Firstly, you can check how it works right now here - https://u.community

We’re utilizing a modified NCDAwareRank algorithm to calculate Importance of entities, it helps us detect and normalize the effect of account clusters focused on boosting a single account or doing circles (in this case they still need interactions from outside the cluster to increase its Importance, so they cannot boost each other higher than to a certain extent)

We’re also using trust action to form a network of trusted accounts, which allows us to regulate the flow of Importance, pessimizing accounts with below-average validity rate.

Consider user A upvoting a post P, authored by user B as an upvote from account A to account B in the context of post P. This uvote decays over time, and another upvote by user A to a different post P1 by user B actually refreshes the link.

This way, a downvote from user A to any post by user B will sever the link, removing the flow of Importance from A to B. So, before you can take something back with a downvote, you have to give something first with an upvote.

Because the Importance rating can be thought of as a Google PageRank descendant, you assume that upvoting works just fine, as by upvoting a post or account you raise its rank at the cost of all other entities, as the sum of all the ranks is always 1. You can think of it as the space engine from Futurama - you don’t push the ship forward, you push all the universe back. You can read more in our yellow paper - https://github.com/UOSnetwork/uos.docs/blob/master/yellow_paper/uos_yellow_paper.eng.pdf

As for the basic concept you can check any PageRank video, for example: https://www.youtube.com/watch?v=P8Kt6Abq_rM

An in-depth explanation by one of our scientists - https://u.community/posts/8004