I had an idea for a web-based app for evaluating instrumental rationality techniques, something like Digg or UserVoice-based forum where techniques get upvoted, downvoted, merged, separated and discussed. However, I don’t currently have a solution for the problem of ‘impulse upvoting’ (“hey, this technique sounds cool, let’s upvote it!”) -- I don’t know how to make the upvotes reflect long-term usefulness of the techniques.
We must also be aware of the PredictionBook problem. It isn’t integrated into LW, therefore it gets little traffic, and, as a result, is underused—as of this writing, it’s not even mentioned in this thread. Lesson to learn: if we want the power tool to work, it must be properly integrated into LW (same login, homepage / sidebar widget, LW moderation policy to redirect relevant discussions there etc.)
So, what suggestions do we have for integrating PB? I’ve been making and linking clear predictions I see on LW, but apparently that’s not enough.
I doubt making every LW account a valid PB account would do the job alone, although it’s a good start.
Perhaps there could be a LW extension which shows a random prediction in the sidebar? Or even just the X most recent predictions, similar to OB and the LW wiki.
I doubt making every LW account a valid PB account would do the job alone, although it’s a good start.
The main directions here are 1) to drive traffic to PB, and 2) to identify and remove the biggest trivial inconveniences. Having separate accounts is definitely near the top of the list.
X most recent predictions, similar to OB and the LW wiki.
Yes, this is exactly what I had in mind—a sidebar thingy (is ‘extension’ the official term?) that serves three key functions:
I would be willing to follow a PB blog if it drew attention to things that are “good” to make predictions about, with testimonials from users. Don’t just encourage people to make predictions, write about why that’s a useful thing to do and the nuts and bolts of it.
I’ve been making private predictions about important work-related events, because I’d like to become better calibrated in this domain. At the moment, I suck at it.
Hm. I would like to be able to write about useful things to do with PB, but Matthew tells me I’m 1 of 4 (or less) regular users, so there aren’t many people to write such posts.
Further, my main goals with PB is to become more calibrated in general—which is almost impossible to give good examples for by its very generality—and as a record of my predictions and politics for the future which I can examine to see how I got wrong (entirely aside from how much I was over/underconfidence) which I also can’t give any good examples for within the next few years/decades.
So I don’t really know what I could do with your suggestions. (I do have a little project involving >100 anime-related predictions, but I don’t know how well that will work out.)
What’s wrong with the obvious solution of having the software downgrade or separately track instant upvotes, and quiz users at occasional later intervals on what techniques actually worked for them over time, and weight those results more highly. Or display separate short/medium/long term ratings so users can decide whether they want to read shiny things or work on long-term helpful things?
I had an idea for a web-based app for evaluating instrumental rationality techniques, something like Digg or UserVoice-based forum where techniques get upvoted, downvoted, merged, separated and discussed. However, I don’t currently have a solution for the problem of ‘impulse upvoting’ (“hey, this technique sounds cool, let’s upvote it!”) -- I don’t know how to make the upvotes reflect long-term usefulness of the techniques.
We must also be aware of the PredictionBook problem. It isn’t integrated into LW, therefore it gets little traffic, and, as a result, is underused—as of this writing, it’s not even mentioned in this thread. Lesson to learn: if we want the power tool to work, it must be properly integrated into LW (same login, homepage / sidebar widget, LW moderation policy to redirect relevant discussions there etc.)
So, what suggestions do we have for integrating PB? I’ve been making and linking clear predictions I see on LW, but apparently that’s not enough.
I doubt making every LW account a valid PB account would do the job alone, although it’s a good start.
Perhaps there could be a LW extension which shows a random prediction in the sidebar? Or even just the X most recent predictions, similar to OB and the LW wiki.
Make it a prediction market and let us bet karma. :)
Please don’t.
Why not? Karma is a more concrete bragging point than a flatter-looking graph. If status-seeking will force people to calibrate...
Same reasons I brought up here a few months ago when this last was mentioned.
Hm, that looks like the Precautionary Principle to me...
The main directions here are 1) to drive traffic to PB, and 2) to identify and remove the biggest trivial inconveniences. Having separate accounts is definitely near the top of the list.
Yes, this is exactly what I had in mind—a sidebar thingy (is ‘extension’ the official term?) that serves three key functions:
Gives new LW visitors an idea of what PB is.
Reminds existing LW users that PB exists.
Drives traffic to PB.
I would be willing to follow a PB blog if it drew attention to things that are “good” to make predictions about, with testimonials from users. Don’t just encourage people to make predictions, write about why that’s a useful thing to do and the nuts and bolts of it.
I’ve been making private predictions about important work-related events, because I’d like to become better calibrated in this domain. At the moment, I suck at it.
Hm. I would like to be able to write about useful things to do with PB, but Matthew tells me I’m 1 of 4 (or less) regular users, so there aren’t many people to write such posts.
Further, my main goals with PB is to become more calibrated in general—which is almost impossible to give good examples for by its very generality—and as a record of my predictions and politics for the future which I can examine to see how I got wrong (entirely aside from how much I was over/underconfidence) which I also can’t give any good examples for within the next few years/decades.
So I don’t really know what I could do with your suggestions. (I do have a little project involving >100 anime-related predictions, but I don’t know how well that will work out.)
What’s wrong with the obvious solution of having the software downgrade or separately track instant upvotes, and quiz users at occasional later intervals on what techniques actually worked for them over time, and weight those results more highly. Or display separate short/medium/long term ratings so users can decide whether they want to read shiny things or work on long-term helpful things?
My only prediction book prediction is about less wrong.