I, Spy – My first design, now on Kickstarter

Hi, nswoll!
This is a complicated question. It's also difficult to address this topic without a great many specifics regarding this game or your designs, but I'll try my best ––

In I, Spy, the penalty for someone knowing your identity with 100% assurance is very high, and you will almost always be fighting an insurmountable uphill battle, unless it's very late in the game. However, they can't know with total certainty until the end of the game, and they are therefore taking an enormous risk if they 'go all in' against a certain Nation while playing a 2-player game. Of course, this possibility could occur, but probably only when very experienced players are playing against rookies, because being able to read someone's intentions goes hand-in-hand with being able to mask them.
You say –– "Or the penalty is too high leading to players playing almost randomly to avoid the other players guessing." I suppose this could happen for a short while in I, Spy, but it wouldn't be very effective, since at some point, anyone who wanted to win would have to make an effort to push their Nation to the front. But if you did so early enough to allow other players to respond, then they would likely exert their limited time and resources into devastating that nation–– but will be less capable of doing so to other nations. (In other words, being in 2nd place by the end of the game is a very powerful position.) Therefore, there's a very high incentive to fool your opponents about your allegiance, and your opponents will be very wary of committing too much. This might be hard to describe if you haven't played I, Spy, but the reality in this game is that it's fairly easy for an given player to severely hurt a given nation, but doing so will result in figuratively hemorrhaging points in all other directions, and helping the other nations, so you have to be pretty damn sure!

Another factor: One maxim of the design is that there is absolutely no action you can perform that serves only one function. For example –– Playing the "Smuggle" card generates resources for your person, and also gives 3 Influence points to France. So when someone uses this card, you can never truly be certain–– Does she really want money... or does she just want to give points to France? Even use of the Sabotage card (the mechanic which severely punishes a nation of your choice, and therefore is the incentive to keep your identity hidden) provides Influence points to Italy, so even this simple action can have a hidden motive. This makes it very hard (and dangerous) to have too strong a conviction about someone else's identity.

But also leads to an inevitable metagame element–– Everything you do in the game helps or hinders at least one Nation (and often more than one) and there is usually more than one way to accomplish any given thing. For example, the actions which allow you to move effectively are the Private Line (which helps Austria), the Overnight Line (which helps Germany) and the Courier Line (which helps Britain). Hold that thought.
Imagine that it's the middle of the game and all the countries are basically even in terms of Influence points. Now, let's say that on my turn, I travel to the German Empire and commit Sabotage, which causes Germany to lose a lot of Influence. Well, on your turn, if you want to move, no matter who you are, you have a huge incentive to use the Overnight Line card – because this way you get to move and it won't be such a big deal to give points to Germany, because they just lost a bunch. After all, if you chose the Private Line or Courier Line, you might be helping me.
So, even though the penalties for being discovered are very hash, there is a natural swing-back for any country that gets too far behind.

I hope that answers some of your question. I'm happy to elaborate about specific elements.