Draft: Newcomb Type Problems and Hypothesis Simplicity
Feb 19, 2025
- related to the thoughts in the act’s moral goodness post
- actually this is isomorphic to the ‘being honest is simpler’
- the reason lying well is difficult is this bias towards simpler hypotheses!
- an awkward and complex hypothesis: “i’m someone who slacks off at work, but during interviews acts exactly like a hard-working person who won’t slack off”
- a simpler hypothesis: “i’m a hard working person who won’t slack off”
- of course this is possible to do
- potentially this has some hints about what sociopathic intelligence is like? that people good at lying are good handling this kind of complexity in your model of what kind of person you are
- the inductive bias towards simplicity has to have some implementation in the brain, right? i know that neural networks have this, but is do different architectures have it to different extents? should we predict that good liars have a stronger simplicity-drive-source, however it might be implemented?
- a bit of an out there hypothesis: in the vascocomputation theory tanha is how the compression-drive is implemented. so maybe good liars are happier?!
- potentially you get fucked the other way if you let too much of this kind of complexity in your life? because then you’re also less biased against a hypothesis like, ‘i will totally go to bed in a sec, just 5 more minutes of twitter’
- do good liars often procrastinate too much?