Fitting in the box

If you don’t ask the question, you won’t get the answer.

Our astronomer does not take part in surveys.  This is due mostly to frustration at those he has responded to, which simply ask the wrong questions.  “A credit-card company will ask about a choice of colors for the cards, do I want my picture on it, where I would like to be able to use it, and whether I am satisfied, highly satisfied, slightly dissatisfied, and the like,” he says.  “But there are no questions about the absurd rate of interest or the illiterate customer service department.”

He then describes an experience with a telephone answering system.  “I wanted to ask a question about a bank account.  I called their customer service number.  It was an automatic tree system, with eight possible first choices.  I could get my balance, or find out if a deposit had been made, or do many other things I didn’t want to do; but nowhere among the possibilities (I spent half an hour climbing up and down that tree) was one that fit my question.  Eventually I said I wanted to set up a new account, and manged to talk to a person.  After being transferred to two other people, I got an answer.”

The two examples point out a major problem with designing something to deal automatically with many different cases: if you don’t think of a certain possibility, you can’t handle it.  When you simplify a part of the world in order to make it easy, you may leave out something important.  In any case, you are requiring people to fit into your boxes, another case of machine learning.

So, what can we do about it?  Refusing to complete surveys is a negative action.  So is ignoring the suggestions of LinkedIn, which conflates market research, legal research and scientific research; and routinely sends our astronomer links to jobs in the medical field because he has “Doctor” in front of his name.  We are pessimistic about the chances of making a large, international credit-card company listen to any one customer (especially if they can point to many “very satisfied” answers in the survey as it is).

We can think more carefully about anything we put together ourselves.  But being all-encompassing is difficult; as our navigator says, “Whenever they make something on board ship sailor-proof, someone comes up with a more resourceful sailor.”  If we’re asking people questions, Social scientists have worked out, through sometimes painful experience, ways to do it–and ways to avoid doing it badly.  In any case, we need to allow for the unforseen, to include a box for “other,” to have an option to write a free-form response.  And to read the answers!