Saturday, November 22, 2014

Is Automation Making Us Dumb?

Source article: Automation Makes Us Dumb

I rather disagree with this column. While on the surface the core claim is more or less accurate (that any new tool that is adopted by humans generally leads for us to become less skilled at completing that task without said tool), this is offset by new skills that we gain through 1.) the skills associated with using the tool to complete the task even more effectively than before and 2.) other tasks that one is able to endeavor on with the freed up time, mental capacity, and energy from utilizing the new tool to complete a once more manual practice. To focus on the skills lost is totally missing the point. It's not like people and society as a whole are substantially worse off since we have become less skilled at translating words to morse code, or reading and writing binary, or navigating a ship via sextant. Instead the new tools have made these tasks unnecessary or irrelevant allowing us to focus on other challenges.

Besides the above focusing on the glass becoming half empty, the author also focuses a common, but misguided, fear. This is the fear of what will we do if our tools suddenly stop working. While it is super common to fear the loss of control in this situation, the fact is that these tools rarely spontaneously stop working and I would argue a better solution is to actually build better procedures for what to do in this rare event. In this article, the author repeatedly harps on the concern of what happens if our flight control systems suddenly stop working, will the pilot be able to step in and save everyone?

While this is certainly a terrifying thought to consider that our pilots are slowing losing the skill to fly planes without the support of a computer system, the reality is that this is a very very rare event and the number of crashes and fatalities per passenger-mile traveled has never been lower. Instead, the automation of a once very manual process has removed hundreds of opportunities for human error making a very complex and seemingly dangerous process of launching hundreds of passengers and their luggage into the air atop a giant tank of explosive jet fuel, into a very straight forward, safe, efficient and convenient process that helps moves millions of people around the world every day. While once we were at the whim of the mental and physical state of our pilot (did he have a bit too much to drink last night? did she just have a fight with her spouse? are the pilot and copilot getting on each other's nerves? is he staying focused? does she have a headache? did he forget something? did he lose focus? did she follow all the steps?), we now have a computer program that can fly our planes without any of these issues. With each flight and each update the system becomes more robust and more effective. While we may never fully eliminate the possibility of a computer error happening, overall we are vastly safer than before.

Essentially what I am saying here boils down to a concept called expected value. In probability theory, expected value takes the value of various potential outcomes and discounts it by the probability of that outcome occurring. For instance, if you have an opportunity where there is a 20% chance you will win $10 or a 80% chance you will win $0, the expected value of this opportunity is $2 (20% * $10 + 80% * $0). This way you know that all things considered, you'd expect to make $2 off this opportunity on average if you ran it over and over again. With expected value you could directly compare this to another opportunity (say where you could win $6 with a 50% probability, or $0 with 50% probability) and decide that this new opportunity is a better deal with an expected value of $3 (50% * $6 + 50% * $0). This concept is widely used in many fields including economics, finance and statistics, but humans often have a hard time internalizing it, especially when we are dealing outcomes that have very low probabilities and very extreme values (eg winning the lotto, or dying in a plane crash).

People love playing the lottery because who doesn't want to have a billion dollars? People love to fantasize of what they would do with all that money, how they would change their lives, and how they wouldn't fall into the same trap that almost every lotto winner has fallen into and blow it all in a few years and end up right back where they started. However, there are countless articles that have been written on the mathematics at play. Basically, yes you could win this big prize, but when you weigh in the chances of this actually happening, your expected value on this lotto ticket is often mere pennies. This makes intuitive sense when you step back for a second, because governments all over this country are running lotteries as a way of fund raising, and this only works if they expect to collect more money than they pay out of average. So essentially when you play the lottery, you are agreeing to pay a voluntary tax to the government, yet millions of people do this all the time often in part because they can't fully grok their expected value of each ticket.

Another area where I often feel people muddy the concept of expected value is in the area of GMO food and the impact it is having on society. In many cases, people have are opposed to this concept of "GMO food" because it hasn't been proven to be safe. What they are really saying however is they are concerned about a very small possibility that a very terrible thing could happen (like we given everyone cancer). This gets much more nebulous of course because it is involving probabilities that we aren't really able to accurately quantify, but we have a pretty good idea that this outcome has a pretty small probability because we humans have been consuming these GMO foods for years or even decades with no clear negative outcome yet. What has happened with high frequency however is that these new GMO foods have allowed us to produce more food at lower cost than ever before, allowing us to go further to combat global hunger than ever before and helping raise the standards of living of billions people around the world. In addition, we know that many GMO foods have allowed our farmers to produce their foods more efficiently with less manual work, fewer chemical fertilizers, pesticides, and herbicides (which we do know are often extremely bad for people), resulting is lower impact on the environment, fewer acres farmed, and more natural habitat retained. We know that these new GMO crops are often better at resisting extreme weather events (like droughts, unexpected freezes, and floods) and diseases leading to more consistent food output and fewer famines. Overall, I would argue the expected value of these GMO crops is clearly much higher than what came before. However many people fixate on that long tail risk of very low probability outcomes distorting their expectation of the risk associated with these new crops.

Overall with all of these things, I think we need to focus our decisions on what is leading us to an overall higher expected value of outcome for society as a whole, while simultaneous building a safety net of procedures and policies to cover ourselves in the event of a low probability adverse outcome and push our systems to constantly drive down both the probability of those adverse outcomes and the magnitude of impact in the even they occur. In this way we can continue to move society forward and balance our fears of something very bad happening. Regardless, I don't think the claims of this author that our society would be better off by resisting automation holds water and we should still continue to strive to replace dangerous manual tasks (like driving a car) with safer and more consistent automated systems where ever possible.

No comments: