Difference between revisions of "User:AngryParsley/IA ideas"

From Lesswrongwiki
Jump to: navigation, search
m (Ideas for intelligence/rationality augmentation/monitoring/whatever software)
(Ideas for intelligence/rationality augmentation/monitoring/whatever software)
 
Line 12: Line 12:
 
***unfortunately this is a surprisingly small group :(
 
***unfortunately this is a surprisingly small group :(
 
**the bar is set very low. it doesn't have to be correct all the time. it doesn't even need to be correct more often than people. it just has to be correct where people are usually wrong. (a dumb chess program will make fewer blunders than the best chess player)
 
**the bar is set very low. it doesn't have to be correct all the time. it doesn't even need to be correct more often than people. it just has to be correct where people are usually wrong. (a dumb chess program will make fewer blunders than the best chess player)
 +
**pretty much anyone in this field is inexperienced, so there's probably a lot of low-hanging fruit
 
*Domains of intelligence/rationality augmentation
 
*Domains of intelligence/rationality augmentation
 
**argument mapping/visualization
 
**argument mapping/visualization

Latest revision as of 16:59, 12 February 2010

Ideas for intelligence/rationality augmentation/monitoring/whatever software

  • Problems
    • people suck at evaluating arguments rationally
    • people don't know what they want
    • people suck at sticking to long-term commitments (example: new year's resolutions)
    • people rationalize past decisions as good to avoid regret
    • lots of software in this domain uses nonstandard/incompatible data formats, vocabulary, or UIs
    • no matter how good any IA/RA software is, most people will ignore/discredit it if it contradicts their strongly-held opinions
  • Advantages
    • massive potential benefit to anyone who cares about truth
      • unfortunately this is a surprisingly small group :(
    • the bar is set very low. it doesn't have to be correct all the time. it doesn't even need to be correct more often than people. it just has to be correct where people are usually wrong. (a dumb chess program will make fewer blunders than the best chess player)
    • pretty much anyone in this field is inexperienced, so there's probably a lot of low-hanging fruit
  • Domains of intelligence/rationality augmentation
    • argument mapping/visualization
      • payoff matrices
    • decision making
      • cost-benefit analysis
    • get stuff done/anti-akrasia
      • start new routines, maintain them
    • quantified self (tangential, but probably popular among the same group of people who would use these other types of software)
    • confidence calibration
      • relatively simple
        • so it's probably a good thing to start with
      • won't keep people coming back. they'll say "oh wow I suck at estimating my confidence" then go on with their lives
      • tiny aspect of this entire field
      • stuff like this already exists, but it's spread out/hard to find/not very good/etc
    • monitor and test cognitive abilities over the day
      • with enough data points, users could then tell if some change in lifestyle helped or hurt
      • with enough users, we could identify generally beneficial lifestyle changes
  • stuff I think isn't very helpful or won't have mass-market appeal
    • learning mental math tricks

TODO

  • decide on source control
    • github
    • google code
    • gitorius
  • IA research statement doc
    • goals

Tangentially-related stuff that is already out there

Useful links