I’ve written a lot about diplomacy on this site. Check out the rest of the blog and let me know what you think.

Sneakiness is not all there is to diplomacy, of course. It is the whole point of Diplomacy the game, which is why, as I noted in my review of the Paradox version that that particular game is probably not feasible with an AI opponent. It is entirely about trust and deceit and requires a human that understands the nature of obligation beyond a pure zero sum model. Diplomacy requires cooperation, Risk does not.

But you can have a good diplomatic model without necessarily making the computer a good diplomat. You can use a simple like/dislike system like that in Imperialism, for example. The Civ IV diplomatic engine is quite good because the pluses and minuses are all out in the open and obvious. Soren Johnson’s argument about cheating AI in Civ notwithstanding, transparency is probably more important to the player than intelligence. I think there is a risk in adding too many variables to a diplomatic model.

My point in this epilogue, though, is that, so far, ancient strategy games have just accepted war as the default relationship in the period which was not necessarily the case. This has been, admittedly, an issue for many strategy games. One of the beauties of RTS games is that they assume the conflict and don’t try to pretend that there are other choices.

So there are a lot more computers playing Risk than playing Diplomacy even though the latter has far simpler rules, simply because the rules focus on things that humans are really, really good at.

Decent diplomacy in computer games really requires radical innovations that we have not yet seen. There will have to be entire new vocabulary, from ways to conduct back-and-forth conversations with the AI players along multiple axes, to ways to mark up maps and draw borders. We’re a long way off as yet.

I haven’t read Eckstein’s book yet. I’d like to, but time-management dictates that I put it off until I teach a Roman history course, or WCS takes its own stab at a Rome game and I engage in some research.

Reading your retrospective outlined nicely how even the best attempts fall short.

That’s really the 64,000 dollar question, isn’t it? Diplomacy is hard to model because it requires an AI that can measure expectations of gain and loss better, especially one that can accept a certain short term loss for a possible long term gain.

You can do abstract diplomacy well; I think that Civ IV has a very good model in certain respects. But I find it remarkable that little of the hardcore player driven pushes for realism and “more history” includes any effort to really capture the dynamism of classical diplomacy, both how uncertain it could be (Rome’s reluctance to do much of anything unless forced to) and how much it covered (demands for repatriation of citizens, demanding foreign hostiles like Hannibal, regime change).