As an ex-physicist this is confusing, just because it seems like a profound challenge to Samuelson, Afriat, and the rest of revealed preferences, but pulls back at the last minute to preserve GARP by rationalizing these changes back into a preferences-oriented utility function.