Algorithmic Intelligence & Market Resillience

Every day the machines are getting smarter. Better tools for collecting, normalizing, and synthesizing insights from information are emerging everywhere. A decade ago, I remember thinking that recommendation panes on ecommerce sites were novel; the sign of an advanced player in the market. Now simple recommendations are easy to make and we constantly rely on software to make much more complex ones.

The promise of these prescriptions is that every task in the world becomes slightly more efficient. Algorithmic prescriptions are designed to help us make better decisions about where to save our money, what materials to use when we build things, who to hire, how to interact with each other - the list goes on.

But even as a devout supporter of the gains that computer aided decisions will bring, it's important to look for some of the risks. Recently, I've been thinking about one in particular; risks to resilience.

Our situations come about based on a complex system of actions. When we walk into a grocery store, the price we see for a basic food is the result of many farmers looking at weather and soil data and interpreting it differently. Some farmers will vary their crop varietal, others will choose different seeds for the same crop, and still other will manage their use of chemicals and fertilizers differently. At the end of the whole process, the best interpreter of information will likely make the best decisions (and therefore the most profit) in the market. But what we see when we walk into the store is a result of all that chaos.

In this world of disparate predictions, any individual participant might stand to benefit from behaving like the best forecaster in the market. Everyone wants to pick stocks as well as Warren Buffet. And the promise of algorithmic intelligence is empowering each and everyone of us to make decisions as well as the best in our markets. The risk, however, is that everyone will be making the “optimal” decision based on imperfect information.

Unfortunately, in a world of unknown information, it’s impossible to make the “right” decision all the time. The world will throw curve-balls your way. It could be an asteroid, a new disease, or something as simple as a cold snap. Invariably, the best forecasters are wrong. And it’s those situations where we’re thankful we had other people making different decisions based on similar information.

Consider this example. Let’s say growers discover a strain of rice that yields 50x more per acre of alternatives in 99.9% of years. Planters around the world would likely swarm towards producing that variety of rice over any number of alternatives - it has an obvious advantage. Then, imagine that a disease hits that strain. Global rice production could be devastated. Depending on how many people flocked to rice production, global food production overall could be devastated. All because farmers made a very rational decision to optimize their operations.

In the real world, we very rarely see these types of situations where one choice is so much better than another. The example is hyperbole at its best. But it's indicative of a significant point: Variety adds to resilience. And optimization reduces variety.

My worry is that even as we become better at accurately forecasting the world’s future state, our attempts at optimization could decrease our resilience. We get more data about the world by connecting up our devices, monitoring our production output, and measuring everything in between. But at the end of the day, we don’t have perfect information. Which means our predictions can’t possibly be perfect. 

But if human decision makers are all entrusting operations to similar data science, we will rationally turn towards similar behavior. Not the ideal position to be in when things go wrong. In this future, will we be less resilient than before? I’m definitely not one to spend money on bomb shelters, hoard canned food, or even learn to hunt. But if there ever is some sort of disaster, I’d be glad someone had made that decision. And in that case, I'd certainly be disappointed if the algorithms made all those bomb shelters and food stores disappear.

For those of us helping guide the world toward a data driven future, these are important questions to keep in mind. How to we help direct our customers the right way and still ensure their markets benefit from the resilience they’ve had in the past? I don’t know the answer, but I do know it’s a question worth grappling with.

------------

Photo courtesy of episos.de