Friday, September 25, 2009

Concentration

Here's a scenario. You are busy doing something. You are asked to spend a few minutes calling up a place, no more than 5 minutes, by a friend or significant other.

Do you do it?

Or do you say "No, I won't do it. I need to get what I'm doing done first, and then I'll take care of it". There are quite a few people in that second camp. Why is that? What is the big deal being interrupted?

Joel Spolsky noted this when he posted the following problem. He imagined two tasks that needed to be done, each lasting 10 minutes. He asked which is better, for a computer to do one task to completion then the other, or to interleave the two tasks. That is, give one minute to one task, then one minute to the other, and so forth.

It may sound identical to you. Both tasks are finished in 20 minutes. Or the second might seem more appealing because each task makes more progress. But consider when both tasks complete.

Start the clock at zero. If you interleave, one task will finish at minute 19, and the other at minute 20. However, if you did one task to completion first, it would complete at minute 10, while the other would complete at minute 20. The average in the first case is a wait time of 19.5 minutes while the average wait for the second is 15 minutes.

This assumes no penalty for switching tasks. In reality, computer tasks pay a small penalty to switch from one to the other. This is called "context-switching". Apparently, for humans, context switching is a very real penalty. Thus, minor interruptions may not be so minor. I find it baffling personally, especially if the time to carry out the task is very little.

The point is this. Some people are definitely inclined not to be interrupted, not even for a few minutes so they can get their task done. I wonder if that is symptomatic of a certain mindset or personality. Is a type A person likely to dislike interruptions and want to be fully engaged, no matter how short the interruption?

What is the thinking behind this behavior?

No comments: