Perhaps the reason emotional reactions seem so untrustworthy is that significant details of this implicit information can be altered without the alteration being emotionally evident. That is, we can change the logic enough that it no longer applies, but do this subtly enough that the emotional reaction is preserved. Therefore, rationalists feel safest if everything is laid out explicitly, so that their emotions are less likely to give them false positives for "convincing argument".

If I can indulge my nerdy side, it reminds me of cryptographic hash functions. The idea there is that the slightest change in the source string ("dinosaur" -> "dinosaus") will result in a completely different hash, one that you couldn't even tell was related to the original. The hash function relating information -> emotion fails at this: it isn't sensitive enough to reflect small but significant changes.