Let's suppose you're having a discussion with someone about free will. They believe that you have it and you believe that you don't. Most of the time when these discussions occur, people are just talking past each other, so you ask them to define what they mean by "free will". They say that they can't define it, so you suggest that they spend some more time thinking about it and that you resume the discussion when they produce a definition. They claim that they don't think there's any way they could ever define it, so you leave the discussion there and conclude that they are being unreasonable.
But are they? Let's suppose someone asks you to define "a pile". This is clearly a useful concept, but it's hard to define. 100 grains of rice is a pile, 1 grain isn't, where do we draw the line? We could pick a number arbitrarily, but perhaps the number of items required to counts as a pile varies depending on whether we are talking about rice or sand or waterbottles with smaller items needing a greater number and larger needing a lesser number. And maybe there's other contextual factors too that you've left out? The word doesn't have a nice crisp definition, you simply learn how to use it from experience. It's kind of the same for words like red and orange where the boundaries are also extremely fuzzy.
This isn't perfectly analogous to the free will situation, but it does highlight the difficulty of definitions. Let's suppose both you and the person you have talked to work together and you manage to identify five models of reality. He argues that the first two would have free will and that the last three wouldn't. That seems all well and good, but what if there's another three models of reality that neither of you had considered and if they had thought of them, they would have considered the first one to clearly be free will and the second one to be a case where you could reasonably call it either way. That would certainly complicate things, wouldn't it? The point is that expecting a complete definition may require someone to have thought of all the possible models of reality, which is a pretty massive, arguably impossible, burden to put on someone.
Nonetheless, I think the discussion can proceed. Even if the first person can't perfectly define what they mean by free will, they should be able to produce some kind of imperfect attempt. They could define a paradigmatic case of what would count as free will for them and what wouldn't count as free will. They could provide a definition that attempts to capture most of it, even if it doesn't completely match the way that they would draw the boundary. Or they could endorse a definition purely for the purpose of this discussion. For example, I might say that I think that grains of rice would be a reasonable cutoff for a pile, even if my actual model is more complicated with 20 grains definitely being a pile and 10 mostly being a pile and 3 grains only kind of being one.