Skip links
Main content

A race between multiple problem solvers

zondag 20 januari 2008 22:23

I have long wondered about the problem that occurs when multiple centers in the mind are occupied by the same task, solving the same problem, so to speak.

The mind is capable of working in parallel. It may have more than one method to solve the same problem and these problem solvers can work on a problem simultaneously.This is a great strength of the mind, both human and certainly machine. The thing is that the problem solvers are not all equally successful, and they are not all equally fast.

These players are racing each other. However, in this case, it may not be desirable to let the fastest player win. The fastest player probably uses the simplest algorithm or is very blunt. If time would allow, the slower algorithm could yield a better result. In short, faster is not always better.

So, in stead of a simple race, we could use the following scheme:

  1. A problem is suggested to the players
  2. The players can sign in on this problem (if they recognize it to be something they can solve)
  3. Depending on external circumstances, a timeout time is set. All problem solvers have until then to solve the problem.
  4. If there is only one player and he solves the problem within the timeout time, this solution will be used and the timeout is lifted.
  5. The timeout elapses. If no player has succeeded, the problem is considered unsolvable and this may trigger other problems. If the problem is solved by one player, this player has won. If the problem is solved by multiple players, a decision criterion will have to determine which is the better player.

The decision criterion can be built by providing feedback to the players. If the actions they have suggested before were successful, they will gain points. And vice versa.

If the agent is under pressure to achieve, he will decrease his timeout time and he will accept a less thorough solution.

This algorithm allows time to play a role in concurrent problem solving, but it will also allow the slower but better problem solver to win.

my agent

« Terug