Unsong is really good, especially if you like ratfic and/or kaballah. Aella said it was her favorite work of fiction

Would Lesswrong comments count? https://www.lesswrong.com/posts/t2NN6JwMFaqANuLqH/the-strangest-thing-an-ai-could-tell-you?commentId=JDEcBaCsn6NjpNWDJ

Aside: This market structure does not incentivize recommendations very well; need to contribute Ṁ25 of initial liquidity to suggest something. I can make some of that back by betting on my own suggestion, but then I fare even worse if you don't like it. It's a net loss unless I can bet against someone on my suggestion and take their money. (Maybe I get trader bonuses on my suggestions? But that rewards controversial suggestions more than good ones.)

@jcb Fair enough. How's a bounty market? Or if I add in your suggestion if you post it as a comment?

@firstuserhere Point is: i'm happy to structure in whatever way is the best for traders as long as I get some of his best writings as recommendations, because he's written a lot apparently

@firstuserhere I think bounties might be more effective for things like this. Though that loses the ability to leverage markets to surface the best options.

Brainstorming other approaches:

• Promise to refund the initial liquidity investment for any answers added. Retains predictive market power, answerers can profit by betting on themselves, you're paying for suggestions. But you're not directly incentivizing good suggestions, someone could submit a poor answer and then bet against it. (Perhaps a discretionary exception for troll or obviously bad answers could keep this at bay)

• Promise a reward for any answers you like (bounty style, just without market support, you have to jump through the hoop of sending mana directly). More incentive for people to make sure answers are good, but more risk in adding an answer, so people might be reluctant to add things they're not confident in.

I did go ahead and add my top recommendation already. I'm just pondering how this could be structured best.

@jcb Yeah the question of structuring is important. I could managram everyone who submits a good option their initial investment + k times initial investment where k ranges [1,5] and thats my rating of how much I enjoyed reading that suggestion.

@firstuserhere Also, bounties require a fair amount of mana tied up from the start. Previously I've tried bounties with 10k distributed for book recommendations for example, and it worked well. This is just another experiment to see if multiple choice would work better but it probably doesn't.

@firstuserhere it seems to me that a hybrid multiple choice + bounty market type could work well for this use case. Or even some basic UI support for sending managrams for answers submitted could go a long way.

I think that having markets to quantify the strength of the suggestions adds a lot over the bounty market type, it just needs some refinement to work really well

@jcb Great! I think I'm going to do the following payout via managram

I send a managram to every good recommendation of amount = 25 + k*25 where k is 0 to 5, depending on how I rate it out of 5.

So, if I absolutely dislike a recommendation, you still get back the initial 25 mana, but only after I've read it. If I like it, then on top of the initial investment made, you earn trading profits + multiples of 25 as mana cash.

To limit spam of lots of recommendations which will be overwhelming, I'll only do this for good faith submissions (which is fairly easy to tell, because manifold is small enough to know this).

It's almost unfathomable to me that you've never read anything from him

@SemioticRivalry That's what everyone who has ever recommended Scott Alexander's writings to me has also said.

Thankfully, now I've read Sort by controversial and the counter is set at 1 instead of 0.