Problem Selection

Problem selection is possibly the most important step in determining the ultimate success or failure of the ES. There are some guidelines .

The 3 P's of Problem Selection:

People

Problem

Payoff

People (and knowledge sources)

Experts

What (who) are they?

Level of support an ES provides

Which expert should you sign up?

the best, individual ... who will be available thruout

Other useful sources

course on the topic

written documents

Problem

Thumbs down:

• if solvable by other technologies

Exception:

you want to acquire the knowledge

to rapid prototype

• red flags / beyond state of the art
• learning
• NLP
• text analysis
• highly perception based
• common sense knowledge needed
• reasoning about: time, analogies

Desiderata:

Poorly structured

Well defined

For us: Derivation/classification (vs. Formation)

classification

diagnosis

selection

right answers

relatively stable domain

Payoff

• Makes money
• Avoids costs
• Improved quality
• Need to capture experties

People - Problem

Look for:

• expert uses experience/heuristics (not flow chart style approach)
• an existing solution
• expert reasons symbolically, not numerically

Problem - Payoff

Look for:

• Leverage
• Limited success is OK

People - Payoff

Look for:

• problems where "non-expert" level performance is not acceptable

3 P's

You need: Management Involvement