Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Yeah, quadratic voting is one of the potential solutions to this, but ultimately that removes the ‘fraud load’ by mostly moving it to another place, to number of accounts. That is arguably worse because it incentivises the malicious users to create as many accounts as there are votes so as to not be subject to quadratic voting (1 vote = 1 energy point equivalency only holding at one vote), and it’s substantially harder to deal with user account fraud than vote fraud.

The goal is to minimise the fraud load and make whatever fraud load you have applied against the strongest parts of your system. It’s like designing a castle except with incentives - it has to fail gracefully and still be functionally successful even in failure.



> ultimately that removes the ‘fraud load’ by mostly moving it to another place, to number of accounts. That is arguably worse because it incentivises the malicious users to create as many account

Arguably the common social media policy of not having your account tied to a unique identity needs to change. It permits all sorts of problematic incentives like you describe, and which lead to bot accounts and more.

Of course, your identity doesn't need to be disclosed in your profile for when anonymity is important.


Adding other factors into this system might help.

What about:

    weighting a vote by the "degree of separation" of the voter? (ie., friends count more)

     hyperbolic discounting of karma acquisition 

     limiting the number of voters per item

    new accounts can't vote until achived a network & karam status (eg., shares, by friends, are upvoted). 

    accounts which primarily upvote/downvote outside of their network are ignored

    etc.
I think this adds up to saying, "small networks of voters" are the primary type of community; and big-network effects are largely discounted. Where things are popular, they are because many different active communities appreciate them.


Yes, these are all different kinds of logic that can be executed, and none of them are ‘right’, it just depends on personal preference and situation.

We call that user-selectability ‘subjective moderation’, but it’s basically a client-side compiler for a people-based raw data feed.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: