> But you are indeed correct that a single metric won't give the best results for individuals and other models would be necessary to act as predictors of desirable content to a specific individual.

I would put it to you that specific individuals are the only kinds of individuals that use any system. Which is a problem when we're talking about creating a single metric that derives the consensus position of all specific individuals seeking their individual goals across an entire operation.

If we could be assured that their goals are at least parallel, we could have some degree of surety that the metric so derived would communicate something meaningful. Consider any online auction house in any MMO that you care to mention. People are there to trade goods for currency. All of their votes/buys are intended to foster that aim. As such, a market price for a product gathered from the consensus position is meaningful. Thus why markets work at all. The consensus intent is meaningful.

Go back to Steemit. The intent of an up vote isn't consistent across in her actors. It means different things to different people – and part of that problem is poor on boarding education, but part of that problem is encouraged by the design of the system. There is no consistent reason to vote things up and because the immediate effect of the vote isn't anything that changes the experience for the voter in any real sense, they have no reason to vote as if it did.

So we can't even accept that discovery methods based on votes are going to really turn up content which users think is good, because good content isn't really the motivator for most up votes. We would like to think it is, and I know I would like to think it is, but it's not.

> Not everybody likes popular content and not all great content is popular or viewed in such a positive manner.

It may even be that content which receives the most votes isn't even near the top of the content that any individuals would like to see. Taken in aggregate, it could just be "not the worst thing," rather than "the best thing," even if everyone was engaged with voting as a signal of content they approve of.

Consider the case in which you have two communities using the same platform. One of them really likes to talk about cryptocurrencies and the other really likes to talk about role-playing games. Not only do they prefer to read about those things, they prefer to write about those things. The majority of the content is focused on that of the community.

You let them vote up content that they like. The communities are roughly the same size. What happens? Well, content which is far enough outside the bounds of either community finds that it has the most votes overall as long as it's at least remotely interesting. This is how you get the Internet overrun by pictures of cats. Pictures of cats are definitely not the best content on the platform, it's just the content which everyone finds least objectionable.

Again, this is in a situation where everyone is actually voting in good faith. As you get further from that possibility, consensus becomes even less meaningful.

> That score still might not translate to an individual's preferences, but it gives us additional information and something we can use to make decisions.

Let's try a little experiment.

Imagine that IMDb scores were set by trying to guess what the final IMDb score in a week after release. How useful would that signal end up being?

For the signal to be useful, it has to originate from a place which is not apathetic but is also not activist. The signal needs to be earnest.

With the overall consensus-derived signal then be useful? Well, no – because all it represents is the consensus of what everyone believes the consensus should be, not what the consensus is. We've already violated the expectation of the signal.

That is what the Q-filter will end up emitting. It will be a more nuanced signal compared to the standard steem voting protocol, but it will still be based on assumptions about the input signal which aren't the case or at least aren't sufficiently guaranteed to be meaningful.

> Also, the effects of apathetic users are reduced. So, the universally mediocre won't automatically float to the top.

But what else can? I mean that in a literal sense. Given that individuals and communities have preferences and even if we assume they vote those preferences, the signals they emit at best are going to be sufficiently fragmented that only the least offensive, most broadly not objectionable content is going to find its way to the top. It has to.

The universally mediocre is the only thing that can float to the top.

Effectively, the Q-filter replaces the tyranny of SP-holders with the tyranny of the majority, but it's still tyranny. But worse, it doesn't give us a signal that we can actually truly interpret as that of quality.

> Lastly, assuming that everything I said fails and turns out to be completely useless, such a way of organizing and distributing rewards does make things more decentralized. When gives some tangible benefit of using this filter to some manner.

The system badly needs decentralization, in a huge way. For all that the rhetoric about the blockchain focuses on its decentralized, impossible to eradicate nature, the entire thing is built around an authoritarian assumption of power and control. I don't even particularly mind that everything on the platform is scaled by your stake. It provides an obvious means of manipulating the commodity, but as a useful platform I can ignore those parts and still continue to leverage the bits that it gets right.

(Giving up the expectation that you will get paid in any significant way for what you create is very freeing for people who have been on this platform a little too long, apparently. I came in with that assumption firmly in place over decades of being an online pundit, so there wasn't much to unlearn.)

> But giving an individual what they want is probably a better target for an ideal filter. But I'm not sure I see a good way of achieving such ends that are also harmonious with some of Steem's stated goals and treating content as something that generates value.

There's no real reason that providing a user content that they are likely to want to engage with breaks steem's stated goals and treating content is something that generates value. Quite the opposite.

If you want content to be earning value, you want that content to be put in front of people who are willing and actively able to provide that value. Burying content in masses of crap, making it hard to discover, is a very aggressive way of making sure that content doesn't earn. Brutally so.

[I wrote a bit on distributed web of trust systems as a means of generating metrics for looking at streams of content on Steemit several months ago,](https://steemit.com/steemit/@lextenebris/steemit-and-the-web-of-trust-a-potential-love-story) and I still think that sort of thing is probably the best way to find and present data to the user. It's completely compatible with things like communities and other self organizing groups. It facilitates rather than impedes.

Unfortunately, it's probably a little too individualist for the folks around here. And it would require another layer on top of the steem blockchain in order to actually record and manage webs of trust.

In fact, [I wrote another article about distributed webs of trust implemented as representative tokens which were issued by every individual user, giving everyone their own currency,](https://steemit.com/steemit/@lextenebris/steemit-and-the-ultimate-recourse-blockchains-for-everyone) but that might actually be too far for anyone to actually comprehend, much less embrace.

But it would be fun.