Winsteps
Friday, August 12, 2005
Models are Stupid
Patricia A. McLagan, 2004: "Every day, we as learning and performance practitioners meet new and often more complex challenges. No situation is exactly like the last. It would be great to have blueprints and guides – or even artificial intelligence programs to ensure we’ll be 100% successful. The reality is that our success relies on our personal ability to come up with the best strategies for the performance problems and opportunities we face. This, in turn, requires a mix of our own experience, analytical and creative thinking, and access to useful models for action."
The Rasch model isn't merely a measurement model, it's a model for action. It helps us arrange our thoughts so we can make orderly decisions. But it's also a warning system.
Tom Peters: "Measures should routinely produce Surprises (if not, discard them)."
If our analytical purpose was solely to confirm what we already know, then we would soon be replaced by robots. The better the measurement system, the more we know that the Surprises are real, not artifacts of the analysis, and also the more we are confident that the Surprises have not been lost in the decimal dust. But let's also remember to keep a sense of proportion. No model, including the Rasch model, is telling us the exact truth. In fact,
Tom Peters: "Models are incredibly Stupid (very rough approximations of reality): Make sure everyone understands that! If a system/measure gives you a stupid answer, it’s probably a stupid system/measure."
The Rasch model isn't merely a measurement model, it's a model for action. It helps us arrange our thoughts so we can make orderly decisions. But it's also a warning system.
Tom Peters: "Measures should routinely produce Surprises (if not, discard them)."
If our analytical purpose was solely to confirm what we already know, then we would soon be replaced by robots. The better the measurement system, the more we know that the Surprises are real, not artifacts of the analysis, and also the more we are confident that the Surprises have not been lost in the decimal dust. But let's also remember to keep a sense of proportion. No model, including the Rasch model, is telling us the exact truth. In fact,
Tom Peters: "Models are incredibly Stupid (very rough approximations of reality): Make sure everyone understands that! If a system/measure gives you a stupid answer, it’s probably a stupid system/measure."
