2019-09-05

The Robot Vacuum Cleaner And The Universal Basic Income

I've blogged about the UBI before. I like to call it the "Basic Excise Guarantee" (BEG), because it is an idea that is almost certain to result in massive new taxation for any society that attempts it.

Bryan Caplan has a good, and short, blog post at EconLog about people who advocate for the UBI. Here's his closing paragraph:
If I were an enthusiastic UBI advocate, I would know this experimental evidence forwards and backwards. Almost all of the advocates I’ve encountered, in contrast, have little interest in numbers or past experience. What excites them is the “One Ring to Rule Them All” logic of the idea: “We get rid of everything else, and replace it with an elegant, gift-wrapped UBI.” For a policy salesman, this evasive approach makes sense: Slogans sell; numbers and history don’t. For a policy analyst, however, this evasive approach is negligence itself. If you scrutinize your policy ideas less cautiously than you read Amazon reviews for your next television, something is very wrong.
I read this, and it got me thinking about Eufy, the robot vacuum cleaner I bought my wife for Mother's Day. She had always wanted one, and I found one for an attractive price, so I bought it. We like it.

Through Eufy, I discovered something important about using robot vacuum cleaners. It's counter-intuitive before you buy one, but in hindsight it is totally obvious. In order to make good use of a robot vacuum cleaner, you need to consciously remind yourself that your floor is being vacuumed by a mobile algorithm, and not by a human being.

How does a human being vacuum a floor? I'm a human being with some vacuuming experience, so I'll tell you how I do it. I start at one end of the room, and thoroughly vacuum the floor by covering every square inch repeatedly, from one side of the room to the other; then I move on to the next room.

How does a robot vacuum a floor? The robot starts at any random point on the floor and moves in one direction until it encounters an obstacle. When it reaches the obstacle, it deploys one of a series of evasive maneuvers. Those maneuvers appear to be:

  1. Turn in a drastically different direction and continue straight until it encounters another obstacle;
  2. Treat the obstacle as a "corner object," and attempt micro-turns until it can find the way around the obstacle;
  3. Treat the obstacle as a "wall," and attempt a 90-degree turn;
  4. Treat the obstacle as a "lump in the carpet" or other insignificant setback, and reattempt the same path to see if continuation is possible.
From the standpoint of a human being, this approach is utter lunacy, because we can see the entire room and already know exactly how to solve the problem. But from the standpoint of a robot, this approach is perfect. The robot has managed to reasonably account for 90% or more of all possible encounters with obstacles, and has figured out a way to process the obstacle without the need for advanced image-processing or computation. Or eyes.

The result of all this is a situation in which a human could vacuum the floor in five minutes, while it might take the robot twenty minutes. Some customers might be inclined to think, "If it takes longer, then what's the point?" But if that's what you're thinking, then you haven't absorbed the economic lessons of comparative advantage. Remember, when a robot vacuums the floor, you don't have to. It might have taken you five minutes to vacuum the floor, but then you'd only have 55 minutes left in the hour to do anything else. If you deploy a robot to vacuum the floor, then you get those five minutes back and use them for literally anything other than vacuuming the floor. That's an efficiency gain.

And if you have more than just one room to vacuum, the robot ends up being really great. I turn the robot on early on Saturday morning, when I'm making breakfast for my daughter and I. I don't have to spend my weekend vacuuming the floor, and my daughter gets to have waffles; everybody wins. But in order to capture this efficiency gain, I have to consciously ignore how I, personally, would vacuum the floor and just let the robot do its thing. It takes more vacuuming time, but it's not my time that's being used for vacuuming, so who cares?

The Universal Basic Income seems to be particularly popular among Silicon Valley tech-types, and it's easy to see why. Rather than sinking lots of time and money into a means-tested welfare system with high administration costs, wouldn't it be better to deploy a simple algorithm, like a "negative income tax," to address society's poverty automatically? We'd reduce administrative costs all the way down to $0, and gain economic efficiency by replacing a complicated system of price distortions with a cash stipend, no strings attached. Sure, we'd lose some efficiency by failing to give the severely needy more money than the just-kind-of-needy, or even than the not-needy-at-all (it's a Universal Basic Income, remember); but means-testing costs time and money, which "we'd" save by out-sourcing our decision-making process to the algorithm.

So, the UBI (the BEG) starts to look a lot like a robot vacuum cleaner for poverty. It's not as efficient as a direct cash transfer to someone in the greatest need, but it a reliable-enough algorithm to do most of the necessary work without having to think too much about it all.

The problem is that when we conduct UBI experiments, the algorithm fails. Rather than modify the algorithm, though, BEG advocates just double-down. Robot vacuum cleaners work because the algorithms were rigorously tested to meet acceptable thresholds; and even then, buying one is a free choice. The UBI doesn't enjoy the same benefits.

No comments:

Post a Comment