Contrarily to LukeProg, knowledge of the Gettier Problem improves one's epistemology because it shows that knowledge equals justified true belief is not a viable stance.
Consider two agents who are communicating with each other in an attempt to reach Aumann Agreement. These agents will certainly need precise words for the following concepts:
Reality: "Is statement A true?"
Belief: "Does agent M believe that statement A is true?" and "With what probability does Agent M believe that statement A is true?"
Map/Territory correspondence: "Does Agent M's belief that Statement A is true correspond to reality?"
Calibration: "Are Agent M's beliefs well calibrated?"
Epistemic process: "What method did agent M use to generate his posterior beliefs?" "Is that method reliable?"
Gettier problems show that you won't be able to project these five dimensions onto a single binary. Which is true but not very insightful. Moreover, I can't imagine that the ability to reach Aumann agreement will ever depend on the definition of "knowledge". Therefore, this is mostly an empty semantics discussion.
It's not very insightful if you already know that knowledge isn't a primitive construct, but justified true belief held sway for an incredibly long time and most people don't realise that there isn't a straightforward definition for knowledge until exposed to the Gettier problem.