Thread by @astralwave

@pathakaakash @johncutlefish What I’d do:

- Ask if you can break down the big decision into smaller constituent decisions.

- Separate into 3 buckets: [reversible+low cost], [reversible+high cost], [irreversible]

@pathakaakash @johncutlefish Then,

Go with gut for [reversible+low cost]. Choose the “default” that comes to your mind and just go.

For [reversible+high cost], ask, how to derisk this? Run experiments, rough modeling.

For [irreversible], declare your assumptions, model it out, and guesstimate.

@pathakaakash @johncutlefish This is assuming all decisions are maximum uncertainty and your downside is unknown.

You still run the risk of a bad outcome you don’t foresee for [reversible+low cost] decisions.

I’d still try to guesstimate potential risk here, and do some rough expected value calculations.

@pathakaakash @johncutlefish The difference is, for the irreversible and low cost decision, you don’t want to waste time generating more options.

@pathakaakash @johncutlefish Bottom line: don’t try to make a “big decision”. Try to make smaller decisions, much faster, instead.

Still, hold the desired outcome and risk appetite for the “big decision” in your head, because you don’t want to lose the plot and end up optimizing locally for minor decisions.

@pathakaakash @johncutlefish Other things worth thinking about:
- Cost of delay if you don’t make the decision
- Do you know the possible option space? Or do you need to discover options? What is the cost of discovery?
- Externalities of this decision. Who/what might this affect? Beware of path dependence.

@pathakaakash @johncutlefish And finally, since this is a real-world decision, you want to find out if there is even spare capacity to be spent on making this decision, given that the decision makers’ time is also a finite resource.

That’s why it’s smart to limit decision size and concurrent decisions.