomfortable doing so.
The findings suggest that although AI is increasingly used for health information and care, public confidence and understanding have not kept pace with its adoption.