One of the most compelling ideas for a game is getting to play a mad scientist. Who doesn't want to design a massive experimental machine and watch it run, laughing all the while?
But there aren't very many mad science computer games. There are some where you play as a mad scientist, but they're not about doing mad science. They're about managing empires or something.
The reason is because it's hard to make mad science into compelling gameplay. It is both too open and too closed. The outcomes are excessively diverse (monsters, weather control devices, teleporters, de-aging drugs, mind control lasers) and the process of creating them is therefore overly simplistic (throw points at the laser gun until you get it made).
There are some mad science tabletop games. They revolve around allowing players to just name their own creations as they see fit in a sort of freeform "generate your own karmic death" engine.
But I want to make a computer game. About mad science. How can I do that?
When we look at mad scientists in fiction, we normally focus on the end result. But the end result isn't the mad science part - it's the karmic death part. Behind every monster, every bizarre contraption there are hundreds of other experiments. Failures and explorations on the subject. Gathering data.
I was thinking about core play in construction games, and about how there is usually a focus, a spine to which everything else is attached. For example, in Kerbal there is a core rocket column, and everything is either integrated into it or attached. That structure resists the stresses of liftoff better than any other. In Sim City, money is what matters, and you tend to have a few high-producing areas of city backed up by power plants, police stations, suburbs, and other support areas that are probably actually costing you money.
What is the focal element of mad science? What does mad science want to accomplish?
If you look at it from a super-long view, there are a massive variety of final goals. But the actual science part? It is all about generating insight.
How can we make our core experiment construction about generating insight?
How about we make the experiments about measuring variations?
Science is about repeatable results. More specifically, if you can repeat a result than you can use that as a foundation to detect variations in other aspects of the situation. If you know that electricity flows through wires and makes them hot, you can try a million different types of wire until you find a filament that allows you to create a light bulb.
When you build an experimental apparatus in my mad science game, you are not building an experiment. You are building an apparatus that supports repeat experiments - variations that attempt to nail down the precise vagaries in subtly different materials or methods. Obviously, the player doesn't weigh in on each variant. Instead, variants are tried automatically, each variant generating a little more insight into the domain, depending on how well you're monitoring it, to an asymptotic maximum.
Let's say you're just starting out as a mad scientist, so you're doing chemical analysis on common air. Your apparatus lets you burn small things in a contained volume of air, so that you can see how different things burn differently in the same volume of the same air. An inverted glass cup is all you need for this apparatus: light something on fire, then quickly put the cup over it. Wait until the smoldering stops, then lift the cup. Smell the contained atmosphere. Poke through the ashes.
This apparatus can be used extensively - it's quite a durable and straightforward system. However, the amount of insight you'll get is pretty low. Even if you run hundreds of samples, all you're really doing is smelling some air and looking as some ash. To get better insight, you'd need better measurement tools. You might create an attached apparatus for analyzing the ash samples (for alkalinity or something), or you might retrofit the cup to give you more controlled smell access, or you might put a thermometer in there to check exothermic levels, or you might add in a camera to look for tint differences in gaseous emanations...
Now when your scientists run variations, there's more meat to be had and they can generate a lot more insight into the nature of combustion. As you can see, the experiment can be built rather freeform out of just a few basic elements, sometimes mixed into one object, sometimes not:
1) Devices which perform a specific process on varied inputs (sometimes requiring human assistance)
2) Devices which separate and contain the outputs (often requiring human assistance)
3) And the monitoring tools that plug into those devices (often requiring humans to note down the values)
4) And, lastly, support devices to generate whatever generic resources (electricity, heat) is required
These can, in turn, be piped in lots of complex and even recursive ways. For example, you're researching magnets. You might expose a magnet to a high-powered electrical arc, then analyze its field strength, then zap it again, then analyze it again... or you might analyze water, boil it, analyze it again, boil it again...
Now the question is: who cares?
Generating insight is a laudable goal in reality, but in the game world, what does it get you?
Well, obviously, it gets you MAD SCIENCE. Once you generate enough insight into various realms of science, you can perform "demonstrations". For example, creating a ray gun. Hypnotizing people en masse. Stopping time. These will get you into the annals of mad science, earning you respect and new resources to use in your experiments. The better your demonstrations, the more of those limited resources you get.
This may seem like a dull goal, but it should be enough. After all, the goal in Kerbal is to land on planets. There's literally no reward for doing so, except that you can plant a flag. The existence of various science journals posting "wanted" ads for various demonstrations is enough to get you to fill them.
There is a question as to how you build demonstrations. To be honest, I think just slapping down a lump sum of insight would be fine. The mechanics of insight need to be carefully worked out: an experiment generates insight, but rather than adding to your insight total, it simply raises your total if it exceeds your total. So you want to have the best experiments you can, rather than running crap experiments a lot. Similarly, having insight on hand lets you use more complex monitoring devices. Spending insight on a demonstration will actually limit your ability to use those complex monitoring devices, which means you probably can't just run that last experiment again to regain the insight you lost...
This is a very simple gating system, but it should allow people to craft experiments not simply to some ideal, but to their particular situation. The potential complexity of experiments also skyrockets dramatically when you start getting into the true mad science areas - unlife, ghosts, time manipulation, weather manipulation, psychic powers... all of them have very different sorts of characteristics which require very different experimental approaches... but it's not really clearly delineated. There is no specific "ghost bottle": the same apparatus elements you might use to test for psychic powers or magic or dimensional flux might be used to look for ghosts. The various elements overlap, and therefore your experiments involve different progressions and mixed from the same base set of items.
Similarly, it's possible to make these elements vary randomly from universe to universe, if you wanted to play a multiverse version where you can travel through dimensions. But that's going a bit too far for the basic design...
The basic design involves applying processes to samples, then moving the samples around. As mentioned, the experiments are about performing the same thing in hundreds of variants - obviously, the player doesn't do it, but the player sets up the in-world scientists to do it, and hits play. Add in monitoring devices and strange magical physics, a respect system based around scientific journals, and a wide variety of scanning devices... and that sounds like a fun game!
I might try making it. There aren't any decent mad science games.