The authors argue that transparency displaces, rather than dissolves, the problem of opacity, since the interpretable predictor used to make a black box algorithm transparent is itself opaque.

The authors hold that epistemic justification through computational reliabilism, on its own, gives physicians sufficient grounds to act on a medical AI's recommendation without further deliberation.

Computational reliabilism (CR) holds that researchers are justified in believing the results of an AI system because there is a reliable process that yields trustworthy results most of the time.

The authors endorse Rudin's view that black box algorithms must be excluded from high-stakes practices such as medicine.

Which of the following is not listed among the four reliability indicators that Durán and Formanek originally propose for computational reliabilism?

In the paper's discussion of methodological opacity, "information hiding" (drawn from Colburn and Shute) refers to:

The authors argue that the worries about patient autonomy raised against black box algorithms (e.g., paternalism, value-misalignment) are primarily caused by:

To rebut the claim that physicians cannot be held responsible for outputs of opaque algorithms they do not understand, the authors draw an analogy with physicians' use of: