The field of economics is desperately in need of a paradigm shift if it wants to play a role in improving the lot of humanity. Instead, mainstream economists keep trying to get away with tiny tweaks in their fundamentally flawed way of looking at the world.

Consider, for example, the concept that still underpins most mathematical models of the economy: that firms and individuals have “rational expectations,” meaning that they all base their decisions on the same unbiased, probabilistic view of the future. A person deciding to buy a chocolate bar, for example, will consider such factors as the future price of cocoa and the interest that could be earned by saving the money, and he or she will use the same forecasts as everyone else on the planet. Given the vast heterogeneity of human psychology and behavior, it’s a preposterous assumption -- one that rendered the models useless during the last financial crisis.

Criticize this state of affairs, though, and you’re likely to be accused of oversimplification. People don’t really have to be incredible calculating machines for the models to be valid, the argument goes. It’s enough if they achieve a sort of rationality over time by learning from their mistakes. Economists call this the “learning literature,” and it sounds quite impressive. It suggests that we can achieve some understanding of how the world works without delving into the complexity of interactions among millions of people with different beliefs and desires.

But hang on. On closer examination, the assumptions in the learning literature are only slightly less crazy than what came before. To get their preferred outcome, theorists generally have to pretend that people start out with an almost perfect knowledge of how the economy works -- the “laws of motion,” as economists put it -- and only need to fine-tune some minor details.

It’s like a physics teacher assuming that students already know by instinct the exact mathematical form of Einstein’s relativity theory and only need to learn the value of the speed of light, which enters that theory. It’s actually less plausible, given the overwhelmingly greater complexity of an economy.

A beautiful illustration of how people might actually learn comes from a group of European economists who ran experiments in which human subjects tried to forecast developments -- such as changes in price and output -- in an economy governed by a few simple equations. Their task was complicated by the fact that they didn’t know the equations, and their forecasts would influence the outcomes.

The results showed -- in agreement with many other studies and surveys –- that people use wildly different methods to form their forecasts. Sometimes they expect trends to continue. Sometimes they abruptly change their projections. The interaction of their differing behaviors led to more realistic economic outcomes than the benign equilibrium predicted by rational-expectations models. These outcomes included persistent oscillations in inflation and output, as well as extended periods of recession driven by several distinct groups clinging to very different expectations of the future.

The message is that relaxing the assumption of rational expectations isn’t a minor thing at all. Include realistic behavior in your models, and you get a realistically complex economy that is very hard to predict and control.

If economists want their critics to go away, they should do more work that explores where plausible assumptions might lead, even if the outcomes aren’t simple or prove unpalatable to their sensibilities. Even in the European economists’ experiments, people’s behavior was artificially restricted: Participants had to form their expectations independently, without listening to others. Include that kind of social influence and you’re likely to get lots of herding behavior -- a recipe for even more chaos.

In such chaos lies economic reality. We need more economists willing to plunge into it.

(Mark Buchanan, a theoretical physicist and the author of “Forecast: What Physics, Meteorology and the Natural Sciences Can Teach Us About Economics,” is a Bloomberg View columnist.)

To contact the writer of this article: Mark Buchanan at buchanan.mark@gmail.com.

To contact the editor responsible for this article: Mark Whitehouse at mwhitehouse1@bloomberg.net.