Critical thinking toolkit
Questions I often find myself running through when thinking or in conversation.
causation != correlation
claim: x => y
details: how do we know x=> y and not y => x ? what are the relevant heuristics and datapoints? how similar are x, y? (visualize as covariance or consume similarity) anyway - are they orthogonal, or do they code for largely the same traits? is there some “hidden” property z that is common to both? example: religious people are less depressed
Limit testing ♾️
edge cases are often the most interesting and worthwhile in understanding things across a variety of contexts (related ideas: taking the limit, finding extrema, my high school ap physics teacher termed this notion vB’s theory of extremes, introducing it as a wedge to reason correctly about physics by employing this thought experimental apparatus. e.g - conservation of momentum. what happens when two balls collide? imagine one ball is massive - the size of the moon. or imagine one is the size of a ping ping ball.
analysis: this works now but what if we changed some relevant parameter theta?
example: i want to get a weighted blanket right now (april in easy coast, Toronto ), but are they amenable to summer? (4 weeks away)
Modeling incentives 🐘
stated incentives = “what you claim to want”
- revealed incentives = “what your actions say about what you want”
- concealed incentives = what do you have to gain from this that you are i) unwilling ii) or unable to plainly state? Elephant in the brian; these underlying incentives are frequently not verbalized, and often subconscious even to the agent themselves (read Hanson and Simler’s book for a thorough coverage of the literature). A lot of this draws on evolutionary psychology.
- Example: what makes someone religious? Ostensibly, it’s their faith in God (brain). But it’s also the community that religion provides, the psychological security of having a cohesive framework for understanding the world, and the costs of defecting from one’s social circles (elephant.
This list is by no means final.