I think the following view is right:
IKSE: p is part of S's evidence if S knows p non-inferentially.
Not only do I think that it's right, I think it's pretty harmless. If S knows p non-inferentially, S's belief that p is the case is non-inferentially justified. So, if we can show that IJSE is true, we can show that IKSE is true:
IJSE: If S's belief that p is non-inferentially justified, p is part of S's evidence.
It seems that IJSE is the sort of thing that Fantl and McGrath defend. They say:
(JKR) If p is knowledge-level justified for you, then p is epistemically eligible to be a reason you have for believing q, for any q.
In defense of (JKR), Fantl and McGrath say, “If your justification for a proposition is good enough for knowledge, then if it isn’t among your reasons for belief, it’s not for shortcomings in your epistemic position with respect to it.”
That seems right, but it gives rise to a worry (for me).
If IJSE is true, anything we are non-inferentially justified in believing is part of our evidence. You can be non-inferentially justified in believing p when ~p. Thus, p can be part of your evidence even if ~p.
That's a worry because I think that only true propositions constitute evidence (here). One response is to deny that there can be false, non-inferentially justified beliefs. I think that's the right thing to say, but it seems we could distinguish between (IJSE) and the following:
(CIJSE) If p is a piece of evidence and S's belief that p is non-inferentially justified, p is part of S's evidence.
Whereas IJSE commits us to a claim about the possession of evidence and a claim about what constitutes evidence, a close principle (CIJSE) only tells us what it takes to take possession of a piece of evidence. It is, for that reason, weaker than IJSE. It also seems that Fantl and McGrath's motivation for (JKR) does not support anything stronger than (CIJSE). Someone who defended (CIJSE) could say that there are non-inferentially justified beliefs where the content of the belief does not constitute evidence. The proposition believed is not part of the subject's evidence, but this isn't due to some shortcoming in the subject's epistemic position.
If we opt for this line, it seems that we have to say that it can be permissible/proper to treat something that is not evidence as if it is evidence. That's not something I like, so I'm not entirely happy with a view that incorporates (CIJSE) without (IJSE). But, the worry is that if I go in for (IJSE), I can be forced to give up the truth requirement on evidence by those who have internalist intuitions about justification ascription (or go skeptical and say that the scope of non-inferentially justified beliefs is limited to a special class of propositions about us).
Maybe I can show that internalists are playing with fire. The dialectical situation is complicated, but the embarrassment for the externalist about evidence is that it seems they have to say that there's a gap between the (i) things that we can properly/reasonably treat as evidence and (ii) the things that are evidence. Maybe internalists have to recognize this gap as well. Consider:
(JE) If S's belief that e is part of S's evidence is justified, e is part of S's evidence.
I suspect that this claim is not one that the internalists can actually defend, but then we either all need to distinguish between the evidence a subject has and that which the subject can properly treat as her evidence or question the justification ascriptions that seem to be at the root of the problem here for externalists.