Large-example-set approaches appropriate for context problems?
Problem is, user interactions don’t usually generate large numbers of examples
Large-scale techniques often have trouble with explaining “what happened”, control issues
Best for unsupervised learning, data mining
Users don’t have patience for slow learning