Parameter settings

[John Anderson]

Unfortunately, I don't think this is one of the cases where we have much
guidance for parameter setting.

The meaning of EGN (or EGS) is relative to the goal value which
multiplies the   probability (PG) and to the values one chooses to set
C.  Basically it is a way of scaling the differences in production
utilities.  So if, in one model, a production A has utility 19 and
another 18 and EGS = 1, one will get the same choice  behavior as in
another model where a production has utility 30 and another 27 and EGS =
3.  The only qualification to this idea is that if utilities get low
enough that they might go below 0 one has to worry about scaling
distance from 0. If all productions have utilities below 0 the goal pops
with failure.

That being said, I would guess that for a normal application with G = 20
and P = 1, and the differences being those that do occur in C, Gary is
right in the 0-1 range he intuits. I would intuit that range for EGS not
EGN if I wanted to state my intuitions precisely.

Remember that EGN is equal to EGS squared times 3.29.  EGN is the
variance and EGS is a parameter of the logistic distribution which is
proportional to standard deviation.

Excerpts from mail: 4-Feb-99 Parameter settings by Gary Jones at psychology.no 
> I'm examining the effect that different mechanisms of development have upon
> a model of adult behaviour on a developmental task. Some mechanisms are
> particularly suited to ACT-R (e.g. strategy choice can be operationalised
> using expected gain noise). So for example, as you increase EGN, the
> behaviour of the model is degraded (fitting the reaction time data of seven
> year old's on the task). Parameters such as EGN and retrieval threshold are
> open-ended though - are there any specific ranges that are normally adhered
> to? (Catastrophic changes in behaviour are not seen in the model until
> EGN=5 or 6; I expect the "standard" range for modelling adults would be
> 0-1).