2017-11-27 concepts

# Decision hyperparameters

## + what I learned or rediscovered recently #

* Decision hyperparameters

Every so often, we face situations that force us to make big decisions. How do we ensure we make the best ones?

One way would be to think really hard about the situation. For example, if we’re buying a car, we may consider the car’s make and model, its price, other cars in the lot, and so on. We could call these “decision parameters,” or parameters that are unique to this situation.

But we could also go one step further: we could focus on the factors affecting our ability to make the decision itself. For example: How am I feeling? Am I being rushed? How knowledgeable am I about the subject? These parameters are common to all decisions we make, so we could call these “decision hyperparameters”[1].

Whenever we make a decision, we virtually always consider the decision parameters. Far less often do we consider decision hyperparameters.

Of course, they are very important:

1. Hyperparameters are adaptable. It’s hard to apply the decision parameters of buying a car to buying a house, but it’s easy to adapt the process from one to the other.

2. Hyperparameters can improve over time. Unless you’re buying a lot of cars, you don’t have many opportunities to get better at this over time. However, you can certainly improve the process of making decisions, since you’re always making decisions and seeing what worked versus what didn’t. For example, if you know you tend to make rash decisions when you’re hungry, a good hyperparameter to tweak may be eating beforehand[2].

Here is a non-exhaustive list of useful decision hyperparameters based on this article:

• Adequate time: Insufficient time spent considering a problem is a great way to make poor decisions. If we don’t have enough time, we can’t consider the consequences of our actions, who else we should talk to, other potential solutions, and pretty much everything else that’s important.

• Considered alternatives: Most problems can be solved with various solutions, some better than others. Cycling through alternative solutions ensures we don’t immediately select the first one, which may be suboptimal (“availability bias”).

• Sufficient expertise: It’s tempting to offer a solution to a problem without considering how much experience we have in the domain. I may think I have a good sense of what is causing the car to stutter, but I’m better off asking an actual car mechanic. If I feel confident, it’s more likely I’m overconfident than right.

• Balanced physiology: This is a big one - it is so easy to make poor decisions when you think you are fine but you are actually not fine. For example, if you’re tired from a poor night’s sleep, or stressed, or irritable, or hungry, or physically uncomfortable, these all can all have huge effects on your decision-making capacity, even though they shouldn’t at all! You should explicitly ask yourself, “how am I feeling?” (Or better yet, have others do this for you.)

• Social pressure: When we’re around different people, we’re often effectively a different person (take how you behave with your parents versus your best friends or partner). This means we may make decisions under social pressure (e.g. colleagues, strangers, the media) that we wouldn’t necessarily make if we were exclusively thinking rationally, so it’s worth considering just how much pressure is being applied in any situation. (This hyperparameter was suggested by `jononor` via Hacker News. Added on 2019-03-04).

• Awareness of cognitive biases: Unlike machines, humans make a lot of irrational decisions because we rely on heuristics and emotions. Being aware of where we commonly go wrong can prevent us from doing so again the future.

While these may seem self-evident, I definitely don’t consider these as often as I should. Sometimes I’ll see a clear course of action, only to later realize that I spent 2 minutes thinking about it and didn’t even consider any alternatives.

I suspect having some kind of checklist is useful. Do you have any to add here? I’m sure I’m missing many.

## + here’s something i wrote #

* The risk-preference payoff matrix

If you were offered an amazing deal - say a great job offer or once-in-a-lifetime investment opportunity - would you take it?

The answer is probably yes… with reservations.

Most importantly, how do we know it’s a great deal? If it’s such a good deal, why are we getting it? And if it turns out to be an okay deal or even a bad deal, what kind of downside exposure are we actually facing under the false pretenses of a good deal?

In fact, knowing what kind of a deal something is has significant implications for just how deep we want to get. I make a case that we can borrow heavily from financial theory to determine which risks we should take and the range of payoffs we should expect.