Ordinary thought depends on recognizing differences. "This is because it is generally useless to do anything that has no discernable effect." Many familiar mental activities can be represented in terms of the differences between situations: PREDICTING, EXPECTING, EXPLAINING, WANTING, ESCAPING, ATTACKING, DEFENDING, ABSTRACTING.
One also can think about differences between differences (interactions). Minsky equates this to reasoning by analogy. "The ability to consider differences between differences is important because it lies at the heart of our abilities to solve new problems. This is because these 'second-order-differences' are what we use to remind ourselves of other problems we already know how to solve."
NB: The notion of detecting changes, and in particular high-order changes, is crucial. For example, change detection is central to lots of connectionist networks (like Grossberg's ART, or winner-take-all networks). The ability to detect higher-order relationships (though not necessarily higher-order differences) is the hallmark of New Connectionism -- it's what hidden units do. (From the Epilogue to the 1988 edition of Perceptrons, I doubt that Minsky would endorse this last point.)