In an earlier post, I had argued that there is something generally fishy about conjunctive characterizations of non-stipulative concepts, such as defining a murder as an act that is both a killing and morally wrong. Natural concepts just don't have conjunctive analyses, and so one can typically find counterexamples to conjunctive analyses simply by looking at cases where the conjuncts are coincidentally satisfied (e.g., something might be a killing but be morally wrong for a reason independent of its being a killing, such as because it is also an instance of promise-breaking, and this does not make it a murder). In a post on prosblogion today, I use this fact to refute a particular argument against Molinism.
What I want to offer here is two hypotheses about why conjunctive definitions sound so plausible to us. The first hypothesis, stated in the prosblogion post, is is that conjunctive claims often carry an implicature of relevant connection between conjuncts. If someone tells me that he went to the store and bought a pound of butter, I assume that he bought the pound of butter at that store (order matters here: "I bought a pound of butter and went to the store" carries an implicature that the pound of butter was not bought at that store; in that case, the connection is a negative one). If I am told that Fred intended to meet George and Fred did meet George, I tend to assume that Fred intentionally met George, though that does not strictly follow.
The second hypothesis is that our minds are designed for finding connections. When we read a set of statements, or see a bunch of evidence, we tacitly assume a relation between them. It is not a matter of implicature, because the phenomenon is more than just a linguistic one. I see an open garbage can, and I smell a stink. I assume that what I see and what I smell are the same thing. Often this is justified. But this habit of mentally inserting connections can be pernicious. When we see a bunch of conjoined statements, it is our natural reaction to imagine a situation where they are co-satisfied in a related way. It was the genius of Gettier to show us that in philosophically evaluating a conjunctive characterization we need to look at cases of unrelated co-satisfaction.
At the same time, it may be that some conjunctive characterizations are close to the truth—to get the truth, all we need to add is that the conditions are satisfied in relevantly related ways. It could be that knowledge will be justified true belief once we understand that it must be justified, true and believed in relevantly related ways. (I wonder if what counts as relevantly related might be contextual.) Of course this is not satisfying to the philosopher—we want to make explicit the relevant relation, but perhaps this just cannot be done.