On models:

Models are an important part of the decision-making process, but it’s also important to recognize their shortcomings. Good models will show trends in entire populations, and can help us decide that perhaps a train is a better overall solution than a bus, but they’re often not sophisticated enough to address the infinite number of factors that influence whether or not a person chooses to actually step onto a bus or train. For example, in a vacuum, a model might predict that train is a better solution than the bus. But what about train vs bus + beautiful bus shelters? Or bus with free wifi vs train without? Or bus without homeless riders vs train with? In the real world, the choice is never simply bus vs train.

As with any design problem, a good solution is one that has been pushed through a series of “what if” cycles. Models are a great starting point to solving the problem, but they certainly aren’t a complete solution.

• When U.S. air force discovered the flaw of averages
• Competing against luck

On civic design not being testable:

Mike Lydon wrote a great book called Tactical Urbanism on how civic solutions actually can (and should) be cheaply tested and iterated on prior to large scale implementation. He offers a ton of outside-the-box ideas that’ll make you rethink what it means to solve urban design problems.