> Yeah, that is my main concern too. Maybe instead of tallying up/downvotes verbatim one could do clustering analysis on the voters. Votes by people who have si

> Yeah, that is my main concern too. Maybe instead of tallying up/downvotes verbatim one could do clustering analysis on the voters. Votes by people who have similar voting-behavior might be of more interest to a user than votes by voters who have uncorrelated or even inversely correlated behavior. That's the essential idea behind Amazon and Netflix recommendations so it is quite workable. >> It would be better as a baseline to use the forum's memory as a filter for each user > Not sure what you're suggesting. I'm saying you have this massive haul of threads of which the majority are never referenced because they happened in the past. Normally users need to search this backlog but often don't know the right search keywords to begin searching. Thus they ask the same old questions of the old-timers who become crankier and crustier by the day because of this. Since the same questions are asked multiple times you should be able to tally the answers together into a cogent form in a similar way to the method that autotldr bot uses on Reddit to summarize news stories. Then you could present this information in a currently running thread that appears to ask the same questions.