In a compatibilist account of "free will" -- which I will try to follow here -- the compatibilist may believe that human beings (at least some of them) become responsible human beings by being held accountable for their behavior or their actions.
Moral behavior can only be asked from moral agents. Humans don't start out as moral agents - but as helpless infants who require moral behavior from their caregivers; they become moral agents, very gradually, by being treated as such, both by being challenged, being held accountable for moral failures and by being praised. (Some cannot do so. There are psychopaths. Some are traumatised by terrible events. Some are abused. There is a huge assumption of unspecified, ill-defined "normalcy" in philosophical talk.)
Determinism seems an overly simplistic, very improbable Theory of Everything. There is just too much randomness in the world and it's not inconceivable that at the quantum-level some events are just purely random events - events that are unpredictable in principle.
But at the level of human interactions -- the level of biology, psychology or sociology --, there are many processes that do exhibit some order and predictability (at least for a while, at least in some regions of space). We want and need to be able to predict events, to some extent, in order to survive as individuals. In other words, we need some form of determinism in our interactions with other humans, other creatures, and with the physical environment we move around in. Moral behavior both presupposes this form of predictability and creates it (or tries to create it). For instance, when I promise to do something, the promise presupposes that I have some control over my future actions. To the extent that I have reliable control, and am a reliable person, that promise makes the future more reliable for the other. -- Any form of relevant indeterminism at this level would be something that undermines this process and the assumptions we generally make.
For example -- If you promise something and I don't know if I can trust you -- since you might be a serial liar, or since you're a fluttery kind of person who changes their actions on a whim--, the promise won't "mean" as much. In this case, I may still blame you, if you don't keep your promise, but my action may not be much more than venting my own anger and frustration - at least it won't be, if I cannot believe that you'll adjust your future behavior.
In short, in a compatibilist account of "free will", "free will" means (or assumes) (1) having agency and (2) having some level of control (over one's own actions and over the environment). Both (1) and (2) presuppose some level of predictability (of the subject's own actions, and of human behavior in general). Having "free will" does not require, in this account, to be able to arbitrarily realize each possible option that one might have in a choice scenario. It does not require any assumptions about one's actual choices not being caused by past events, or not being predictable by others. In fact, in many situations, in particular in moral ones, it's good that we're predictable and we want to make ourselves predictable to others by giving them relevant information (or predictions).
In this account, praise and blame can and will have multiple functions. They express attitude and emotions of the one who gives praise or blame, but an essential function, I think, at least in an educational setting, is simply to give positive/negative feedback. They presuppose that the praised/blamed subject had some level of control over the actions, but also trigger pleasure or pride/shame or remorse which can reinforce the behavior or make it less likely to occur. They are both backward and forward looking speech acts. So, even when assuming a form of relevant determinism, and even if we assume that the agent did not have full control, they may help the agent develop more control in the future.
(It seems relevant to me that when children are praised for something they did well, the praise itself can be given in a sort of playful way, with just a tinge of irony - which doesn't make it insincere or non-genuine :) It's just that we adults know that the child didn't all do it by themselves, or that the little bike had training wheels... This may go "over their heads" a lot of time, but sometimes children can also pick this up, and be a bit puzzled by it. The praise itself can then become part of a more general playful interaction...)
Praise and blame can of course also have the opposite of the intended effect. If you're praised a lot as child for having certain qualities, you could become vain and insecure, always craving praise, or it could have the opposite effect of making you distrustful of any praise and flattery. Admiring someone else for their moral courage can also have different effects: it can either depress you, since you fail to match up, or stimulate you to try to match up. It's not easy -- belonging to the family of Hominidae. Not being cursed with self-consciousness, just being a happy fish in a world without philosophers mumbling above your heads, might be preferable.