After posting an article last month detailing the output of a collusion detection algorithm I had applied to posts on Steemit, I was initially met with a lot of backlash from presumably "false positives" who were very offended that their name had been included among lists of potential voting rings. First of all I want to whole-heartedly apologize for any offence caused. This was never the intention. So for future collusion detection reports, I would like to qualify that these are not accusations. I am working on a pipelined process which includes a well-known collusion detection algorithm to throw up suggestions of behavior similar to that of voting rings. While it is not perfect, it is not without some valid output either. I hope to, with the help of others, improve the accuracy of this process, over time and eventually reduce the amount of false positives it throws out.
However, while I was initially very upset with the reaction my report had received among the community, slowly but surely, I began to also receive a great deal of respect and encouragement from others, who have convinced me to further pursue the project in the hopes of reaching a valuable end product which can act as a force for good on Steemit. Voting trails are a very real concept in the community. Some embrace them and believe there to be nothing wrong. Others believe it to completely take away from the experience of a natural social network where users upvote out of interest, not mutual financial gain. My opinion on this is somewhere in between but really is irrelevant to the task at hand which merely seeks to research how prevalent this really is. And if it is very prevalent, I want to ask the question - what can/should we do about this, if anything?
I am working on a whitepaper detailing the algorithm being used and some assumptions I have made during the process to produce the output seen last month. I hope to publish this whitepaper shortly. This should clarify some things and open to the door for discussion regarding the improvement of other things.
I also hope to begin reapplying the algorithm to the latest steemit posts in order to again throw up some output for discussion. Note that this will again be naming some potential voting rings. This time I hope that I have qualified satisfactorily that these are not accusations of guilt in anyway, but rather the output of an algorithm which one day could be used more accurately to offer greater transparency regarding the use of voting rings in the community.
Stay tuned :)
