How should bounties be done on the site?

Question

As discussed here, it may be a good idea to have the SE-concept of "bounties" on this site. There is no (as far as I know) Q2A plug - in for that, so we may need to have a workaround for this, as we have for say. close votes, suggested edits, etc.           

My proposal 

Administrators to transfer rep from the bounty giver to the answerer.  

First of all, we need to setup a thread, called "Bounties" and link to it from the navigation bar.    

Here, anyone is able to propose any bounty for any question. Then, the bounty-giver should clearly state:  

• The question it putting a bounty on
• The kind of answer he expects    
• The amount of rep at stake 
• The amount of time he will keep the bounty open for (this can be set to "forever", too)  

As on SE, a bounty can be divided among answerers, too.       

On Q2A, an administrator has the priviledge of being able to deduct, and add reputation from or to users. As soon as the bounty is put at stake, the amount of reputation at stake will be deducted from the offerer's account by an administrator. 

Once the bounty-giver certifies that he will give a bounty to a certain answer within the time frame specified, the administrator can then award the bounty to the answerer.  

It is important that 

1. The bounty is not awarded to the bounty-giver, because then the bounty will just attract nice answers, then the bounty-giver can give a crappy answer, and lose nothing from the bounty.           
2. The bounty-giver or anyone else should not be allowed to change the time frame, since that will give no value to the bounty time frame, and people will not respect the time frame gvien.    
3. If possible, the bounty list is to be placed in a prominent place in meta, and not only the navigatuion bar.       
4. In contrary to SE, the question can still be closed if there is a bounty on it. Or else the feature can be abused to avoid question closure, as is done sometimes on SE.            

---

Any issues with the above procedure? 

Any better idea (hopefully feasible with the Q2A software, we really can't afford to have something like what SE has for this simple feature)?       

Anyone opposing the idea of a bounty in the first place?          

[BUMP] 

Answer 1

The problem with using bounties as transferrable currencies is that then it stops being a good measure of the reliability of an author--- the author that places bounties on everything seems more unreliable. It is possible to institute an "ask to answer" system where the reputation of the answerer is increased, but the reputation of the asker is not decreased. If you make the rep gain proportional to the rep of the person asking, and perhaps award points to both the asker and the responder according to the rating of the question/answer, you can encourage people to ask folks to answer questions, without harming the reliability metric aspects of reputation.

But I think that it is not really necessary, as bounties didn't really work to make the questions get an answer reliably, although sometimes it helped, it wasn't reliable, mostly because of the week time limit. Bounties for research level problems really should last indefinitely, perhaps even slowly increase with time.

You can replace bounties with real, actual, honest to goodness money.  Erdos's created this system of "money problems" in his field, and these were a huge boon for mathematics. He would place a money bounty on problems, from 100dlr to 1,000dlr, and he would pay it out of pocket if someone solved it! He wasn't particularly wealthy either, it was an actual financial cost to him. But if you published a solution, you would know that a check would come in the mail from Erdos. This was a tremendous thing for combinatorics, a lot of problems got solved this way, and Erdos was very good at placing appropriate bounties according to the difficulty. People solved it, not because they needed the money, but because it was considered a high honor to have gotten a check from Erdos.

In this case, the Erdos money problems were all research level problems, although at various levels of difficulty. A 1000dlr Erdos problem was usually very difficult, requiring a major idea, while a 100dlr Erdos problem was approachable, but still difficult. I think that this model is very effective, and this way, 30 people who care about a problem can each contribute 10dlr, and make a 300dlr Erdos problem, and then the solver gets 150dlr, and this site gets 150dlr (but only if the answer is satisfactory and complete, of course).

This is really ok, if the site is a non-profit, and the money sums involved don't become astronomical, like the millenium prizes.

Comments

Hm, if the bounties don't cost the asker, then it would be easy to place bounties everywhere.  

---

The idea of replacing bounties with real money may not be good for following two reasons -

1. If someone is not able to pay any money, their problems may not be taken seriously.

2. Someone may get his/her research problem solved by paying a good amount of money as an anonymous user, and then get it published somewhere :)

---

I disagree with the idea that an interesting problem with no money will not be taken seriously, it wasn't true of problems other than Erdos's problems. It's just a good way to give a real motivation to solving problems. It is perfectly fine for a person to ask for a solution to a problem, pay money, get an answer, and publish an extension, this is what it is for! The site is visible, and people will find it, and this is an extremely good way to get academics to participate. It is no more explotative than finding something on stackexchange, extending it, and publishing the extension, as might have happened regarding the question "What is the significance of the new theorem in QM".

Regarding the "placing bounties everywhere", it's ok, so long as it's money bounty, and it's also ok if it's not real money, and there is a standard bounty size, it's like an extra upvote which doesn't contribute to the question, but to the correct answer.

---

The bounties being real money makes sense only for the refreeing section, in the other sections, nobody would be willing to spare real money for answers.