I spent a few minutes today starting to learn how to code a curation bot. I started from this helpful post by @xeroc. My general plan is this:
- Learn how to do it.
- Use the bot to generate a stream of articles which I can manually upvote (I'm not a big fan of downvoting.)
- Continuously tune the filtering to suit my preferences.
- As I come to trust the bot, begin letting it make some decisions without my help.
- Continue to improve the bot according to the following goals:
- Only upvote posts that I would have upvoted manually (i.e. automatically curate the site according to my own tastes).
- Raise revenue through curation.
- Increase its autonomy.
Basically, I want to automate the same things I'd be doing manually. I want good voting decisions in order to help raise the value of my long-term position in steem power, and I want short and long term revenue. I assume most bot authors are pursuing similar goals.
However, the comments in @xeroc's post caught my attention. Apparently, the idea of creating upvote bots is somewhat controversial. For example,
Wanna to vote against the voting bots. Sorry.
and
This is not good for the system really. To be clear I'm not "blaming" you - people will do what they have to do given a certain rule set. It's just that the system needs to tweak the incentives so that scripted participation, instead of human judgement in terms of quality, is not rewarded. Otherwise what's the point?
and
Why make it easier for people to bring vote quality down? If someone else wants to write their own anti-social bot, they can do it themselves.
Needless to say, the fact that I started this project implies that I disagree with those commenters, but I'll try to be open to persuasion. At heart, isn't Google's PageRank just a massive curation engine? Couldn't some well-tuned algorithms help to improve the feed quality on steemit? To me, curation seems like a perfect job for an algorithm (specifically, an evolutionary algorithm).
Yes, early generation algorithms will have problems, but over time they will improve. In the long run I think that - as with Google's search rankings - a hybrid of human and machine curation will produce the best results. What do you think?