Do You Feel Lucky? - by Kenny Collins


Here's a thought experiment. After a night of insobriety, Person A is drunk driving home from the bar, and makes it home safely. Person B has a similar experience and is also driving home drunk, but hits someone on the way home. They both arrive home. How should you (or society) deal with them?

Well, most likely they will both be prosecuted for drunk driving, but Person B will also be charged with some form of assault. Seems to make sense.

But wait - Person A and Person B both engaged in the same activities and made the same choices. Why is it that Person B is being treated more harshly than Person A?

Another question you could ask is - why is Person A being punished at all? He got home safely, and no one was hurt in the scenario.

There are plenty of analogous situations. Imagine a sports star who makes a risky and showoff-ish play in a key moment in the game. If the move pays off, he will be heralded as a hero by his team and fans, with the highlight clip going viral for the next week. If the move fails, he might be reprimanded for not playing it safe, benched by the coach, and shunned by the community.

The general case involves an actor (driver, player) who engages in some action (drunk driving, risky play) that causes a result (hit pedestrian/safe drive, score/miss). It is our job, as a society, to decide on the reaction - whether he should be given positive (cheered on), neutral (left alone), or negative feedback (jailed/benched).

There are some different "policies" we can use to treat these cases. One policy would be to penalize the actor just for engaging in the action, regardless of the result. This can be thought of as an "action-based" policy, since the treatment is dependent on what action the actor decides to do.

Another policy could be to deal with the actor depending on the result of the action. We can call this a "results-based" policy, since our reaction is based on the results of the scenario. If the result is positive, we give them positive feedback. If the player scores, we cheer. If they turn the ball over, we boo.

A consideration that matters here is if the connection between the action and the result is "lucky" or not. Luck, chance, or probability is by no means a simple concept, but we can treat it as the amount of influence that the person has on the outcome. Once I flip a coin, I have little control over whether it lands on heads - thus it is highly lucky. However, if I place a coin down, I have high control over how it lands - thus it is highly non-lucky.

With lucky events, even if the event goes my way the first time (i.e. I flip a coin and it lands heads), it is unlikely that it will go the same way if we repeat it (i.e. more flips will have the coin not land heads). In other words, it is not very reproducible. With non-lucky events, the results are very reproducible (i.e. I can place a coin down as heads many times).

Different activities can range in from being highly lucky vs highly non-lucky. Roulette is a game highly dependent on luck. Poker has lots of luck elements but also takes a lot of skill. Soccer takes a lot of skill and only has small elements of luck (e.g. temperature, wind speed, players being hurt, etc).

However, a lot of times it is very unclear if something is due to luck or not. Professional forecasters (weather, sports, politics, economics etc.) can leverage their most complex statistical regressions and machine learning models and still get it completely wrong. Average people can (seemingly) randomly predict some once-in-a-lifetime event / occurrence correctly to their own gain (buying/selling Bitcoin or GME stock at the right times).

Having good intuition is a legitimate skill, and one that is inherently not easy to articulate. Hence, someone with such intuition could be mistaken for someone who is just "very lucky", and vice versa. Take a poker player at a casino who is able to call his opponent's bluffs correctly. Is he drunk and just happening to guess correctly? Or does he have great social reads to notice his opponent's tells? Or the hedge fund managers and investors who predicted the Wall Street crash of 2008. Were they doomsayers / market bears that happened to be right? Or did they read the signs correctly, and could do it again?

Let's go back to our potential policies. "If you got it right, you deserved to have gotten it right" is a fairly robust results-based mindset. We cheer for the player when they score, regardless of how lucky (devoid of skill) the shot was, because what matters is that the points count. Since we now know that some things that appear to be due to luck actually have non-luck elements (random prediction or soothsayer?), this meritocratic policy assures that people with good intuition do get rewarded for their good results (for reproducible actions over time).

However, this would lead us to advocate for Person A, the drunk driver who didn't hit anyone, to get away without punishment. Other inebriated people might be spurred on by this policy to drive while drunk more, if they are confident enough in their abilities to operate a car. Is that desirable?

On the other hand, we can decide to punish people for actions we think are stupid, and reward them for actions we think are smart - aka an "action-based" policy. Drunk driving is idiotic and introduces unnecessary risk into a society, so we punish it regardless of whether an accident actually occurs. But then we'd have to bench the player for making his risky play. This is a viable course of action, but it does require consideration. Even if it could have lost the game, it doesn't make sense for the player would made the winning shot to be totally punished.

In reality, a mix of these policies is often the way to go. The law is "action-based" for punishing people engaging in drunk driving, and "results-based" for punishing drivers who happen to hit pedestrians (hence further punishing bad drivers). The coach might reprimand the player for showing off, utilizing an "action-based" policy, but the spectators are sure to be "results-based" and scream in delight if the goal is made.

The policy to choose depends not only on the luck involved with the action, but also the positive/negative utility of the result. Having citizens hit by cars is a serious issue, and we (as a society) would like very much to create an incentive structure to avoid that result. (It can also be argued that even if no one was actually hit, Person A might have caused an unsafe environment (e.g. people swerving to avoid him), which is not what we want to encourage.) For actions with less negative results (missing the shot is not the end of the world) and positive externalities (show-off plays do look cool), we are sometimes okay with somewhat encouraging these risks.

Risks are part of life, since there is uncertainty involved with almost any choice you make. These kinds of "policies" are interesting, because we are considering how to treat a risky decision after the fact, when all the uncertainty is gone. Our reaction to the action/result will in turn influence our (and others') actions in situations similar to those.

You can applaud people for doing (what you deem) the right action, or instead for achieving the right result. You can shun people for doing (what you deem) the wrong action, or instead for achieving the wrong result. Choosing which reaction is suitable, is of course, the hard part.

Good luck!