Spotlight: Combining Deductive and Inductive Reasoning

Deductive and Inductive reasoning are complementary reasoning processes.

Deductive Reasoning 

Deductive Reasoning refers to the process of concluding that something must be true because it is a special case of a general rule that is known to be true. For example:

• When it is day on one side of the Earth, it is night on the opposite side. (the general rule)
  
• It is daytime in Toronto. (a fact)
  
• It is nighttime in Beijing. (a fact)
  
• Therefore Toronto and Beijing must be on opposite sides of the Earth. (a deduction, and a specific instance of the general rule)

Deductive reasoning produces knowledge that can be logically proven. That is, its conclusions will be absolutely true if its general rules are perfectly sound and its facts are true.

Deductive reasoning is logically sound, but it is seldom feasible from a time-and-cost perspective to apply to diagnostics. Fault isolation from first principles (deduction) requires time consuming and costly representation of the first principles on which it may reason.

 Inductive Reasoning

Inductive reasoning, in contrast, uses specific examples to create a general rule. Each example supports the rule, but the rule cannot be proven absolutely. For example, no-one has ever seen a palm tree growing natively in Toronto. Using induction, we can create a rule that "Palm trees are not native to Toronto". This rule serves us well but cannot be proven absolutely - because one day, someone just might find a palm tree growing natively in Toronto, and the rule will be invalidated.

• Inductive reasoning may not be provable, but it offers practical efficiency advantages, and is often seen in bad troubleshooting habits. A mechanic who reasons that "any time the GEN FAIL light was on, it was the generator, so I'm going to start by changing the generator" (induction) will burn quickly through your inventory of generators.
  Therefore, neither deductive reasoning nor inductive reasoning, in isolation, is ideal for troubleshooting. Thankfully, SpotLight takes a pragmatic and rational approach to combining inductive and deductive reasoning.  
  
• Inductive Reasoning is used in the process of creating the knowledgebase. Each SpotLight "Solution" is a general representation of a failure mode based on the best knowledge available. A Solution can be based on real failures that have been encountered in the field or it can be based on a theoretically possible failure mode anticipated by engineering analysis or modeling predictions. 
  

Deductive Reasoning is used during a SpotLight troubleshooting "Session" to determine if the problem at hand is a specific instance of any of the general rules ("Solutions") in the knowledgebase. A Session will assist the technician in gathering the symptoms and test results (facts) necessary to conclude that the problem on this equipment is a specific instance of a Solution in the knowledgebase.

The GEN FAIL example

Suppose that our total worldwide experience to date shows that thousands of incidents where the GEN FAIL light turns on were caused by only thirty different failure modes. Let's further suppose that design engineers have determined that there are 1000 potential failure modes that could (in theory) turn on the GEN FAIL light - 30 of these have happened; 970 have not.

The SpotLight knowledgebase could hold the 30 known failure modes, and chances are very high that it would meet the users' needs. After all, no-one has ever encountered those other 970 faults before - so apparently they are not as likely to happen, for some reason.

In reality, you'll create a knowledgebase with some number between 30 and 1000, depending on how much operating experience you have with the equipment. The more you have, the more you will rely on field experience for your knowledgebase because those are the weaknesses in your equipment that are revealing themselves through real operations. The coverage level to apply will be a pragmatic decision trading off cost and coverage level, and will depend on your business and technical drivers.

Summary

SpotLight uses Induction to build a knowledgebase, and Deduction to find the ones that match the symptoms of a problem during troubleshooting - a very pragmatic, rational, and effective approach that overcomes the short-sightedness of the purely inductive approach and the cost/complexity of the purely deductive approach.