Wednesday, April 19, 2017

Does Big Data Make Centralized Control the Better Option?

I grew up in the waning years of the Cold War. I remember doing a project in High School on the possibility and repercussions of nuclear war. I got to see the fall of the Berlin Wall and many other events that went along with the crumbling of the world's other superpower as their system of government, communism, proved to be inferior to the combination of democracy and free market capitalism.

One of the standard arguments I recall for why communism failed was that it was inefficient and wasteful, particularly when it came to the allocation of resources. A standard example was that local farmers knew better when to plant and harvest than the central government, but that in the USSR they had to do those things when the federal government told them to. By contrast, in the US, farmers would make decisions based on their experience to try to maximize their yields. They did that because that was how they made money. A year of bad yields would be an economic hardship in the best case and possibly lead to losing the farm.

A thought that struck me recently is that this particular argument against communism might not apply anymore. The reason is that we have moved into the era of "big data". In fact, I wonder if centralized control might not have the advantage these days, assuming that centralized control does things in an optimal way.

To illustrate this, consider the farming example I just gave. The local farmers have a lot of experience they can fall back on, but these days I'm guessing that even better decisions for allocating resources could be made using a data set that included crop yields across an entire nation for many years, combined with detailed weather data during that same time and other potentially relevant information.

Consider this article about a company that uses predictive AI to place orders in advance to improve shipping efficiency and reduce the number of returns. It is just one of many examples of how computer systems can now make decisions much better than humans are capable of because they can pull in much larger quantities of data.

Of course, the key here isn't centralized control, and I'm sure that many people would argue that centralized control still fails because it lacks the motivation to do well that individuals have. In that regard, this isn't a call to switch to communism. Instead, the key here is the data, and this is an area where government can play a role. Even better than one big centralized decision making process is a system where everyone has access to all the relevant data, and they can all try out various ways to process it to make optimal decisions.

In that regard, I think that government could play a role by making data available and helping different groups to make their data accessible and consistently formatted so that it can be more broadly used. This doesn't just apply to crops with data on weather, planting dates, harvesting dates, and yields by location, this could also be useful for a lot of data related to health including air and water quality and potentially consumption habits. I don't have a full mental image of what exactly this looks like in a broad sense, and I can clearly see challenges related to privacy issues. Still, I feel that we need to push for making more data generally available so that individuals and companies can utilize it to make better decisions. Regardless of where the control comes from, the way forward for efficient decision making is clearly availability of useful data.