 As we saw last week on Risk Bites, models are incredibly powerful. Whether we're talking fashion models or climate change models, they provide us with insight into the world we live in and enable us to make predictions about things we haven't yet been able to observe or measure. But models are dangerous if they're misunderstood or misused. So before Risk Bites gets back to the riveting world of dose-response models, here are three rules of thumb to help use rather than abuse scientific models. 1. Never confuse a model with reality. This may seem like a no-brainer, but it's easy to become seduced by models and forget that to shed light on some things, models have to get other things wrong. Imagine a three-dimensional anatomically correct plastic model at the heart. As something that can be pulled apart, reconstructed and generally played around with, it offers tremendous insight into the organ. But using it as a replacement for a real heart would be a grave mistake. The wrong that makes this model powerful is that it's made of plastic, not human tissue. The statistician George Box infamously captured this idea with his phrase, all models are wrong, but some are useful. Some people worry that this undermines the credibility of models, like climate change models. But instead, it acknowledges that an effective model will always have its limitations. The trick is to know what they are and how to use them. 2. An unvalidated model is an invalid model. How do you know that you can trust a model? The best way is to have it predict something that you already know the answer to. If the prediction and reality are close, your confidence in the model to be able to predict things you don't yet know will go up. This is validation. And without it, any model is worthless. Imagine that a child living next to a power line tragically develops leukemia. It's easy to build a model that associates the power line to the disease. But until that model has successfully predicted other confirmed cases of leukemia under similar circumstances, it remains a fantasy. Until it is validated, it is invalid and potentially very dangerous if misused. 3. Models are a means to an end, not an end in themselves. Models are tools. They have a purpose, which means that the right model needs to be used for the right purpose. And using the wrong model for the wrong purpose can get messy. The bottom line is that when the limitations of a scientific model are understood, its predictive powers are validated and its purpose clarified, it can be an exceedingly powerful tool. But like all powerful tools, you want to be sure that the person using it knows what they are doing. With that brief introduction to scientific models, risk bites will be introducing simple dose response models next week as the first step towards predicting risk from exposure to a hazardous substance. Until then, stay safe.