Wednesday, November 29, 2006

Human Factors in Incident Investigation

For those of you who know me personally, you know that I injured my knee last weekend and may need surgery. One of the common areas of human factors consulting that I get called in for is to do incident investigation of these kind of events (usually a worker or a customer in a workplace or store) to find out what happened and why. In my case, the injury was the result of many aspects of human factors.

At the most basic level, I was injured because I stepped out of the way of another pedestrian walking down the sidewalk in downtown Boston. My foot landed half on a curb and half over the edge. I slipped straight down, cracking my kneecap and partially tearing my anterior cruciate ligament.

The abundance of HF causes are typical of the kinds of analysis I do in my consulting. Here are some examples:

The person I was avoiding was not looking where he was going. Why not? It was the day after Thanksgiving (the biggest shopping day of the year) and I was walking in one of the busiest shopping districts in Bostong. His attention was dominated by his current task (Christmas shopping) as well as the very salient sales signs in the store windown. This leaves very little attention for where he was going. And since he probably did not have much of an association beween walking down the sidewalk and risk, there was little drawing him towards looking at his path.

I noticed him just before he walked into me. So my time to make a response decision was very limited, leading to a quick decision. The initial schema activated was to step to the side because that is the most common response (having the lowest threshold for activation). Since I had no time to evaluate the quality of the decision or its risks, I simply stepped to the side. This did not allow time to look at the ground first. This led to the poor location of my step and my fall.

After the fall, my knee clearly hurt, but really only when I tried to bend it. There was no visible damage except a small scrape. Since I had experienced minor injuries many times before, some of which involved more pain than this event, I was certain that it was minor. My schema for this kind of injury is a little rest and a little ice and I would be fine in a day or two. This conclusion was further supported by the cost/benefit analysis. This response was easy and cheap and had a happy ending. Going to an emergency room to get it checked out was expensive, very time consuming, and could end up with a sad ending (bad news).

Then, confirmation bias set in. Even though the pain grew over the next few hours, I was able to attribute that to the fact that I kept walking on it and didn't have time to ice it. The swelling could be attributed to the same thing.

8 hours later, I went to the hospital. At some point, the cost of not going, plus the predominance of evidence that my first hypothesis was wrong, cause me to inhibit the hypothesis and consider others. Going to the hospital was both my only other existing schema and the recommendation of a reliable source (my brother the surgeon).

In this case, I was not trying to allocate blame or develop design solutions to prevent the class of incidents from occurring again. But if this had been such a case, I would probably have found no individually to be legally liable and perhaps recommended signs in busy retail neighborhoods reminding people to watch where they are going. The design of the signs would have to be salient to attract attention (with humor rather than fear), and change often to avoid habituation.