Table 3 Pharmacists’ perceptions of AI recommendations

Positive

Auxiliary tool

“If my dose is similar to the AI dose, I probably change it. But again, I'd probably end up splitting the difference rather than just blindly accepting what the models are telling me, I probably would use it as a facet of what we're trying to the dosing tools.”

Double-check dosing regimen

“You know, if I was very different from the AI, I would double-check myself. I would take it out and take it seriously, and maybe I missed something. So I think it would be beneficial even though I may not agree with its dose.”

Develop optimal dosing scheme

“Then, if it's discrepant, I think it would cause me to put some more thought into the patient, so it would help challenge my initial decision, which would help me make a more informed and thoughtful decision.”

“I think I think it's a great project, and I think it would be really helpful and useful because everyone's scope is fairly narrow.”

Decision making on uncertain cases

“I would be on edge and conflicted for this case. With his age, I’m less sure that I would use q8h dosing. But I would take pause to think about it, and AI would help me make the decision”

“I think if, in this case, it's the AI suggested that there was a 70% probability that I was overdosing, I would potentially modify the dose.”

Career development

“AI is a new system. So if we could utilize it, that would also be like another piece to the puzzle that we can potentially use. So yeah, if I knew that the AI was staying me to go higher, I would probably use it.”

Negative

Self-confidence

“I was pretty confident that I was right. No amount of computer would be able to say to kind of convince me otherwise.”

“I would say I wouldn't worry that the models indicate me inappropriate, but I like my dosing better.”

Mistrust of AI

“I think if in this case, it's the AI suggested that there was a 70% probability that I was overdosing; otherwise, I won't change my dosage. In short, I don't fully trust it”

Unclear rationale for recommendations

“The reason I wouldn't change my dose scheme is that I don't know why, what factors AI is based on to tell me I'm underdosing or overdosing.”

“I would say no unless I could be given a reason why the AI model says that it's overdosing. I would want to have those responses to what the reasons would be”