Saturday, November 13, 2010

retrodict

retrodict \ret-ruh-DIKT\

Definition
verb
: to utilize present information or ideas to infer or explain (a past event or state of affairs)

Hindsight is 20-20? Not exactly. I really like MW's word of the day today. It means inferring backwards from the present state of things to explain the a past state of affairs we would otherwise have no or little information about - the origin of the universe, or the customs and beliefs of some ancient civilization.

While it's quite tempting to think of retrodiction as "predicting" the past, the reasoning employed isn't quite the same. Prediction is largely an inductive process. We see specific events happening, we develop theories about why (induction) and then project those theories into the future based on the current state of affairs and our theories. (Every time it has rained, the grass has gotten wet, therefore whenever it rains, the grass will be wet.) Retrodiction is largely abductive. It sees the present state of affairs and proposes an sufficient explanation. (The grass is wet, so it must have rained last night.) Of course, retrodiction is tethered to prediction. A theory which does not produce good predictions is unlikely to be employed in retrodiction. And thus we can "test" our retrodictions about past events.

To complete the picture, when we explain present events, we are using deduction.  It's raining now. The grass is wet. Therefore the rain made the grass wet. You'll note, of course, that there are other things which must also be true. There cannot be another source of moisture. There cannot be anything stopping the rain from reaching the grass. We make these and many other observations when we come to the simple deduction that we don't need to water the grass since it's raining.

So if science is largely a matter of induction, it is fair to say that it's mostly about prediction? I'd say yes. Even when science seeks to explain the past or present, it does so by creating theories that will predict the future. This is necessary for a theory to be testable.

No comments:

Post a Comment