Tell HN: An idea on karma
3 comments
Karma is a tricky thing. I've been playing around with a toy version of HN for a couple of days now to see if I can somehow find a better way to use the 'votes' in order to maximize the user experience on the site.
Karma is only one aspect of that.
Every time a user votes a submission (or a comment) up or down there is a complex relationship between the two users.
It is based on any combination of the following factors:
- quality
- like / dislike
- agreement / disagreement
- mood of the voting user
To mine all those from a single click it is necessary to track those clicks over many interactions. Only that way does it become possible to get a sense of what is on the users mind when he/she clicks that 'vote' link.
Votes on urls that are submitted are not subject to the 'agreement / disagreement' element, but all the others apply.
Allowing people to 'climb' faster based on click patterns is a very complex data mining issue. The 'dummy buttons' idea is an interesting one, to some extent this already happens on HN (the up/down vote buttons do not always work the way you think they do).
One hard thing to achieve is to get rid of the effect simply measuring karma has on HN. Because the users see the points and the karma they collect there is a feedback loop in HN that tends to get out of hand.
Comments posted early on in a thread will get some karma regardless of how good they are in the longer term.
Ideally a 'good' user (so one that in the long term will acquire access to all the features) should be identified early, but then that would open the door to gaming the system even further. The road up the hill is pretty quick anyway for most users, as far as I can see you get all the goodies from 200 points upwards (other than administrator privileges), an active user could collect that in a week, and by then will be reasonably well versed in HN etiquette to not abuse these powers.
Karma is only one aspect of that.
Every time a user votes a submission (or a comment) up or down there is a complex relationship between the two users.
It is based on any combination of the following factors:
- quality
- like / dislike
- agreement / disagreement
- mood of the voting user
To mine all those from a single click it is necessary to track those clicks over many interactions. Only that way does it become possible to get a sense of what is on the users mind when he/she clicks that 'vote' link.
Votes on urls that are submitted are not subject to the 'agreement / disagreement' element, but all the others apply.
Allowing people to 'climb' faster based on click patterns is a very complex data mining issue. The 'dummy buttons' idea is an interesting one, to some extent this already happens on HN (the up/down vote buttons do not always work the way you think they do).
One hard thing to achieve is to get rid of the effect simply measuring karma has on HN. Because the users see the points and the karma they collect there is a feedback loop in HN that tends to get out of hand.
Comments posted early on in a thread will get some karma regardless of how good they are in the longer term.
Ideally a 'good' user (so one that in the long term will acquire access to all the features) should be identified early, but then that would open the door to gaming the system even further. The road up the hill is pretty quick anyway for most users, as far as I can see you get all the goodies from 200 points upwards (other than administrator privileges), an active user could collect that in a week, and by then will be reasonably well versed in HN etiquette to not abuse these powers.
>> Some advantages: - Members can climb the 'ranks' faster which satisfies peoples' thirst for power (who doesn't like being promoted).
That's a huge disadvantage. I used to run a site that had a karma system, and the more karma you had, the bigger the impact when you gave someone else positive or negative karma. That "thirst for power" just got bigger and bigger and bigger, and it ended up almost ruining what had previously been an extremely good community. When we (the other admins and mods) decided to drop the system, it only returned to normal after a lot of the previously-helpful senior members were given temporary bans so they could 'cool off' for a while.
The more complex karma becomes and the more people keep obsessing about it and wanting to improve it, it's just going to impede the real progress of the community.
That's a huge disadvantage. I used to run a site that had a karma system, and the more karma you had, the bigger the impact when you gave someone else positive or negative karma. That "thirst for power" just got bigger and bigger and bigger, and it ended up almost ruining what had previously been an extremely good community. When we (the other admins and mods) decided to drop the system, it only returned to normal after a lot of the previously-helpful senior members were given temporary bans so they could 'cool off' for a while.
The more complex karma becomes and the more people keep obsessing about it and wanting to improve it, it's just going to impede the real progress of the community.
The greatest danger of a karma system is for it to become a means unto itself. I think you've fallen headlong into that danger. You don't have a single sentence about how your system would make the site better, which leads me to the conclusion that it isn't even a goal of yours. Without that as a goal, you're not going to create a decent design except by sheerest accident.
I was wondering about karma on HN and other sites, like stackoverflow, which are community driven. The essence of karma is to be a self-controlling mechanism which empowers users based on their past contributions.
One aspect of this is to award 'power' to users (like moderating and downvoting) after certain karma point thresholds. My idea is to speed this process up by promoting 'powers' in a more flexible way.
How it would work - a 'predictive karma algorithm': Analyse usage patterns of members with higher 'ranks' (i.e. more 'power') and try to match those usage patterns to new members' behaviour to determine if they qualify for an early promotion in the 'ranks'.
This could entail giving dummy access to all functionality to new users, like downvoting and deletion. The dummy buttons would not actually do anything but the actions are stored in a DB to be used in the usage pattern comparison algorithm.
Some advantages: - Members can climb the 'ranks' faster which satisfies peoples' thirst for power (who doesn't like being promoted). - It would show full site functionality at an earlier stage so users can see the site's richness immediately. - Speeds up members' site loyalty.
Some disadvantages: - Could become complex to create a well balanced algorithm - Measuring any comparative advantage over using thresholds is difficult.
It is just a seed of an idea but I would think some kind of a 'predictive karma algorithm' could further improve the user experience.
Your thoughts?