Another dilemma?

Following up on my posts (this and this) regarding puzzles generated by moral uncertainty, here is another curious case.

Dr. Alice Kowalska believes that a steroid injection will be good for her patient, Bob. However, due to a failure of introspection, she also believes that she does not believe that a steroid injection will be beneficial to Bob. Should she administer the steroid injection?

In other words: Should Dr. Kowalska do what she thinks is good for her patient, or should she do what she thinks she thinks is good for her patient?

The earlier posts pushed me in the direction of thinking that subjective obligation takes precedence over objective obligation. That would suggest that she should do what she thinks she thinks is good for her patient.

But doesn’t this seem mistaken? After all, we don’t want Dr. Kowalska to be gazing at her own navel, trying to figure out what she thinks is good for the patient. We want her to be looking at the patient, trying to figure out what is good for the patient. So, likewise, it seems that her action should be guided by what she thinks is good for the patient, not what she thinks she thinks is good for the patient.

How, though, to reconcile this with the action-guiding precedence that the subjective seems to have in my previous posts? Maybe it’s this. What should be relevant to Dr. Kowalska is not so much what she believes, but what her evidence is. And here the case is underdescribed. Here is one story compatible with what I said above:

1. Dr. Kowalska has lots of evidence that steroid injections are good for patients of this sort. But her psychologist has informed her that because of a traumatic experience involving a steroid injection, she has been unable to form the belief that naturally goes with this evidence. However, Dr. Kowalska’s psychologist is incompetent, and Dr. Kowalska indeed has the belief in question, but trusts her psychologist and hence thinks she does not have it.

In this case, it doesn’t matter whether Dr. Kowalska believes the injection would be good for patient. What matters is that she has lots of evidence, and she should inject.

Here is another story compatible with the setup, however:

2. Dr. Kowalska knows there is no evidence that steroid injections are good for patients of this sort. However, her retirement savings are invested in a pharmaceutical company that specializes in these kinds of steroids, and wishful thinking has led to her subconsciously and epistemically akratically forming the belief that these injections are beneficial. Dr. Kowalska does not, however, realize that she has formed this subconscious belief.

In this case, intuitively, again it doesn’t matter that Dr. Kowalska has this subconscious belief. What matters is that she knows there is no evidence that the injections are good for patients of this sort, and given this, she should not inject.

If I am right in my judgments about 1 and 2, the original story left out crucial details.

Maybe we can tell the original story simply in terms of evidence. Maybe Dr. Kowalska on balance has evidence that the injection is good, while at the same time on balance having evidence that she does not on balance have evidence that the injection is good. I am not sure this is possible, though. The higher order evidence seems to undercut the lower order evidence, and hence I suspect that as soon as she gained evidence that she does not on balance have evidence, it would be the case that on balance she does not have evidence.

Here is another line of thought suggesting that what matters is evidence, not belief. Imagine that Dr. Kowalska and Dr. Schmidt both have the same evidence that it is 92% likely that the injections would be beneficial. Dr. Schmidt thereupon forms the belief that the injections would be beneficial, but Dr. Kowalska is more doxastically cautious and does not form this belief. But there is no disagreement between them as to the probabilities on the evidence. Then I think there should be no disagreement between them as to what course of action should be taken. What matters is whether 92% likelihood of benefit is enough to outweigh the cost, discomfort and side-effects, and whether the doctor additionally believes in the benefit is quite irrelevant.