JMIR Publications

ABSTRACT

Background:

The use of artificial intelligence in medicine will change multiple areas of medicine, including uncertainty in medicine and medical education.

Objective:

This viewpoint aims to comprehend how artificial intelligence (AI) will impact uncertainty in medicine, specifically epistemic, systemic and linguistic-based reducible and irreducible forms of uncertainty in medicine, and how it will impact medical education.

Methods:

Theoretical analysis.

Results:

Using Djulbegovic, Hozo & Greenland’s model of uncertainty as a guiding framework, we analyzed and predicted the impact of medical AI on reducible and irreducible uncertainty in medicine. We argued that in their strongest forms, AI technologies have the potential to address reducible epistemic uncertainty due to measurement and systematic error, subjective uncertainty due to information absence or abundance, uncertainty arising from patient beliefs, values, and preferences, and linguistic uncertainty due to unclear definitions of illness. In addition, these technologies cannot fully overcome epistemological limitations of random error, the problem of induction, and model uncertainty, which underlie many forms of knowledge production in medicine. AI will positively and negatively impact the areas of communication, information, knowledge, perspectives in medical education, and to address the negative impacts, medical education needs to be reformed.

Conclusions:

AI may hold promise to address reducible forms of uncertainty, but cannot extinguish the irreducible forms of uncertainty due to epistemic issues in practice. In addition, medical education needs to be reformed to prevent medical students from learning incorrect information and harming patients. Clinical Trial: Not applicable.