AHP and ExpertChoice

I propose to discuss, share ideas, etc. on using Analytical Hierarchy Process(AHP) and AHP-tools like Expert Choice or Decision Lens. I have a copy of ExpertChoice on my PC and can give a quick demo if anyone is interested. - Brian Lawler

Here is an an example of an AHP evaluation:

 There is a decision to be made - Where to go for vacation?.
 There are criteria for making that decision - Scenery and Nightlife
 There are options available - Las Vegas and Yellowstone Park
 We can compare the importance of the criteria
   Scenery is more important than Nightlife
 Each option can be compared relative to each criteria
   With regard to Scenery, Yellowstone is better than Las Vegas
   With regard to Nightlife, Yellowstone is worse than Las Vegas
 We can then logically organize the comparisons to make a decision.
   Yellowstone Park is the better option by the logic that Yellowstone Park did 
   did better at one criteria and Las Vegas did better at the other but Yellowstone Park
   did better at the criteria that is more important.

There are things about this that seem to really work well

• It forces the decision process to be logical - which can be lost in a points and scores approach.
• When the initial results aren't logical, this can be used to pursue additional criteria and options that rationalize the process
• Logical comparisons work really well with executives. They easily express preferences in a relative manner. Asking them to use a points/scores/etc. doesn't work well.

One problem is that PCs aren't good at logic but are good at number-crunching. So, the commercial tools that support this generally use a numeric approach - usually based on matrices, Eigenvalues, etc. This is convenient for large hierarchies but isn't without shortcomings. Does anyone have any experience tailoring this approach?

It seems that this technique could be pursued as a Constraint Satisfaction Problem. Does anyone have any insights into this?

With regard to complex systems, AHP seems a good approach (at least often used) for complex systems of decision criteria and options ("complex" in that lots of interactions between lots of components, simple in the small but complex in the aggregate, sensitive to small perturbations).