You are viewing a single comment's thread from:
RE: Towards a reputation system suitable for SteemSTEM
The engagement metric rewards replying, period. I haven't looked specifically at #steemstem replies, but your top three accounts for enagement are:
1 erh.germany 2.886
2 agmoore 2.76
3 steemit-italia 2.623
The first two seem legitimately engaged. Half of the third account's replies are boilerplate:
Congratulazioni!
Hai appena ricevuto il voto del trail di @steemit-italia
...
This seems like a poor thing to reward in a reputation score. (Particularly since you weight towards longer responses.) I would suggest looking at:
- Does anybody reply to the comment? Do they upvote it? Bonus points if it's the original author?
- Is the comment unique, or the same comment on many posts?
- Are they engaging with a broad spectrum of writers or just the same writer over and over?
Of course, all these can be gamed too.
What you suggest is a much more complex thing. I wouldn't mind though to have a program capable of doing it. On the other hand, it raises the question in me whether I would like to be the results revealed to the public. When it's simple statistics, I am fine with it. But when I feel like I am under a more sophisticated microscope I get uneasy... hmm... but then - If I would suck by this criteria, I wouldn't be listed anyway.
If I am in favor with a particular author and encounter his content more than that of others, am I than a bad community member? ;-)
Well, it would seem a little odd to claim you were a highly-regarded member of the #steemstem community if you were just specifically a @lemouth fan. :)
I agree.
But as far as I know the community, almost nobody would behave in that way to comment only on one author. Without having the numbers of evidence I dare to make this assumption:)
We manually removed all real bots, but steemit-italia looks like being half a bot (which is why we left it there). We definitely need to improve the engagement metric (at a reasonable cost).
Concerning your suggestions, thanks a lot! They are good and I like them. I don't know what will be part of the next version, but some of them may (I have to think further about the how to do what in practice).