The Generic Task Approach
Research at The Ohio State University Laboratory of Artificial Intelligence Research (OSU LAIR) in the early to mid-1980's focused on attempting to
"match technique to task" by identifying types of problem solving which
they called "generic tasks". Each generic task is a basic combination of
a knowledge structure and an inference strategy that is suited for solving certain
kinds of problems. The idea is to model expert reasoning with several
problem solvers, where each problem solver performs a generic task, and all
the problem solvers cooperate to solve the problems presented to them.
Each generic task is characterized by information about the following:
- The type of problem (the type of input and the type of output). What is the function of the generic task
- The representation of the knowledge. How should knowledge be organized and structured so to accomplish the function of the generic task?
In particular, what are the types of concepts that are involved
in the generic task?
- The inference strategy. What inference strategy can be applied to the knowledge to accomplish the function of the generic task?
Some of the tasks that were indentified included:
- Hypothesis Matching. Match a hypothesis to a situation using a hierarchical representation of evidence abstractions. Such a tool can verify that a certain hypothesis is valid (or not) in a given case.
- Classification Determine what categories or hypotheses apply to a situation given a description of the situation.
- Abductive Assembly Construct composite hypotheses in order to account fro some set of data.
- Object Synthesis by Plan Selection and Refinement Design an object using hierarchical planning.
In addition, research into deep models looked into
ways to represent the knowledge that underlies our understanding of
how devices work- Functional Representation of Devices Example use was to generate causal explanations of diagnostic conclusions
For a bit more about these tools