[MUD-Dev] Cognitively Interesting Combat (was Better Combat)

[David Kennerly]

Paolo Piselli wrote:
> --- David Kennerly <kennerly at finegamedesign.com>
> wrote:

>> I doubt how theorems of games can deny their pixelated roots in
>> discrete mathematics, which may include game trees, tree
>> algorithms, and game mechanics.  I have no disparages for what
>> cognitive

> IMO, the goal at hand is not to come up with an algorithm for the
> purpose of solving the game, the goal is to model a human's
> cognition of the game for the purpose of evalutation and
> adjustment of the game.  It would be very hard to argue that a
> human tic-tac-toe player does a brute-force expansion of the
> entire game tree, or that a computer abstracts chess board-states
> in the same way that a professional Go player does.  Do the
> algorithms developed for Deep Blue give us any insight into how
> people think about chess, or how to teach people to play a better
> game fo chess?

> Don't get me wrong - I'm not a zealous supporter of ACT-R or
> anything (http://act-r.psy.cmu.edu/), but I do think that a
> production model can be a useful tool for thinking about how a
> human player experiences a game.

We agree more than this transaction has implied.  I welcome
cognitive models, too. I look forward to the insight they provide,
and am eager to replace my faulty intuitive models of a given set of
players' minds.

Let me try anew: Cognitive psychology is a welcome inclusion for
gameplay analysis, but it does not yet exclude modeling the
structure of a game.  Even the cognitive scientist Allen Newell used
a game tree to model how humans play chess.  Of course, as all of us
agree, humans and computers don't use the same algorithms to search
a portion of the game tree.  But understanding game structure is
still a prerequisite to understanding human strategy.  For example,
in Newell's study, players used what he called progressive
deepening, which in his own words was a modified depth-first search
on a game tree.

You know the details better than me.  As an outsider looking in at
cognitive science, I see the field, if anything, including quite a
bit from computer science.  I could not recommend exclusion of game
trees and tree algorithms--not until demonstrating the advantage of
excluding these models of gameplay.  I welcome that demonstration or
any other evolution of the science of games.

>> As a tuple, it is (1/3, 1/3, 1/3).  Randomly select one of R, P,
>> or S.  Game theorists call a mixed strategy over all elements
>> with a uniform distribution a pure random strategy.  In the case
>> of RPS, it is the solution.  Having solved it, it's no more
>> interesting than another solved game, such as tictactoe.

> I would argue that Tic-Tac-Toe is still more cognitively
> interesting than RPS because the production-model for generating a
> random move from [rock,paper,scissors] is far simpler than the
> production-model for playing tic-tac-toe.

You'll not argue that with me!  :) Of course tictactoe is more
interesting than rock paper scissors; but that hardly matters after
having solved tictactoe.

>> A solved game, for a rational player, is just a chore.  By chore,
>> I mean a procedure without interesting decisions.  By a

> I definitly agree with you there, anything that can be
> proceduralized can be reduced to one rule: execute the procedure,
> and is incredibly boring unless there is something interesting
> about the skill involved in executing the procedure.

Then we agree above, too: Solved tictactoe and solved RPS are chores.

>> What is necessary seems, as best as I have found any (always)
>> necessary trait, is NP-hardness.  That is, no polynomial-time
>> complexity algorithm exists to solve it.  According to a
>> researcher, Tetris is NP-complete.  But I am wildly generalizing.
>> Is Puzzle Fighter NP-hard?

> This guy agrees with you:
> http://www.ics.uci.edu/~eppstein/cgt/hard.html

> However, I do not.  There is a difference between the size- and
> time- complexity of an algorithm, and the complexity of the steps
> involved in executing that algorithm.  Just because I have to
> expand B^D nodes of a search tree does not mean that I'm having
> fun expanding those nodes.  Solving 3-SAT problems always seemed
> pretty boring to me because the search process is mechanical.
> Tic-tac-toe is NP-hard, isn't it?  Does that make it interesting?

No, of course not.  The comparison mirrors my own examples of
calculus and FPS.  I wondered if NP-hard is necessary for fun--not
if NP-hard is sufficient for fun.

> It doesn't take exponential complexity for a card-counting
> Blackjack AI, but people have fun playing this game.

Yes you're right; my generalization was false.  NP-hard is not
necessary for fun.

>> But before continuing, I need to know: What is a metric of
>> cognitive demand?

> I would propose the following metric, given a production-rule
> model:

>   The cognitive demand that an instance of combat places on the
>   player can be measured on several axes: first, by the number of
>   different rules that fire during the combat task; second, by the
>   rate at which rules fire; and third, by the total number of
>   rules is fired.

....

> Does this metric sound reasonable, at least for production-models?

Yes.  And the example elucidated the principle excellently.

David
_______________________________________________
MUD-Dev mailing list
MUD-Dev at kanga.nu
https://www.kanga.nu/lists/listinfo/mud-dev