Industry News
In Japan, 51% of doctors consult generative AI when deciding on treatment options, while 61% still feel uncertain about the answers it provides.
A survey conducted by Deloitte Japan provided insights into how Japanese doctors utilize generative AI: Over half (51%) of them use it to gather information when determining treatment plans for patients.
The proportion is even higher in the field of cancer. Among doctors specializing in solid tumors, 59% use these methods when selecting treatment options, and 46% utilize them for clinical diagnosis – both figures are higher than the overall average. This is quite understandable; given that tumor treatment guidelines are updated frequently and there are numerous clinical trials underway, it has become increasingly difficult for individuals to keep up solely by relying on personal knowledge and literature.
The frequency is also quite high. Over the past year, 30% of people engaged in this activity “several times a week”, 26% “almost every day”, and 19% “several times a month”.
The turning point comes in the latter half of the report: 61% of the respondents expressed concern or discomfort regarding the accuracy and credibility of the information generated by generative AI systems.
Using it frequently on one hand while feeling uncertain about its reliability on the other – this combination is actually not considered contradictory in hospitals. Doctors never blindly trust textbooks or information provided by pharmaceutical companies either. The problem lies in the fact that when AI makes mistakes, its tone remains just as assertive, and the claims it makes may often turn out to be fabricated.
The term “investigation” simply means “gathering information”; it does not imply that doctors hand over the decision-making power to algorithms. However, in practice, this distinction can become quite blurred. With limited outpatient time, if the first suggestion provided by AI seems reasonable, it is likely to become the starting point for further consideration, and all subsequent decisions will be based around that initial proposal.
The figure in Japan is actually higher than many people expect, and this also reveals another fact: even if hospitals do not provide official tools, doctors will still find other ways to use them – and it is very likely that they will turn to publicly available chatbots. The question of whether medical record information is being used in these chats was not addressed in the investigation; this is what hospitals should really be concerned about.