# Jeff Vancouver's Mastermind control system

[From Bill Powers (960709.0300 MDT)]

Jeff Vancouver (960709) --

I think that a public discussion of your program would be useful to
others.

In the second frame, you have the start of a control system with the
reference signal labeled "goal (find solution)". The value assigned to
the reference signal is 1. This is fine.

The next step would be to define a perceptual input function, with a
perceptual signal labelled "Perception: solution found". If the value of
this signal is 1 there is no error and the control system has succeeded.
So the first question you have to answer is "How does this control
system perceive that the solution has been found?" I don't understand
the game of "mastermind" that's being played (you might explain the
point of the game for us dummies), but this is the place to design a
the game and reports it as being solved or not solved. When I see that
input function I should understand the point of the game. So I would
suggest that you need to define "solved" in a way that depends on
variables relating to lower level perceptions ("lower" in a general
sense, not the levels I have proposed).

So: how do you know when the solution has been found?

ยทยทยท

----------------------------------------------------------------------
Best,

Bill P.

[from Jeff Vancouver 960711.10:00 EST]

[From Bill Powers (960709.0300 MDT)]
I think that a public discussion of your program would be useful to
others.

This will be difficult without the benefit of the program, but we can let
others filter as they see fit.

In the second frame, you have the start of a control system with the
reference signal labeled "goal (find solution)". The value assigned to
the reference signal is 1. This is fine.

The next step would be to define a perceptual input function, with a
perceptual signal labelled "Perception: solution found". If the value of
this signal is 1 there is no error and the control system has succeeded.

Yes, although I just labelled it "perception" given that it changes
during the run. The brief play-by-play indicates that the idea is to get
the perception to equal 1.

So the first question you have to answer is "How does this control
system perceive that the solution has been found?"

If the simulation finds the solution, the "game" states "solution found!"
If you then press the "continue" button you will see the value in the box
change from 0 to 1 and then the perception change from 0 to 1. THe input
function just "reads" the box. You would conceive of the box as the
lower-level input functions that provide meaning to the words "solution
found." I am not interested in modelling those functions. Can I not
simply say they are a given?

I don't understand
the game of "mastermind" that's being played (you might explain the
point of the game for us dummies).

In the game of mastermind the player is presented with an array of four
squares (pegs in the real game). I have simplified it to only two
squares, but the idea is the same. Each square contains one of six
colors. The colors and their positions in the solution set are not shown
to the game player. The player's job is to figure out what the colors
are. The player has rows for placing colors. When the positions of the
colors in the a row matches the "solution," the puzzle is solved. In
the simulation, the model is the player. You are observing the model
play. You can see the solution, but the model cannot.

To play the player picks colors from the set of six and puts them in the
first row. In this version each color must be different (in one version
of the game a color can be repeated in the solution set, and thus in the
row). After a row as been set, the computer (another player in the real
game) gives you "feedback" about the colors picked. If a color matches a
color in the solution set, but it is not in the correct position (column)
the player gets a mark (a white peg in the real game). If a color
matches the position as well, the players gets a different type of mark (a
black peg). Again, depending on the version, the player is generally not
told which colors are matching and which are not. THe player must
determine, from counting the marks, the colors and positions in the
solution set. Once the feedback is given, the player tries another row,
until the solution is found or the player runs out of rows.

I have programmed the game into the computer and am using it to run
participants, which I will then match against the model. However, first I
am using it for some of those conventional research methods that I know
you all don't think much of.

BTW, I will get back to you about that fuzzy glasses analogy (I think it
is appropriate), but now work beckons (I'm looking for a new job next
year - got to get those pubs out). However, given that I want to use
this model in my job talk, it is a top priority for me. I appreciate you
taking the time to look at it.

Later

Jeff