This will be a trip into theory-land, so it may be frustrating to users who dislike talk about "vaporware" and want to see how something works in practice. I understand where you're coming from, but I submit it's valuable to raise these questions early. This is in any case not intended to supplant discussion about how things are things are currently progressing.
First, though, consider the benefits that such a search engine could bring, both to content consumers and content providers, if it really did return results sorted according to average community preferences. Suppose you wanted to find out if you had a knack for publishing recipes online and getting some AdSense revenue on the side. You take a recipe that you know, like apple pie, and check out the current results for "apple pie". There are some pretty straightforward recipes online, but you believe you can create a more complete and user-friendly one. So you write up your own recipe, complete with photographs of the process showing how ingredients should be chopped and what the crust mixture should look like, so that the steps are easier to follow. (Don't you hate it when a recipe says "cut into cubes" and you want to throttle the author and shout, "HOW BIG??" It drove me crazy until I found CookingForEngineers.com.) Anyway, you submit your recipe to the search engine to be included in the results for "apple pie", and if the sorting process is truly meritocratic, your recipe page rises to the top. Until, that is, someone decides to surpass you, and publishes an even more user-friendly recipe, perhaps with a link to a YouTube video of them showing how to make the pie, which they shot with a tripod video camera and a clip-on mike in their well-lit kitchen. In a world of perfect competition, content providers would be constantly leapfrogging each other with better and better content within each category (even a highly specific one like apple pie recipes), until further efforts would no longer pay for themselves with increased traffic revenue. (The more popular search terms, of course, would bring greater rewards for those listed at the top, and would be able to pay for greater efforts to improve the content within that category.) But this constant leapfrogging of better and better content requires efficient and speedy sorting of search results in order to work. It doesn't work if the search results can be gamed by someone willing to spend effort and money (not worth it for the author of a single apple pie recipe, but worth it for a big money-making recipe site), and it doesn't work if it's impossible for new entrants to get hits when the established players already dominate search results.
Efficient competition benefits consumers even more for results that are sorted by price (assuming that among comparable goods and services, the community promotes the cheapest-selling ones to the top of the search results, as "most desirable"). If you were a company selling dedicated Web hosting, for example, you would submit your site to the engine to be included in results for "dedicated hosting". If you could demonstrate to the community that your prices and services were superior to your competitors', and if the ranking algorithm really did rank sites according to the preferences of the average user, your site could quickly rise to the top, and you'd make a bundle on new sales -- until, of course, someone else had the same idea and knocked you out of the top spot by lowering their prices or improving their services. The more efficient the marketplace, the faster prices fall and service levels rise, until the prices just covered the cost of providing the service and compensating the business owner for their time. It would be a pure buyer's market.
It's important to precisely answer the question: Why would this system be better than a system like Google's search algorithm, which can be "gamed" by enterprising businesses and which doesn't always return the results first that the user would like the most? You might be tempted to answer that in an inefficient marketplace created by an inefficient search result sorting algorithm, a user sometimes ends up paying $79/month for hosting, instead of the $29/month that they might pay if the marketplace were perfectly efficient. But this by itself is not necessarily wasteful. The extra $50 that the user pays is the user's loss, but it's also the hosting company's gain. If we consider costs and benefits across all parties, the two cancel out. The world as a whole is not poorer because someone overpaid for hosting.
The real losses caused by an inefficient search algorithm, are the efforts spent by companies to game the search results (e.g. paying search engine optimization firms to try and get them to the top Google spot), and the reluctance of new players to enter that market if they don't have the resources to play those games. If two companies each spend $5,000 trying to knock each other off of the top spot for a search like "weddings", that's $5,000 worth of effort that gets burned up with no offsetting amount of goods and services added to the world. This is what economists call a deadweight loss, with no corresponding benefit to any party. The two wedding planners might as well have smashed their pastel cars into each other. Even if a single company spends the effort and money to move from position #50 to position #1, that gain to them is offset by the loss to the other 49 companies that each moved down by one position, so the net benefit across all parties is zero, and the effort that the company spent to raise their position would still be a deadweight loss.
On the other hand, if search engine results were sorted according to a true meritocracy, then companies that wanted to raise their rankings would have to spend effort improving their services instead. This is not a deadweight loss, since these efforts result in benefits or savings to the consumer.
I've been a member of several online entrepreneur communities, and I'd conservatively estimate that members spend less than 10% of the time talking about actually improving products and services, and more than 90% of the time talking about how to "game" the various systems that people use to find them, such as search engines and the media. I don't blame them, of course; they're just doing what's best for their company, in the inefficient marketplace that we live in. But I feel almost lethargic thinking of that 90% of effort that gets spent on activities that produce no new goods and services. What if the information marketplace really were efficient, and business owners spent nearly 100% of their efforts improving goods and services, so that every ounce of effort added new value to the world?
Think of how differently we'd approach the problem of creating a new Web site and driving traffic to it. A good programmer with a good idea could literally become an overnight success. If you had more modest goals, you could shoot a video of yourself preparing a recipe or teaching a magic trick, and just throw it out there and watch it bubble its way up the meritocracy to see if it was any good. You wouldn't have to spend any time networking or trying to rig the results, you just create good stuff and put it out there. No, despite whatever cheer-leading you may have heard, it doesn't quite work that way yet -- good online businessmen still talk about the importance of networking, advertising, and all the other components of gaming the system that don't relate to actually improving products and services. But there is no reason, in principle, why a perfectly meritocratic content-sorting engine couldn't be built. Would it revolutionize content on the Internet? And, could Search Wikia be the project to do it, or play a part in it?
