As an example of deduced intent let's envision a system where every action I commit over the course of a year can be observed and recorded for inference analysis to determine if I am evil or good. Assume that in all ways I meet the standard definition of good... I help people across the street, save puppies, etc. At the end of the year the BEST that you can conclude is that I behaved ethically (good actions), but that says nothing about MY intentions. I may in fact be evil and am only ingratiating myself within a community in order to later kill them all. There is a reason that within philosophy that morality and ethics are separate words. There is a reason that when viewing demographic data people say "I can tell you what 9 out of 10 people in the group prefer, but I can say nothing about any one individual within it."
As to the preference between inferred vs explicit intent, you're placing your own judgement of the value of inference vs explicit above that of the user which will inevitably lead to frustrations on the part of the user. In the simple case of a catalog there are two distinct intent patterns that users engage in... one is the specific intent to find a known product and the other is what I refer to as "discovery." Discovery is, as the word suggests, finding things that the user did not know existed or knew very little about.