Political science has plenty of literature about sophisticated voting systems (vs the simpler systems in general use today): there one can find a good part of the theoretical background. (I am focusing here on the feedback as fundamental to the goal.)<p>An example: <i>Numbers Rule: The Vexing Mathematics of Democracy, from Plato to the Present</i> by George Szpiro is a primer.<p>Of course it depends on the goal defining "improvement" - heuristics, informative/insightful/funny, educational, sinister goals (compulsion, sensationalism etc.).<p>For example, votes should be clustered and weighted (the Professor has more more insight than the Child - so "weighted" - in some areas - so "clustered").<p>Also: technically, that platform should be heavily modified to allow for fruitful discussion.
There should be a trigger mechanism that recognizes the type of spam seen, for example, following Twosetviolin posts. There is a specific person who can only barely play piano that posts his songs in <i>every</i> comment thread, even though his links are clearly attempting to derail every conversation in these threads.<p>It should be super easy for YouTube to notice this link over and over and over again in the comments for a single video. There isn't a case where this would be acceptable behaviour, so why not sense it and end it? It is <i>the</i> definition of spam.
Use the same systems as HN: down voted comments disappear, getting down voted stops you commenting, new users are rate limited as are replied on the same thread.