7
\$\begingroup\$

I've been researching the Belief-Desire-Intent model and the way they used it in the original Black and White Game(which is the same, but they also defined decision trees as opinions).

I have a question about the distinction between opinion and belief.

From what I understand, beliefs are linked lists in the form of attribute: value.

So let's say a tree would be something like: Condition: Inanimate Type: Plant

So let's say I try to apply the command cut on it and the npc cuts down the tree in like 2 turns, by doing 50 damage each turn. So the newly formed belief is that the hp of a that tree is more than 50.

From what I understand, that this is now a newly formed belief, so the new belief about the tree would be: Condition: Inanimate Health: >50 Type: Plant

What about the fact that the npc now knows it takes 2 turns to cut down the tree, is that now a newly formed belief or an opinion? If it's not, what kind of opinion could be formed out of that?

Help from anyone who's dealt with this would be appreciated as I'm finding it hard to find open source code examples of this.

\$\endgroup\$
  • \$\begingroup\$ So from what I've gathered since, opinions are formed based on experiences with objects that would satisfy a desire. So let's say your desire is hunger, you're looking for objects to satisfy that desire, you've eaten a rock, a tree(inanimate) and a cow(animate). Eating the rock and tree only increased your hunger, while eating the cow decreased it. The common belief you have about them is whether they are animate or inanimate objects. So now you form the opinion, that animate objects satisfy your hunger more.That's as far as I've gotten so far, anyone who knows more feel free to comment! \$\endgroup\$ – itsmetheperson Nov 3 '15 at 1:32
  • \$\begingroup\$ Shouldn't this question be asked at Philosophy Stack Exchange? \$\endgroup\$ – The Light Spark Nov 3 '15 at 11:45
  • 1
    \$\begingroup\$ @TheLightSpark it's a software model based on a philosophical concept, which was then adapted for game artificial intelligence \$\endgroup\$ – itsmetheperson Nov 3 '15 at 14:52
2
\$\begingroup\$

It would seem that you are missing some beliefs from the model.

There's two reconciliations I see:

NPC Does 50 Damage is an Existing Belief

While one belief, that the tree has >50 HP is true. It's based on an existing belief that the npc does 50 damage. So, you may be missing the npc's beliefs of its own abilities.

Then there is a second belief that is missed, and that's that the tree has <=100 HP. If the npc believes it does 50 damage, and has to use two turns to destroy the tree, then the new belief is that the tree has to have less HP than or equal to 100.

NPC Only knows how many turns to cut down a tree

The other possibility is that the npc will not know how much HP the tree has, but rather how long it takes for them to cut down the tree. Then their beliefs about many things will be relative to turns, and not relative to the underlying HP stats. It can only build beliefs based on what it observes.

In which case the belief would be described as taking 2 turns to cut down the tree.


I'm not very familiar with the actual system, but from a logic stand point that seems to be the issue. You can't believe the tree has any specific number of HP, unless you have some way of knowing how much damage you do.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.