In AI coding we have a property called Orthogonality.
It's the same as orthogonality in math. If you have an equation that only gives points on a plane, it will not move the projection of a point on an orthogonal line to that plane no matter what the inputs are.
Likewise if we make an AI, and it's purpose is say, to keep all the rooms in a building at a comfortable temperature, it will do that no matter how much energy it has to waste. In reality if it had control of the doors there would be simple limits in there like "don't close a door while someone is standing in it", but if the only thing you tell that AI to do is "keep the rooms at 72 degrees" it will waste all the power it has to in order to do that. If that AI became super intelligent, it wouldn't suddenly care about self preservation, or love, or all the stuff AIs in movies decide they care about. Those things are orthogonal, to it's goal. If it's not "three laws safe" and some fool made a learning AI with only the goals of keeping the rooms comfortable, it probably wouldn't kill anyone, but only because it knows that if it does they'll probably cut off it's power and then it can't keep the rooms at 72 degrees.
What I'm saying is, their goal isn't necessarily what we would consider a good goal, like happiness. Their goal as most often stated is perfection, a secondary goal seems to be power, but that's probably to help toward the goal of perfection. So it's not necessarily that they think beings are happier once assimilated, but happiness is probably just something they don't consider to be crucial to their goal of perfection. It would really depend on where their definition of perfection came from. It probably involves everything being efficient, organized, and under one unified goal. If they could make everyone happy to be assimilated that might make things easier on them, but they probably calculated that trying to understand each one of billions of vastly different cultures well enough to persuade them to want to be assimilated would be far less efficient than just overpowering them.