A Design Rationale for Creanimate

This is a revisionist rationalization of Creanimate, using the learning environment design framework. Creanimate wasn't developed in this framework, but I think using the framework helps explains why it succeeded as well as it did.

The framework consists of the following key questions that need to be answer when creating an initial design for a learning environment:


What mistakes do people make and why does it matter?

The main mistake is not asking questions about why things -- animals in particular -- are the way they are. This matters for several reasons. First, if you constantly ask such questions, you have more fun. Your life is full of little "that's weird, I wonder why..." puzzles.

Second, you impress other people as being smart. You seem to know more and can explain more things.

Third, occasionally, those bits of knowledge and explanation have practical use, e.g., you figure out where mice are hiding in your house and how to get rid of them, or you come up with a novel design based on something created in nature.


Why do people make these mistakes?

There are several reasons:


Why don't they learn from these mistakes?

This is a "lost opportunity" situation. There's no obvious failure event or feedback. Nothing goes wrong, in any obvious way, it's just that certain good things never happen.


What design can overcome the learning obstacles?

One design would be to engage students in conversations with people who are really good at asking these kinds of questions and coming up with possible answer. Kids are often turned on to fishing, hiking, etc. when they do it with someone who knows about lot about fish or plants or wildlife.

It's important though that these be personal casual interactions, not field trips with 2 dozen students with a quiz to follow. Furthermore, the guide has to be adept at emphasizing how questions come up all the time that she doesn't know, but she can guess at and maybe know where to find the answer. Finally, the student has to have some initial interest to get involved in the first place.

Creanimate uses a simple question -- "if you could change an animal any way you want, what would you do?" -- to trigger the initial involvement. Creanimate then engages the student in a dialog involving repeated "why?" questions and possible answers based on real animals.

These questions are organized around the general templates "what actions (flying, eating, digging, ...) can a feature (wings, claws, sharp teeth) enable?" "what other features can enable the same acton?" and "what primary goals do different actions achieve?" This exposes the student to questions, cases, and the key abstractions that tie them together.


What are the challenges with this design?

Challenge: How to let students pick any animal and any change and still have Creanimate understand what they chose and what cases are relevant to it.

Responses: Offer a menu of options that should interest as many students as possible. Also allow an open-ended choice and try for partial understanding using natural language techniques.

Challenge: The program knows all the answer and asks all the questions.

Responses: This is a problem but it's mitigated somewhat by several things. First, the program doesn't know the answer to why the student's animal is the way it is, only to why other animals are they way they are. Second, the student controls whether or not a line of questioning is followed.


What specifically will students do?

This would be filled in with a detailed description of how Creanimate works.