Sunday, February 12, 2012

All Models Are Wrong - #CCK12

I was watching a wonderful slideshare by Jurgen Appelo called Complexity versus Lean, when I came across one of those pithy quotes that pulls ideas together for you. On  slide 74 of 92, Appelo says, "All models are wrong, some are useful." I liked the quote so much I googled it and discovered that, according to, George Box, the industrial statistician, is credited with the quote ‘all models are wrong, some are useful’. I will leave it for another day to figure out what Box and Appelo mean by the quote, but I know what it means for me today.

Just yesterday, I left a comment on Joanne's post Rhizomes and Canons - Week 3 - CCK12, in which she says YES to many Connectivist principles that she's learning in CCK12, but NO to the idea that there is not "some knowledge that is settled and unassailable." She particularly wants to reserve a special place—perhaps a sacred space—for her own knowledge of God and human love. I have great sympathy for her point of view, being a believer myself, and yet I think she is confusing knowledge of the thing with the thing itself. This is not uncommon. It's what most of us do most of the time, and usually, it gets us through the day quite nicely.

But this quote—All models are wrong, some are useful—puts this habit of mind in perspective for me: all our models of God are wrong, some are useful. Likewise, our models of reality are wrong, some are useful. This, I think, is a core principle of rhizomatic/connectivist thinking. Reality does not sit still for our models. While we must create models of reality to attain some kind of sanity, the models are at best useful for a time. Eventually, all our models slip into an intolerable inconsistency with the reality they once mapped, and they cease to be useful—or worse, they become harmful and dangerous.

This is why Deleuze and Guattari insist that one of the core principles of the rhizome is cartography, or mapping. We must be in a constant process of mapping reality, especially that reality that we think we've already nailed down. Reality slips, and our models no longer map to it. We must build models and constantly rebuild our models. Models such as Connectivism, then, are wrong, though I and others find it useful for a time—perhaps a life time. I'm satisfied with that, but also vigilant. We are in most danger not when we think we know, but when we think that our knowledge is permanent, settled and unassailable. Nothing stops learning as quickly as settled and unassailable knowledge.


  1. Keith, you post leaves me thinking about the possible differences between a model and a map. If a river changes course, we redraw the maps, but the underlying model which says that water flows down hill and pools in low spots(unless otherwise forced) does not change. Isn't a model more durable than a map? - Bruce

  2. Bruce, I think you are onto a fine distinction between models and maps, but it's a distinction that I was not making in my post. A model may, indeed, be more stable than a map, but the point I was trying to make here is that neither models nor maps capture reality in an unassailable and eternal truth. Both our maps and models have a limited shelf-life, and when we try to force them to outlive their usefulness, then we usually do harm to ourselves, to others, and to reality.