Well, the Kerbal update came out and, yeah, blew me away. It features a career mode where you actually do science to gain technical advances. It's a little opaque and oddly balanced at the moment, but it's a killer mode. The core idea builds on Kerbal's strengths: there's only a few scientific techniques (a bay full of goo, taking surface samples, etc), but they yield different scientific results used in different places, so you have to travel to different places to get more science. Kerbal's strength is the rocket travel, so it makes sense that their science would revolve around travel as well. Ship the same seven instruments to thousands of different places on different worlds.
But it really got me thinking about the nature of science as gameplay. There's a part of me that loves the apparatuses of science. The weight and grandeur of a telescope array, for example, is very compelling. The immensity of CERN's colliders is astounding. This is sort of the opposite of having seven basic instruments and carrying them to more and more exotic locations.
It's clear to me that you can think about science in games in a few different ways.
One is as a simple gating/pacing method, similar to how you would view money. Invest in science, build a more advanced murder-bot. This is the Civilization method of science, the Starcraft version. There might be a tech tree, or maybe not - the important thing is that "science" is an amorphous concept that responds to simple, abstract investment.
Then there is the idea of science as a driving force. This is science used as an excuse to get the player to do more, explore more, expand their horizons. Kerbal's new career mode uses science to push you to go new places. I might use science as an excuse to build ever more elaborate facilities. The end result might feel the same as the gating/pacing method - a certain amount of science and a tech tree - but earning science comes with a lot of complex baggage that pushes gameplay forward.
Lastly, there is the idea of science as actual research (and development). I haven't seen any games like this. I've seen a few where you do science, but they don't progress according to the science you do - the science is just standard puzzle gameplay. The concept here is that you actually perform science, which allows you to perform better science due to the advancements you make.
For example, there are a few puzzle-chemistry games. If they were to fall into this third category, you would be allowed to mix the chemicals in a far more open manner, and the chemicals you created would then become available in the game world (with the efficiency and runoff that your method uses). Everyone might discover an oxidizer for their rocket fuel, but there would be many variations and many different manufacturing processes.
Obviously, this is a nasty problem to tackle, because your science needs to be incredibly robust and permissive, allowing players to create a huge variety of results and apply them in a bunch of weird ways. If you can create a chemical that is an oxidizer for rocket fuel, you need to have the concept of "rocket fuel" as well as a robust method of modeling how combining chemicals to create rocket thrust would work. In turn, this ends up being quite limited: you're not really able to invent new concepts, you just fill in the existing concepts in your own way.
Well, if that's the case, you might as well just have science as a driving force and make manufacturing the main gameplay.
It might be possible to allow for "open" science somehow, but I can't think of a way. In short, science as actual research is a phantasm. It's not a mechanic that can be used.
But you can use science as a driving force, so that's what I recommend and will design for in the near future.