Thinkering

the unnameable

Naming patterns of thought takes effort. We don’t name all of them. Maybe intuition/instinct is having as-yet unnamed thoughts.

This is distinct from having the unconscious part of the brain make our decisions for us.

The uncertainty hierarchy

I have come across different levels of accounting for uncertainty.

1. X will happen, plan accordingly.
2. X or Y or … Z will happen - think of contingencies.
3. X or Y or … Z will happen with different likelihoods - balance effort on the alternatives.
4. X or Y or Z will happen with different likelihood and expected payoffs (use kelly’s criterion?).

So far so good. X or Y or Z could open up other opportunities, so we can iterate but we basically end up with the same structure as in 4.

People usually go to level 4 in discussions. What I do not see in discussions is the next level - taking into account the shape of the distribution of the payoffs - and I don’t know why. For any given situation, I could not quickly tell you what distribution I would prefer. It’s not automatic.

Forecasting through sensemaking

Many non-math situations (and some math problems) involve thinking about the effects of many variables on a system with few potential answers/outcomes.

The disciplined way seems to be to look at the variables one by one and their relationships to each other and the outcome. A mental model of the system is then constructed and the outcome is predicted. This can be slow.

I think harnessing sensemaking is worth trying as a neat hack in these situations. Humans seem to have evolved a powerful sensemaking mechanism (Read: The Black Swan). Sensemaking can be roughly understood to be the ability of the brain to explain outcomes after the fact.

Instead of iterating through the variables, invert. Consider all possible outcomes/answers and ask how did we get here. The brain seems to be able to prune a lot of variables this way. I find it easier to solve problems when I know that the solution is easy to find.

Apart from the speedup due to pruning variables, this works because we often only need to consider a few outcomes reducing total computation time.

Nothing is a substitute for exhaustive search and this will likely miss extreme outcomes caused by a few variables but I think its neat nonetheless.