I'm not sure it's accurate to say that it's a recommendation system. It can function as a recommendation system in part but it's more accurate to say that it's exactly what it says on the tin: "a system which manages quanta of trust."
Trust can mean a lot of things. Trust can mean that a user is considered "a good part of the community." Trust can mean that a source provides content which is in line with with expectation. Trust can mean exactly what it does in common parlance, that you extend a level of predictive expectation to another entity.
It's certainly possible to get into various types of trust which all function simultaneously, but at this point Steemit really needs some sort of lensing system, some way to slice through the piles of content which get dumped on it every day in ways that are meaningful to individual users.
I'm not one of those people that sees a blockchain and thinks I have a hammer and every problem is a nail. I'm actually pretty sure that any sort of web of trust system would have to be built on another type of data store, and that could certainly be built into an entirely different platform front end.
But that wouldn't improve Steemit. And that would be a shame.
Well, i see.
As an analyst once I created a content-rating on a classified selling site. it was a hell of a task, I spend 9 months to did that. At it was a simple target function - a score of selling an item and getting revenue to a company. With tons of information in hands.
I'm not sure if we need to put that many resources there.
But someone could experiment with steemit sidechain attaching a content score to each content and each author.
That's a lot more complicated problem then we even need to think about in this context.
Essentially a web of trust is a sparse relationship matrix. You don't adjust trust to all entities, just those you want to. The results provide an ordering of the content but don't change the content. If there's any innovation required for the process, and given that we were doing this 10 years ago with a lot less resources at hand, it's in resolving the matrix for an individual view.
The big hook is that we need to stop trying to think of things from a global perspective because all that does is encourage trying to "game the system" from the top down. Effectively. That's why we see the undesirable whale and bot behavior that we do. It's incentivized because there is a global view.
My feeling is that a blockchain of any sort is the wrong technology to bring to bear on this. If I'm honest, I have to admit that I'm not really sold on a cryptocoin being a really effective way to reward creators and curators, but since that's the underlying premise that we except when we deliberately engage with Steemit – there you go.
I will say this: somebody needs to put the resources into doing this well and doing it soon, if not Steemit then Busy or someone else who really wants to be a successful social media platform first so that they can reward creators and curators secondly, because if there is no successful social media platform integrated, there's no rewards to hand out. If no one uses the system, nobody gets paid.
This is at least one reasonable approach.