From: Clark Quinn (510) 768-2408

>'Instruction' is a totally different thing as 'learning'. We know
>'eveything' about instruction, but what do we know about learning and
>'accumulate knowledge'?

Um, quite a bit, actually. From the physiological, behaviorist, cognitive, and machine learning fields are useful indications.

>For a number of years modelling about fenomena's isn't always the way for
>scientists to look to thinks; but it was and is an important way to discuss
>about complexity of unknow phenomenas in relation to parameters. It is a
>part of concept-mapping

Kurt Van Lehn has made a powerful argument for the importance of modelling (e.g. his keynote at AIEd 93), and exemplified it in his work.

>I want to change that. Recent I developed a (concept) model about learning,
>as I call that. More in common: it is a model to 'reach a goal' (target).
>When a student points out (different) targets (in time) (also called
>'qualities') (Juran, MIT, 1989) my model 'calculates' the results (in time,
>hours, weeks, etc.) UNTIL the output (the 'quality' or 'level of knowledge')
>MINUS the level of the goal (target) is zero. Then the student stops
>learning: his goal is reached. (And he takes another goal.) If his
>motivition is high or/and if the study-ability is good, the target is
>reached quicker.

I had a look, but I find that the electronic metaphor is a bit limiting. I prefer symbolic representations myself, including rules. But I'm trying to build systems that help people learn, and it's not a simple accumulation model.

>In this paper there are a lot of pictures and schemes and there is an (java)
>applet to test your own hypothesis in relation to two of my parameters
>(prototype; 2/21/2000). You see output in time and accompanying 'speed of
>accumulation' or 'level of knowledge' or 'competence'. I like to discuss my
>ideas and concepts about this paper. My adress is:

I've similarly seen a hydraulic model of learning, but in all cases they look like simple feedback systems and those have been shown to not capture performance appropriately.

This, however, isn't really my field, and I suggest you pursue some of the other literature. John Anderson and his ACT model, and Paul Rosenbloom and the SOAR model are good starting points. Good luck,

-- Clark

From: Errol Thompson Lecturer in Business Information College of Business Massey University at Wellington Private Box 756 63 Wallace Street Wellington New Zealand Email:

I have retrieved and had a quick browse of your paper. I have to admit that I have not seen anyone who has endeavoured to produce a model like yours before.

Enschede, march 2000.