I'll take them in order.
Filtering the undesirable content:
Each visitor has a preferences page where they can change their 'vote' for what weight they would give to certain thresholds. The threshold for what kicks certain info out permanently could end up at, say 5%. If so, that would mean that after passing 100 total votes, if more than 5% was claiming that was irrelevant, offensive or (can't remember the other 3rd factor) but it's kind of like being counter trolling/flaming... if more than 5% say it's bad, then it vanishes. If this seems too low, people will naturally trend toward increasing their threshold setting which averages the overall one slightly higher.
We have not even considered some type of freshness mechanism. We did discuss the issues that arise from storing links vs. actual content and what happens when a referenced link dies. Our consensus was that depending on our ability to acquire massive storage space, we might go with something similar to Google's 'cached' pages where full credit is attributed. This addresses issues of page changes as well as acts as a type of archive for what someone later recanted.
A side issue on this is whether or not what some random page says should even be considered as submittable fact. Ideally, the actual research, study or policy should be referenced only. TBD
Wiki's referening is passive and does not provide any info on that ref. This is designed so that whichever reference is most relevant is the one that jumps out at you. Insignificant or minor ones get much less attention. Think of those word clouds.
Aside from an "Explicit" tag (thresholded and voted as well), no content will be censored. The very structure doesn't allow it. Without going into minute details, multiple votes are seen as one person changing their mind or reasserting their same vote - i.e. they're only kept for trend type graphs. Rogue page changes can't happen because of the law of averages, once a page passes into popularity.
The format for data storage is such that the only way to see the info is to use the site correctly. The only way to input any info is to vote correctly. Even the creators won't have access to go change any info by altering data because it's not stored in accessible tables and such. In this way, even government subpoena's can't get force something to become secret, hidden, altered or removed. Picture P2P file sharing combined with Win2k distributed file system and cloud processing ala SETI's project.
We predict lots of interactivity, just on an anonymous basis. People have an inherent interest in seeing some instant result change before their very eyes. They can watch a topic in live mode with a specific demographic breakdown and see the average and trends change. If they make a significant contribution, they will want to watch the results as their argument changes opinions of the others. It's kind of like Up and Down votes where opinion and merit remain separated.
Also, as we've seen here on DP, if something gets Downvoted 'for no apparent reason', the poster feels the need to break down the issue into sub-sections to isolate where they went wrong. This is at the very heart of the project because it causes the inter-linking and dependencies on which our human decisions are made. It's this aspect laid out for new people that has the greatest impact on coaxing more research.
When on one topic, there is available only the info that's relevant and directly focused on that topic. Links will show side bars should one want to take that path but in not doing so, there is almost no distraction. Favorite topics will appear on the left for close watching as will a drill-down category list and recent, popular, controversial, live, etc. Lots of options are available to travel around but once you're in one place, it's very focused. Think of a personal conversation.