October 30, 2012 § Leave a comment
So, what’s up with magic. First, let’s assume that any post (position or response) on WhyWeVote can be either a “good” or a “less-than-good” post. A “good” post is one that helps us make a decision, that educates us, that is valuable to whatever workflow we are pursuing. A “less-than-good” post… not so much.
So, our goal is to give good posts exposure. So we came up with a constantly evolving algorithm to sort through the posts to:
- allow good positions get more exposure
- prevent mobs from suppressing good positions due to bias
- prevent old good positions from disappearing amongst the new noise
At the end of the day, WhyWeVote is a sort of a social brain. At any node/post, we are trying to identify what is the most important information for that specific context and the “magic” sorting algorithm seeks to do just that.
We expect the algorithm to evolve, but if you get frustrated and want something different, you can always switch to sorting by time.
November 7, 2008 § Leave a comment
OK, this is really pissing me off. Not because Obama won or because McCain lost – both are strong candidates and that is a different story. But I had two of my friends tell me that they were able to vote without any verification of their identities!!!!!!!!!!!!!!!!!! Why do we have a democracy if the voting process is not being maintained?!?!?!?
I think there should be a move by the government to bring voting online… or offline with a requirement for use of your passport.
November 4, 2008 § Leave a comment
Now that we are getting a few more votes, I am able to start playing around with the sorting algorithms.
So, when adjusting the algorithm, I am going with the following:
- never touch the content – i.e. the algorithm should be content independent. If there is a bias in the results, blame (and fix) the algorithm
- take how people vote, factor in how young is the vote (time stops being a factor after a short while), add some randomness to keep things interesting…
November 3, 2008 § Leave a comment
OK! So, 2:23 AM and we got the first version of our sorting algorithm. I’ve spent a long time thinking about bias and how we want our system to behave… Came to the following conclusions:
- Anonymity: Information should be evaluated on the merit of information and information only. Hence the system is anonymous. This should prevent people who like posting garbage simply to get to the top from doing so.
- Bias: bias is not when a system present one-sided information. Bias is when you can’t get any other information if youa re looking for it. I don’t care whether people support Obama or McCain – both have strong points to them and there are reasons to vote both. Recently I had a Vote “Everything” with content “www.barakobama.com”. Now, come on, you could at least put it into the link section. So this is not a very informative post, so it should end up lower on the queue.
- Volatility: I want to make sure that the algorithm is difficult to figure out so that gaming of the system is minimized. Thus I introduced some randomness. This will also allow information that is not initially recognized to get on the radar.
Right now the algorithm is fairly simple. However, I’ll be fine-tuning it to try to make sure that it favors good, well-sourced, reliable information. I can already see some of the problems that will arise, but it will be interesting to jump off that bridge when we get there.
Please post your votes and vote. I am very curious how this will turn out… if it turns out…
November 2, 2008 § Leave a comment
OK, I completely changed the architecture of tagging dropping the issue table. This is not something you will see on the surface, but it will affect the future development for the better.
Had some very interesting discussions on where this may grow. One interesting point that came up was the fact that I am forcing information to be anonymous. Why do I do that? Because I am a fim believer that information should be evaluated on basis of information rather than who submitted it. Once we start having problems with people submitting explicit lies, then we will see how to deal with this, but, in the meantime, it is important to state explicitly that we want to try to avoid any sorts of labels for information. We also want to try to minimize the effect of the mob… Digg has been an excellent example: several overactive individuals completely ruin the environment for everybody else (irrespective of whether you are left or right, just so long as you are reaonable).