Fair warning: I’m going to take a less-than-scientific approach to Type III errors, so if you came here to learn how to avoid or control these types of errors in your experiment or analysis, it’s probably best to go back to Google.
That being said, I’m still going to define the types of errors:
- Type I Error: a false positive; finding a correlation when there is none
- Type II Error: a false negative; finding no correlation when there one
- Type III Error: finding a positive result to the wrong question
We are an answer-seeking species. We not only want answers, we want them now. When we sit around with friends and family at the dinner table and someone asks a question, at least one person pulls out their smart phone and searches for the answer. We no longer live in wonder; we live in answers. Where we once wandered through an open field of curiosity, we now slog through a swamp of reasons.
This isn’t wholly a bad thing, as it has given the layperson access to information that might otherwise be unavailable due to access or cost. But this monsoon of answers does cause a problem.
We stop questioning. There are a multitude of reasons we stop questioning from childhood to adulthood, such as being afraid of the word “no”, being trained to act instead of question so much, or being told it’s better to just accept the status quo. After all, people who don’t do anything about their questions are sometimes just seen as annoying or complaining. Our need to ask questions hasn’t gone away so much as our need to ask MORE questions has.
Alas, we get to Type III errors. We ask a “good enough” question and search for the answer. When we find that answer, we give up and move on, assuming we asked the exact question we needed. Sometimes we are actually given the question and told to find the answer. It’s ironic that we don’t question that question, isn’t it?
“Jack, what are the latest metrics on our new ad campaign?”
“Jill, can we afford this new proposed project?”
Some of you may immediately be saying to yourself, “We don’t ask those questions where I work; we’re enlightened.” But what about your personal life?
“What do we need in order to save for retirement?”
“Can I afford to go back to school?”
Think about the last time you asked a question. Now when is the last time you asked a second question after that? Third? Fourth? Have we gone completely off your radar?
We pick a question that makes sense and stop asking. But perhaps a better solution is to question the assumptions of that original question and ask more of them. In A More Beautiful Question, Warren Berger suggests having a question session that is filled with simply asking as many questions as possible related to the topic. You won’t need all of them, or probably most of them, but one question can and most likely will spark an idea to another question. This works best with multiple people, so if possible, include others on your question session. For those close to me who are reading this, they can attest to the number of questions I ask them in hopes of sparking more.
Think of it this way: If you only travel one road, you’ll never see what is on the other roads.
Berger suggests the Why-What If-How process. First, you face to something you don’t understand (which admitting such can be a whole article by itself). You have to take a step back and ask a Why question. This detaches you from the situation. He even makes a great suggestion that I love: ask yourself “Why is this my problem?” and if it’s not, ask “Why should this be my problem?” Next, you come to the What If question. This connects possible ideas, whether they already exist or are brand new. Eventually when you come up with a bunch of “What if…” ideas, you can hone in on one. This brings you to the How question. Here you are making a small bet on an idea. You’re putting the idea into theoretical action. Then, you can move into real action.
It is never a bad idea to let your question(s) gestate for some time without an answer. This is especially true for ‘sleeping on it’ – it is amazing what your brain does while it’s asleep, connecting ideas with new ones, breaking old connections, and thinking in a practically limitless stream of (un)consciousness.
The definition I used of the Type III error uses a specific word that I’m less-than-a-fan of: wrong. “The right answer to the wrong question.” Wrong implies a kind of evaluation, which is essentially what we are trying to decrease. Think of it as asking a better question or a different question, or just simply asking a bunch of questions! If you need a starting point, try this:
Why are you looking at the situation like this? What if you looked at it differently? How can you look at differently?
See what I did there? (Don’t answer that.)