Matrix Games Forums

Some previous posts have got me thinking about the problem of AI in a complicated strategy game like EiA. The fundamental contradiction seems to be a rational AI is predictable and an irrational one might as well be random. The problem is to make something to mimic a human player, a kind of Turing-test opponent. I’m not a programmer so I don’t know how much of this is feasible but here goes with some observations and suggestions:

How about a simple semi-hidden personality for each leader of a country and a rating for each player’s foreign minister. The personality could be factors like caution/boldness, intrinsic liking/dislike for other leaders/nations (which would change with how those leaders behave) secret obsessions (like a title or piece of territory or an ideological bent the AI player prefers or finds repulsive), vacillation/resolution, how the player reacts to bulling/sweetening and the like. These could be given simple factors which are only known to the players through the rating of their foreign minister. Thus Tallyrand would report relatively accurately that the Duke of Saxony would be greatly offended by a certain maneuver but probably won’t do much about it and will get over it quickly where a lesser minister would be clueless. The likelihood of an offer being accepted or causing offense would be another example of ministerial advice. The secret obsessions would especially serve to give character to the AI players and of course, everything would change with a change of regime. Hidden information is one of the things a computer can do well which are difficult in a paper game and would add a lot to a game like this, it would be a shame to miss the opportuity to do everything a computer can to improve AiE.

First of all, I did want to mention that your analysis of the pros and cons of EiA was on the mark. My euphoria was too high initially to acknowledge any defect in EiA.

Turning to AI, I am not a programmer, but I think that giving hidden personalities is something that is already attempted. The problem is implementation thus far has proven illusive.

One alternate approach that I would suggest is to utilize the preexisting gray matter, which is built into any game with two or more players. Instead of having full AI control over uncontrolled major powers, they could be given limited AI control for housekeeping and other simple decisions. Existing major powers would attempt to influence uncontrolled powers, in part by paying money, but more significantly, by specific actions in game. For example, the French player creating the Rhine Confederation would push Prussia away from France. A good model and illustration of what I am suggestion would be something along the lines of Days of Decision II.

With enough “influence” or whatever it is called, a controlled major power would take control of an uncontrolled major power for a specific purpose. As a practical matter, this approach would probably require the assumption of a perpetual French-British war, with the uncontrolled powers swinging back and forth towards the two poles. Of course, game balancing restrictions would have to put in place.

This approach obviously has trade offs, but it does work for face to face cardboard games.

In fact, I do a lot of Game AI programming, and I have to tell you, most of the time, it really isn't that complex. Unfortunately, with the CPU-clock budget that most games have, not much time is left over for AI. It's getting better, but it's not there yet. The thing is, with a game like EiA, the developers have the opporunity to start using some of the concepts from AI that aren't relly used in games because not only is video hardware getting better at doing a lot of the processing required, but a game like EiA is NOT going to tax the system with graphics the way a game like UT2k3 will.

The thing to do, if you ask me (and no one did, I know) is to create expert systems for each of the countries. This would be analgous to your personalities that you talk about, but much moreso.

I actually spent a great deal of time a few years ago on this kind of concept. I created a combat decision making engine that would not only try to think of the best chit choice for a particular situation, but would try to predict the choice of his opponent and use that as a factor in his decision.

When I play EiA face-to-face that is almost always the majority factor in my choice for combat chit.. "what do I expect my opponent to choose?" And that is something that an expert system is uniquely qualified to do.

An AI that would make good decisions is, of course, a good thing but I think too little attention is paid to making AI more like a human in having a ‘semi-readable’ character. One of the main faults of computer games generally is they basically either present information or hide it but that information once revealed is reliable in a way that is unlike real life. To put it another way they make the players decision a calculation or a guess but seldom an estimate. It is in the difference between a calculation and an estimate that some of the most interesting and complex mental processes become involved. If you actually want to simulate command you have to have scope for players to really use intuition.