Sorry your browser is not supported!

I have a question about artificial intelligence ethics...

Do you think it is okay to give the computer an advantage against the player in certain situations? Assume there are multiple difficulty levels...do you think it is okay to give the player advantages on the lower difficulty levels and the computer adavantages on the higher difficulty levels? The advantages I am talking about are like giving one player more points than another.

What do you think? Voice your opinions please

Sure, that sounds like a nice way of doing it.

Just remember, that if you make it so that on the harder difficulty levels, it's too hard, and on the lower difficulty levels, it's too easy, you wont be able to get the nice balance that is required.

The way that I recomend you judge the levels, is you do the lower difficulties for the begginers, and then as they get better, and when they can beat that difficulty, they select the next level which provides the same challange that the lower level did, and so on.

I take it this is for Firewall?
Hows it going with getting that published? Good i hope

Jess.

I donno, I play Battlefeild 1942. And I always thought it was annoying how on hard when you kill a bot they lose 1 point, when they kill you you lose 10! Thats not fair!

I like it best when the AI just gets harder. (which ironicly is the hardest to program)

I agree with Blazer....

When I first played DOOM (the ancient) I was remarkably naive at the manner of approaching the game and the situation it posed. As my skills developed, the opening became a *boring intro* to what was to come, when the levels got more complex. So perhaps the best choice is to make the opening levels "dumb" but not "too dumb" as to be random, then increase the "smarts" used by the enemy as things go on.

My classic example of this -- from memory -- would be Galaxian, where it starts out as basically "Space Invaders" at first until the aliens begin attacking in groups of bombing runs. Again, this is a next step in the AI given that you understand how to attack the step-by-step nature of Space Invaders, then must face coordinated attacks by wings of approaching ships.

So, it is safe to say that one must balance between "very smart" enemy and very dumb ones, with a progression of the former for the latter.

ie; start out challenging then unlease hell!
S.

I don't think the Galaxian example is actually an example of A.I. The game code more than likely states to create a group of enemies and move them in a bombing run group, rather than multiple enemies that were randomly spawned deciding that they stand a better chance of killing the player if they group together and perform a bombing run attack.

I've never programmed A.I. before, and as such I don't know how many of these suggestions are implementable to the DB programmer, but having thought about it, I've come to the following conclusions about your A.I. 'ethics' problem. Good A.I. progression of difficulty can encompass many features of the enemy playing-style, ranging from totally decision-dependent to options on the outside of actual decision-making:

Increase in number of possible options...
Such as different types of enemies being equipped with different capabilities. Not really an A.I. example but it ties in well with a later point.

Quicker decision-making...
The dumber NPCs will have a longer delaying mechanism in their decision-making code, the length of which will decrease as the difficulty of the NPC increases. E.g. in a war game, the regular infantry will take longer to make a decision because of their lesser degree of training, whereas the Special Forces infantry will have excellently short decision-making times.

Larger span of awareness...
The range in which the NPC will respond to the player, for example the range of hearing or sight that the NPC has. Again, this isn't really an aspect of A.I., but it is important when used in conjunction with the other aspects.

Probability of choosing the most effective option...
The NPC has to make a decision from a multitude of different options. Different levels of NPC intelligence will precipitate differing tendencies to choose the most effective or lesser effective responses. E.g. a soldier trying to find someone on his own is more likely to get killed... he did not choose the most effective option. The more intelligent soldier would get his colleagues to help him search for someone, or would sound an alarm.

Thanks for the responses...

I use most of that stuff in the AI already, for example...

-The computer will not use implode drones against large groups of robots unless the difficulty is above 6
-The computer will not create groups of shooters for defense unless the difficulty is above 4
-The computer will not use explode drones against bases unless the difficulty is above...um...some number, I forgot...

Druadan: I never thought about reaction time before, thanks for that tip The span of awareness does not quite work with Firewall because all of the robots have ranges in which they can see and battles usually occur only on borders, however, they could detect long range robots attacking from far back...Thanks

Jess: I haven't found a publisher yet, but am talking with a couple and am almost ready to release a beta for them to try out...

Any other comments?

You mentioned something about computer players having more points. You could make an option where players could set handicaps.

The thing I think is most annoying about FPS AI, is that on harder difficulty, the bots know exactly where you are all the time. You could be sniping in a tree, and you'll get shot before you can see the bot yourself. That's a seriously unfair advantage, and very frustrating to play with.

I agree Phaelax...that is going way to far with AI, even though it is actually easier to make AI all-knowing than it is to make it see only particular things...

Actually, my situation is an RTS game...

Aramil: There is a difficulty setting in my game, which acts as a handicap. The more difficult it is, the smarter the computer is.

Any other people here working on RTS projects? If so, do you make the resource allocation dependent on the difficulty level?