RE: A Different Approach to Curation: Quality Based Consensus and Quality Discovery
It's an interesting approach, but as with any attempt at measuring subjective quality - it has major holes.. At least if I have interpreted your idea as you intended.
- There is an assumption that if 'everyone agrees that the post is quality, then it is quality'. As most of us know, it is entirely possible for a majority of a crowd to think in a way that lacks enlightened awareness of the deeper issues and so to make decisions, en mass, that are of a low quality but which the majority prefer anyway. If they learned certain things they were ignorant of, then they would choose differently. So in a way, what this assumption gives us is not a measurement of quality, but a measurement of the perceptions of a given crowd in a given time period.
This in itself isn't necessarily out of alignment with what the community, as a group, might want to occur - but it has the effect of skewing things away from 'the absolute finest' and towards mediocrity - since the bell curve principle means that the vast majority of people are not top experts in everything and thus cannot assess everything, on average, in a way that is more than mediocre. There will still be greatness uncovered, but the algorithm needs to allow more flexibility to be sure to identify it.
To put this another way, if everyone knew what greatness was - there would be a whole lot more greatness ;)
- On a more mechanical level - unless i missed something - isn't it possible for people to create large numbers of accounts and use them to skew the averages?
cheers for imagining!
p.s. something is broken with the text input here as the numbering of my comment points here refuses to be updated and saved!
To address your first concern, there is a disagreement penalty built into the formula which sorts the posts. This helps protect the minority opinion. If you choose the variance based measure carefully, it would require a large super majority of voters to cancel the effects of that disagreement. That being said, smaller users with differing opinions than the community might get drowned out, but that currently happens anyway. Your groupthink concern definitely has some validity, although all social media suffers from mediocre content that appeals to the masses. Definitely a tough problem to think about.
To address your second concern, you could create a large number of account to skew the simple average and take more of the curation rewards, but you would have to buy Steem Power for this attack to impact the rating that is used to sort and display the posts as a filter since it is a weighted average.
The disagreement penalty will have an effect of limiting 'dissent' and thus will erase the voice of the true experts who disagree with the crowd - yes. So I just want to be clear that this is not a true solution to the issue of identifying actual quality, but rather it is a way of increasing the responsiveness in the system to the ideas about quality held by the majority in the community.
I'm not sure it's possible to do much about this problem without either having a computer that is literally more intelligent than most humans or by possibly using reputation in a clever way that factors in 'voices of authority'. The problem with alleged voices of authority is the same as with the alleged measurements of quality - in that if measurements of reputation are defined by the crowd - en mass - and the crowd - en mass - tends towards mediocrity, then those with the highest reputation will also be skewed towards mediocrity. lol. This is not entirely fair or true of course, since in the real world there are other factors involved that make a reputation score serve it's purpose in a useful way - to some extent (although it can be gamed).
Yes, you would need to invest in Steem Power to make the scam work but you could also withdraw it all when you are done by powering down - so it's a scam with a high entry price, but with potentially high rewards. Your investment of a million dollars to create new accounts might artificially inflate the price of steem, leaving you a nice golden getaway car when you are finished vampiring the rewards pool!
Well, the disagreement penalty may limit "dissent", it also can encourage dissent. If a post is highly rated, and then an expert disagrees, he adds variance to the system, decreasing the score. If you have a large enough minority of experts, they should be able to hold their own in disagreement.
But if you dissent in that you think something is quality and everyone disagrees then penalty hurts you, so you are better to just join the crowd.
Yes, I would agree that this is more of a "community" rating rather than one of objective quality. Objective quality is impossible.
I actually have some good ideas for how to overcome these problems that don't require any changes to the existing algorithms - but I will need to write them down in detail. I'll link back to your post here when I put the new post together. Well, it's a 'good' idea according to my own subjective assessment anyway. :)