The referenced work by Andrew Critch on hedonic awareness is not yet published science. It’s his private work that he developed at the University of Berkeley for a course on psychology for mathematicians; brought to the Center for Applied Rationality; and then developed into a CFAR workshop unit.
Dang it. I was trying looking him up until I saw the Author’s Note.
For anyone who has taken the workshop or is otherwise familiar with his work (or similar work), could you provide a summary? I’m sure it’s more complicated than portrayed here, but is keeping a bag of chocolates with you and rewarding yourself like you were training a pidgeon a decent start? I’d love to try it out.
The other obvious question is if/when this work is going to get published in journals? This is exactly the sort of work that if can if presented well can give CFAR a reputation for real science (which among other things helps nicely with grants and the like). Moreover, this is precisely the sort of thing that should be well known if it is accurate, and if it isn’t is the sort of thing that careful peer review will likely find holes in.
The most immediately useful thing I learned from Critch is that the human mind is sophisticated enough that it can give itself chocolates without chocolate. If you get good at noticing your thoughts you can give yourself reinforcement entirely in the confines of your brains, eg by thinking “smiley face!” or something whenever you notice yourself thinking about getting some exercise.
Unless I missed something, the little I had to read about Critch’s unpublished work on hedonic awareness seemed to be a rephrasing of Skinner’s Operant Conditioning/Reinforcement theory?
As for the use of imagined positive reinforcer, that seems very similar to covert positive reinforcement (part of covert conditioning) which should be easy to find scientific tests on if you have access to libraries.
The only difference here is that the behavior itself is not imagined. I’m inclined to believe that the situations are similar enough that the tests on covert positive reinforcement could be applied. The perception of the behavior itself being real may have some effect on our perception of the imagined reinforcer, but there’s not enough reason to believe it would majorly change the effect of the imagined reinforcer on average.
The reply was about how drethelin’s situation where a real situation/behaviour is repeatedly associated with imagined reward, is very similar to covert positive reinforcement where one imagines even the situation/behaviour itself. I’m confused on the relevance of mentioning the original comparison between actual/imagined reward in the context?
We have a situation where there are scientific/empirical tests performed on ‘a real behaviour with real positive reinforcement’ and ‘an imagined behaviour with imagined positive reinforcement’ that seems to support each other.
In fact covert conditioning does have the requirement that the patient imagine the situation sufficiently vividly. There’s no reason to believe that if the patient imagine (or perceive) the situation too vividly (or too real) it would somehow affect them less.
As for the use of imagined positive reinforcer, that seems very similar to covert positive reinforcement (part of covert conditioning) which should be easy to find scientific tests on if you have access to libraries.
The only difference here is that the behavior itself is not imagined.
Whoops, I misread that last line as “The only difference here is that the reward itself is not imagined.” Thanks for catching that.
Dang it. I was trying looking him up until I saw the Author’s Note.
For anyone who has taken the workshop or is otherwise familiar with his work (or similar work), could you provide a summary? I’m sure it’s more complicated than portrayed here, but is keeping a bag of chocolates with you and rewarding yourself like you were training a pidgeon a decent start? I’d love to try it out.
The other obvious question is if/when this work is going to get published in journals? This is exactly the sort of work that if can if presented well can give CFAR a reputation for real science (which among other things helps nicely with grants and the like). Moreover, this is precisely the sort of thing that should be well known if it is accurate, and if it isn’t is the sort of thing that careful peer review will likely find holes in.
The website for the course in question (as well as Critch’s contact information) can be found here.
The most immediately useful thing I learned from Critch is that the human mind is sophisticated enough that it can give itself chocolates without chocolate. If you get good at noticing your thoughts you can give yourself reinforcement entirely in the confines of your brains, eg by thinking “smiley face!” or something whenever you notice yourself thinking about getting some exercise.
Wow. If that works thats genuinely an incredibly powerful technique.
Has there been any empirical testing comparing that to controls or to external rewards?
Unless I missed something, the little I had to read about Critch’s unpublished work on hedonic awareness seemed to be a rephrasing of Skinner’s Operant Conditioning/Reinforcement theory?
As for the use of imagined positive reinforcer, that seems very similar to covert positive reinforcement (part of covert conditioning) which should be easy to find scientific tests on if you have access to libraries.
The only difference here is that the behavior itself is not imagined. I’m inclined to believe that the situations are similar enough that the tests on covert positive reinforcement could be applied. The perception of the behavior itself being real may have some effect on our perception of the imagined reinforcer, but there’s not enough reason to believe it would majorly change the effect of the imagined reinforcer on average.
The only difference between actual rewards and imaginary rewards is that the former isn’t imaginary?
The reply was about how drethelin’s situation where a real situation/behaviour is repeatedly associated with imagined reward, is very similar to covert positive reinforcement where one imagines even the situation/behaviour itself. I’m confused on the relevance of mentioning the original comparison between actual/imagined reward in the context?
We have a situation where there are scientific/empirical tests performed on ‘a real behaviour with real positive reinforcement’ and ‘an imagined behaviour with imagined positive reinforcement’ that seems to support each other.
In fact covert conditioning does have the requirement that the patient imagine the situation sufficiently vividly. There’s no reason to believe that if the patient imagine (or perceive) the situation too vividly (or too real) it would somehow affect them less.
Whoops, I misread that last line as “The only difference here is that the reward itself is not imagined.” Thanks for catching that.
Yw and thanks for the clarification. No more confusion then. :)
I only have anecdata, though Critch might know. It’s the kind of thing that I imagine is hard to run tests on due to problems with compliance
“Psychology for mathematicians” sound to me like the coolest thing ever to be thought at an university.
I’ve wondered before whether Eliezer would use science from after 1992 that Harry couldn’t have read. Now I need wonder no longer.
Brings a whole new meaning to the phrase “timeless physics”.
(Yes, I’m aware that this joke would have been funnier if it was a physics paper)
Actually timeless physics is being treated as Timeless Science in HPMOR—nobody in 1991 should’ve heard of Julian Barbour yet.
Timeless physics predate Barbour.
Ah! I knew there was a funny joke in there somewhere.
While the actual physics has Time-Turners.
Twist, Harry’s Dark side is actually an embedded copy of a modern day wikipedia?
Test: Watch Harry closely to see if he starts randomly deleting parts of himself!
Inner Slytherin: Quirrel can’t be David Monroe, because I asked him about the class of 1943 and he didn’t know the teachers’ names!
Inner Ravenclaw: [No original research]
Inner Hufflepuff: [Request for Speedy Deletion]
Oh please, like a Ravenclaw would be offended by original research.