Then sense-data spoke up: “What assurance have you that your reliance on [reason] is not like your reliance on sense-data? Indeed, you used to have confidence in me. Then the reason-judge came along and gave me the lie. But were it not for the reason-judge, you would still accept me as true. So there may be, beyond the perception of reason, another judge. And if the latter revealed itself, it would give the lie to the judgments of reason, just as the reason-judge revealed itself and gave the lie to the judgments of sense. The mere fact of the nonappearance of that further perception does not prove the impossibility of its existence.”--Al-Ghazali "Deliverance from Error" translated by Richard McCarthy (12) [slightly modified--ES]
While I had noted a lengthy extract of Al-Ghazali's Deliverance in William James, who appreciated its mysticism, but mistakenly thought that Al-Ghazali was reporting factual autobiography (rather than, as Stephen Menn argues persuasively,* deploying a strategy of self-presentation taken from Galen),+ I was unprepared for Al-Ghazali's Deliverance. I found the anticipations of Descartes's self-presentations uncanny (and Menn very illuminating on this point). The passage quoted above occurs in Al-Ghazali's description of his (adopting Menn's terms) skeptical crisis--a crisis initiated by the variability of intellectual and religious traditions. The point of Al-Ghazali's argument is to point the way toward a human, (again using Menn's terms) 'cognitive faculty' superior to reason (which reminds me of Plantinga's Sensus Divinitatis in Plantinga's more Spinozistic/mystical moments). This faculty is central to the practice of prophecy and the scientific fields of medicine and astronomy (not astrology, which he ridicules).
Interestingly enough (and again Menn notes this, too), this more authoritative faculty faculty is not introduced to vindicate belief in or inferences from miracles. Rather, this faculty provides the kind of inspiration that is common to some kinds (but of course not all) of scientific discovery and mystical experience. While these days it is uncommon to see scientific discovery and mystical experience in common terms, a century ago at the height of Positivism, it was a kind of (Romantic) trope, perhaps made prestigious by some comments by Einstein, to allow that the context of discovery is in some non-trivial sense non-rational. I mention this not to defend Al-Ghazali (nor Plantinga's Sensus Divinitatis), but to be clear about the fact that if we do reject Al-Ghazali's superior cognitive faculty we may also forego resources for projects we might wish to pursue.**
One can discern three commitments in Al-Ghazali: first, from a first-personal perspective one has no right to assume that one's reason is operating faultlessly. Second, it follows from this, that reason is not self-vindicating (because it does not provide epistemic security). Third, therefore reason need not be a bedrock epistemic authority. Here I have presented the three lines as a cumulative argument, but one can also see them as three distinct claims. Al-Ghazali here clearly anticipates the thrust of Hume's argument at Treatise 22.214.171.124-2, although they work with a different conception of reason, and Hume is uninterested in exploring the possibility of an alternative, (higher) cognitive faculty. [So Al-Ghazali and Hume part ways, to some degree, over the manner in which they resolve or overcome their skeptical crises.]
I mention Hume here because, in general, whenever I encounter contemporary philosophers who purportedly respond to skeptical challenge(s), they freely permit themselves the use of reason and will charge the skeptic with inconsistency for relying on reason (or even arguments). Al-Ghazali and Hume are both sensitive to the fact that the application of reason, and the authority of canons of rationality need not to be taken for granted. To note this is not to offer a criticism to the content of much philosophy, nor a critique of reason; but it is to become aware of how much is presupposed by folk that think of themselves as self-critical.
Al-Ghazali does not point to an actual error by human reason; he just notes its possibility [for the significance of modality in Al-Ghazali, see here], and this generates (the possibility of) a regress in which no (human) authority is self-vindicating and bedrock. (Al-Ghazali blocks the regress not by fideism or appealing to tradition, but by insisting on the role of the inner light in providing security.) I could have stopped here, but Menn's piece also calls attention to Al-Ghazali's concern over the tacit corruption of intellectuals/experts by the embrace of state funding and grants (who become no better than experts for hire despite their self-understanding as disinterested servants of the truth); Menn notes that Al-Ghazali provides enough hints for the careful reader to discern how Al-Ghazali indicts the practices of his past self. Perhaps the public articulation of such self-criticism is only possible from a more secure vantage point.