Digital models that simulate the dynamics of a system are increasingly used to make predictions about the future. Although modeling has been central to decision-making under conditions of uncertainty across many industries for many years, the COVID-19 pandemic has made the role that models play in prediction and policymaking real for millions of people around the world. Despite the fact that modeling is a process through which experts use data and statistics to make sophisticated guesses, most consumers expect a model's predictions to be like crystal balls and provide perfect information about what the future will bring. Over the last decade, we have conducted a series of in-depth, longitudinal studies of digital modeling across several industries. From these studies, we share five lessons we have learned about modeling that demonstrate (1) why models are indeed not crystal balls and (2) why, despite their indeterminacy, people tend to treat them as crystal balls anyway. We discuss what each of these lessons can teach us about how to respond to the predictions made by COVID-19 models as well models of other stochastic processes and events about whose futures we wish to know today.

模拟系统动态​​的数字模型越来越多地用于对未来进行预测。尽管多年来建模一直是许多行业不确定性条件下决策的核心，但COVID-19大流行确实使模型在全球数百万人的预测和决策中发挥的作用。尽管建模是专家使用数据和统计数据进行复杂猜测的过程，但大多数消费者希望模型的预测像水晶球一样，并提供有关未来将带来的完美信息。在过去的十年中，我们对多个行业的数字建模进行了一系列深入的纵向研究。根据这些研究，我们分享了从建模中学到的五个教训，这些教训证明了（1）为什么模型确实不是水晶球，以及（2）为什么尽管不确定性，人们还是倾向于将它们视为水晶球。我们将讨论这些课程中的每一项可以教给我们的知识，即如何应对COVID-19模型以及我们今天想知道的谁的未来的其他随机过程和事件的模型所做出的预测。