Simple or Simplistic

“Everything should be made as simple as possible, but no simpler.”

— attributed to Albert Einstein

“When does simple become simplistic” is a question I’ve been musing about. It comes up related to mission research quite frequently, in various forms. Here’s three ideas I’ve run across that seem to bear on it.

1. “The map is not the territory.” A 1:1 map is nothing more than the territory itself. A map is useful because it reduces a larger thing to a manageable size. It is useful because it eliminates everything that gets in the way of accomplishing a specific task. When you want to know where the nearby gas station is, you don’t need to know a count of the number of pine trees in the area.
2. A story I often remember from my days at the World Christian Encyclopedia: David Barrett told me, once, that people expected mission research to be like a high resolution black and white photograph—clear, detailed, crisp. They thought of mission research as a place where daily, even hourly reports were flowing in from the field, summaries of who had been evangelized and how it had gone, etc. But of course, mission research isn’t like that at all. He said it’s more like a Monet painting—fuzzy, but instantly recognizable. Little dabs of color provide just enough detail.
3. Heuristics, or rules of thumb, work most of the time. They are very simple and usually memorable. “Correlation is not causation,” is one a lot of us know. However, while “rules of thumb” work on “common sense” cases, the outliers can really trip you up. They can be simple and easy in some cases, and overly simplistic in others. Correlation is not causation, but it can be a strong clue, because causals are correlated.

Something simple (like a map) can be very useful. Something complex (like an AI LLM) can also be useful. Something complicated has too many details—it’s too difficult to use it, to ask it the question you want to pose. something simplistic eliminates too much detail—it gets rid of all the complications but also the important nuances. Because it doesn’t have enough detail, you’re in danger of getting the wrong answer and making the wrong choice.

• Imagine asking Google Maps where the nearest gas station is. It tells you, but doesn’t tell you that just a little bit further away (‘no longer nearest’) is a gas station that is 10% cheaper.
• Imagine asking an AI LLM why the war in Sudan is occurring, and having it respond simply, “because the RSF & the SAF don’t like each other.”

“Simple” and “complex” are neutral descriptors. “Complicated” and “simplistic” are judgements of usability. How do you keep something from becoming simplistic? Making the ‘it’s too simplistic’ judgment depends on the thing being judged, its situation, the question being asked of it—there’s no way for me to say when ‘simple’ becomes ‘simplistic’ outside of the context. But one general (possibly simplistic) way I know to try and figure out whether something is too complicated or too simplistic is to be very clear on what what question you are trying to solve. Then, go out and get examples of the kind of question—ask focus groups, friends to brainstorm, etc. Get lots of random examples of the question, some of which you wouldn’t have thought of—those are outliers to test the system. Then, apply those questions to the model/data set/app/research project/thing—do you get back the right response? do you ever get back a response that surprises you, but in the end appears right?