Robert Nozick holds that people want real experience and not the subjective one.
I used to think he is entirely mistaken. It just sounds bad. It depends on how you phrase the question + most people cannot realistically understand the question.
Now I have two more points.
1) In line with the idea that many of our intuitions are rational as general decisions rules (Gigerenzer "ecological rationality" and "simple heuristics that make us smart").
Possibly we have a strong intuition against false deals and fake stuff. The whole experience machine sounds like cheating. If taken to its logical conclusion, it is a good deal when taken seriously, but most people are following their thinking rules, yielding a strong no. Would they be able to truly comprehend it they would have liked it.
2) Goals are another fancy issue. Do we want to feel good? or do we want status, sex and food?
It sounds contradictory, but human motivation need not be defined in feeling good. Humans seem to have in fact multiple motivations, and not all of them are always compatible, nor can we say that people know how to trade off between different goals.
If the evolutionary embedded goals are specific, we can understand why people refuse the experience machine. We have in mind the embedded goal of food not of its experience.
Does it imply that it is not experience that matters? surely not. Because it is quite possible that in a sense people prefer the good experience to the real stuff.
In a sense the want to feel good has better rational justifications, and is more coherent when looking at life, than when taking the various goals themselves seriously.
[PS. goals as means: Sometimes, these goals bring the good life indirectly, as when being immersed in an "important" task makes a person happy. He must truly believe in the essentially valuableness of the goal in order to have the experience.
If goals are means for the good life, than they will tell us to refuse the experience machine, but it will be misleading. Since they are means and the experience of the machine is the very end.]
Some will say that it is not good, since we need a way to determine what people want. How will we determine preferences, wants and goals, if people are so unreliable?
The question is mis-defined. Because it is very possible that we do not have a way to ascertain what people want. Maybe people themselves do not know.
This is also not contradictory to the idea that we should use certain rules for practical purposes, because even without knowing "really" and "for sure", we may take the rules that seem best for us.
But I believe that if we want to know what people want we got to look at the variety of people actions AND related thoughts. Analyzing actions and preferences in light of what people think and know when they decide and act. This is very subtle and unreliable. Smart skeptics will say impossible. But we have a glimpse.