skip to main | skip to sidebar

Monday, December 3, 2007

Cognitive errors in I.T.

Cognitive errors are starting to be seriously evaluated in the medical profession where lives depend on proper analysis and effective diagnostics. I see a great need to apply these lessons in the I.T. arena as well. One common error know as Anchoring occurs when the answer that immediately comes to mind is the one that is focused on (often excluding all other possibilities). This often seems to be tied to availability bias where the likelihood of an answer being correct is judged by how easily it can be brought to mind.

Common causes of Anchoring in I.T. include:

  1. Recent exposure to a solution e.g. "We'll use the umptyfart design pattern. I just read an article that said it is an excellent solution to this problem." Or "I heard Joe was having a similar problem the yesterday and he reformatted his hard drive and reinstalled everything and it worked fine. Let's try that."
  2. An unduly strong reliance on prior history e.g. "We're having a problem with the release of the XYZ application. It must be the config file again. "
  3. Confusing correlation with causation e.g. "We pushed out the ABC web application last night and now the LMNOP service is down; The new release must have broken it." Or "After (some candidate) was elected the economy got (better/worse); Clearly they caused the turn in the economy."

While the above reasoning could increase the likelihood of an answer being correct it does not necessarily prove it to be true. Problems arise when other possibilities are not considered or evidence is not evenly evaluated. As a result of this the correct answer may not be found, critical time may be wasted, or unnecessary/harmful changes may be made. It's good to ask "If it's not this, what else could it be?" and briefly explore each possible cause mentally before deciding on a course of action. Often it can be useful if you can quickly eliminate several less likely but plausible causes before investigating a very time consuming but more likely one. Sometimes you luck out.

Be wary of any possibility that you automatically eliminate as this can be rooted in a denial bias. As an example, I recently I cleaned my wife's car just before we left the house on a date and found shortly after that I could not find my phone. It occurred to me (repeatedly) that my phone could be in the garbage bag, but I ignored this suspicion because I knew I would never throw my phone away. Thirty minutes later after prompting from my wife I was standing outside by the garbage dialing my cell phone with her cell phone; its characteristic chirp was clearly audible. In my defense I suspect that the phone fell from my shirt pocket into the bag when I bent to get garbage from the floor-board, but my wife insists that I actually threw it away.

Two other common cognitive errors that are interrelated are Cherry-picking and Confirmation bias. Cherry-picking is process of choosing only examples or data that support your position while confirmation bias is selective thinking where only items that confirm your conclusion are considered.

An interesting book on cognitive errors in medicine is "How Doctors Think" by Jerome Groopman.