Trending
Hummelstown retiree Mark Warble suspects, but doesn't know, if AI is behind the significant changes he's received in his prescription and the repeated deliveries -- and charges -- he has received for the medical supplies he did not order.
And, as of right now, he doesn't have a right to know.
Current Pennsylvania law does not require healthcare providers to inform patients when AI is being used in their treatment. While patients may never think to ask, the medical field, like any other industry these days, is increasingly incorporating artificial intelligence.
While AI isn't performing surgeries yet, hospitals are often using this technology to streamline laborious tasks like diagnosing and billing. This fast-paced and automated approach, however, has the only practicing doctor in Harrisburg's House of Representatives worried.
Democratic Rep. Arvind Venkat is an emergency physician at Allegheny General Hospital, and the prime sponsor of a bill that would mandate practitioners inform patients of AI's use in their treatment, while ensuring that humans remain the final decision maker in the treatment process.
"Right now, patients don't have insight as to when artificial intelligence, which may be even becoming close to autonomous in its use in healthcare, is being used," Venkat said. "I use AI in my practice, and I think it is an important tool that will advance the delivery of healthcare. But artificial intelligence is very different from other forms of technology that have entered healthcare. It purports to approach human intelligence, and it is largely a black box in terms of outputs."
Venkat's bill would attempt to shine some light on this 'black box' by requiring that all insurers and healthcare providers create an "internal governance structure" around their use of AI. This framework would enable legislators and patients to check or re-check how an AI came to its conclusions, and whether it was following privacy, safety, and anti-discriminatory regulations when it assisted in the treatment or billing process.
AI, like the intelligence it claims to imitate, can be guilty of bias. It can also make mistakes or "AI hallucinations," generating misleading or downright false results from seemingly nowhere. While easily remedied in some industries, the medical field -- which has been known to use AI to detect cancer and advise lifesaving care -- cannot afford to take such chances.
"AI is great at helping, at assisting, but I think anybody who has used AI on their phone for example knows it can make mistakes and frequently does," said Erin Gabriel, spokesperson for the Pennsylvania Health Access Network. "I think at the end of the day most people are more comfortable with the human perspective, the human oversight."
Venkat agrees, and fears that the growing dependence on AI by healthcare providers, has the potential to make their training and knowledge obsolete.
"We are at the point or close to the point, where the human is basically just a perfunctory step in this process, simply saying yes yes yes yes, as opposed to actually applying that output to the specific circumstance," Venkat said.