|PHASE I||PHASE II||PHASE III|
|To explore alternative||To explore both direct||To explore the four|
|Representational Devices||and indirect ways to||methods of composing|
|(RDs), e.g, Rules, Tables.||decompose problems.||plans.|
|1. Recognize RD implicit||1. Consider several||1. Note composition|
|in specification.||commonsense plans.||methods in generated|
|2. Select alternative||2. Identify intermediate||2. Attempt to compose|
|from stock of RDs.||steps in commonsense||plans with other|
|3. Rewrite specification||3. Include intermediate||3. Note changes needed|
|to use each RD.||steps in decomposition.||to use other methods.|
|4. Evaluate advantages and||4. Evaluate advantages||4. Evaluate advantages|
|disadvantages of each RD.||and disadvantages of||and disadvantages of|
|FIG. 6.7. Summary of three heuristics for generating variability.|
structs!) normally covered in an introductory programming course and teach variability, all in a single semester. The next question to ask is: How can we assess whether students are learning something and whether that learning transfers? It is extremely problematic to grade exams and homework where the student's task is to show variability in their designs: What are the criteria? And transfer, well that one is even more problematical.
The cognitive science approach to education is a heady enterprise; researchers and practitioners feel that they are truly breaking new ground and they can make a difference in education. Thus, these worries should be treated as new mountains to climb and should not hold back our efforts at exploring what really counts as expertise and what ways there might be for teaching it.
This work was sponsored by the National Science Foundation under grants MDR-8751361 and IST-8505019.