Friday, February 18, 2011

Amazon's A-B Testing and Government Applications

One of the benefits of scientific research is that by being empirical or experimentally driven it avoids circular and endless discussions on what works or should work. You can just check it. You test to see what works. Jeff Bezos described in a Stanford lecture how Amazon utilizes this testing in it's site live to verify improvements. The ability to test which option is best increases efficiency as it cuts down not only on inefficient options but also on the time that would have been spent discussing the pros and cons of each approach. At Amazon this approach is called A-B testing. Where one option "A" is tested against option "B" and the results are seen immediately. This data-driven approach is what makes science edge out over other disciplines, and it's one that seems sorely lacking on the social sciences, where argument rather than data still reigns king. It doesn't matter that "trickle down economics" didn't have it's intended effect if the data is analyzed, what matters is that the argument sounds right and wins out over it's opposition.

I think it's high time that this data-driven approach be brought to our public institutions like government. In Amazon's elegant A-B testing approach (possibly with the addition of a C control) would be fantastic to test approaches with real data as opposed to talking heads on congress or the media. Think about simple things such as tax laws. Imagine if instead of it being driven by mere legislators it was goal driven and experimentally tested. Many laws fail in its intended purpose because debate or rationally doesn't drive human behavior. We are emotional creatures and not driven by cool dispassionate reason. Society's glorification of reason is a offspring of the Enlightenment the period that saw the birth of the nation of the United States. The Enlightenment which held reason above the then wide authority of the church, found liberty and practicality in its look at rationality. But now we're hitting the limits of such view. We are seeing the failure of reason. We need to use experimentation more. Forget about what "should" work and see what actually does work. What's great is that this is being used by businesses now. We don't have to invent it. Amazon does it with it's A-B testing and gets real time results. Can't we do it with laws too? How much easier would it be to vote on which tax law to approve or which ballot initiative if the data was already in? It may be hard to convince someone that changing something can be beneficial without this data but far more dangerous are things that sound rationally right but experimentally fail to accomplish their set goals, like the supposed "benefit" of trickle down economics.

No comments:

Post a Comment