Whatever search engine the Wikia company produced, it would probably have such a large following among the built-in open-source and Wikipedia fan base, that traffic wouldn't be a problem -- companies at the top of popular search results would definitely benefit. The question is whether the system can be designed so that it cannot be gamed. I agree with Jimmy Wales's stated intention to make the algorithm completely open, since this makes it easier for helpful third parties to find weaknesses and get them fixed, but of course it also makes it easier for attackers to find those weaknesses and exploit them. If you think Microsoft paying a blogger to edit Wikipedia is a problem, imagine what companies will do to try and manipulate the search results for a term like "mortgage". So what can be done?
The basic problem with any community that makes important decisions by "consensus" is that it can be manipulated by someone who creates multiple phantom accounts all under their control. Then if a decision is influenced by voting -- for example, the relative position of a given site in a list of search results -- then the attacker can have the phantom accounts all vote for one preferred site. You can look for large numbers of accounts created from the same IP address, but the attacker could use Tor and similar systems to appear to be coming from different IPs. You could attempt to verify the unique identity of each account holder, by phone for example, but this requires a lot of effort and would alienate privacy-conscious users. You could require a Turing test for each new account, but all this means is that an attacker couldn't use a script to create their 1,000 accounts -- an attacker could still create the accounts if they had enough time, or if they paid some kid in India to create the accounts. You could give users voting power in proportion to some kind of "karma" that they had built up over time by using the site, but this gives new users little influence and little incentive to participate; it also does nothing to stop influential users from "selling out" their votes (either because they became disillusioned, or because they signed up with that as their intent from the beginning!).
So, any algorithm designed to protect the integrity of the Search Wikia results would have to deal with this type of attack. In a recent article about Citizendium, a proposed Wikipedia alternative, I argued that you could deal with conventional wiki vandalism by having identity-verified experts sign off on the accuracy of an article at different stages. That's practical for a subject like biology, where you could have a group of experts whose collective knowledge covers the subject at the depth expected in an encyclopedia, but probably not for a topic like "dedicated hosting" where the task is to sift through tens of thousands of potential matches and find the best ones to list first. You need a new algorithm to harness the power of the community. I don't know how many possible solutions there are, but here is one way in which it could be done.
Suppose a user submits a requested change to the search results -- the addition of their new Site A, or the proposal that Site A should be ranked higher. This decision could be reviewed by a small subset of registered users, selected at random from the entire user population. If a majority of the users rate the new site highly enough as a relevant result for a particular term, then the site gets a high ranking. If not, then the site is given a low ranking, possibly with feedback being sent to the submitter as to why the site was not rated highly. The key is that the users who vote on the site have to be selected at random from among all users, instead of letting users self-select to vote on a particular decision.
The nice property of this system is that an attacker can't manipulate the voting simply by having a large number of accounts at their control -- they would have to control a significant proportion of accounts across the entire user population, in order to ensure that when the voters were selected randomly from the user population, the attacker controlled enough of those accounts to influence the outcome. (If an attacker ever really did spend the resources to reach that threshold point, and it became apparent that they were manipulating the votes, those votes could be challenged and overridden by a vote of users whose identities were known to the system. This would allow the verified-identity users to be used as an appeal of last resort to block abuse by a very dedicated adversary, while not requiring most users to verify their identity. This is basically what Jimmy Wales does when he steps in and arbitrates a Wikipedia dispute, acting as his own "user whose identity is known".)
This algorithm for an "automated meritocracy" (automeritocracy? still not very catchy at 7 syllables) could be extended to other types of user-built content sites as well. Musicians could submit songs to a peer review site, and the songs would be pushed out to a random subset of users interested in that genre, who would then vote on the songs. (If most users were too apathetic to vote, the site could tabulate the number of people who heard the song and then proceeded to buy or download it, and count those as "votes" in favor.) If the votes for the song are high enough, it gets pushed out to all users interested in that genre; if not, then the song doesn't make it past the first stage. If there are 100,000 users subscribed to a particular genre, but it only takes ratings from 100 users to determine whether or not a song is worth pushing out to everybody, that means that when "good" content is sent out to all 100,000 people but "bad" content only wastes the time of 100 users, the average user gets 1,000 pieces of "good" content for every 1 piece of "bad" content. New musicians wouldn't have to spend any time networking, promoting, recruiting friends to vote for them -- all of which have nothing to do with making the music better, and which fall into the category of deadweight losses described above.
An automeritocracy-like system could even be used as a spam filter for a large e-mail site. Suppose you want to send your newsletter to 100,000 Hotmail users (who really have signed up to receive it). Hotmail could allow your IP to send mail to 100,000 users the first time, and then if they receive too many spam complaints, block your future mailings as junk mail. But if that's their practice, there's nothing to stop you from moving to a new, unblocked IP and repeating the process from there. So instead, suppose that Hotmail stores your 100,000 received messages temporarily into users' "Junk Mail" folders, but selectively releases a randomly selected subset of 100 messages into users' inboxes. Suppose for arguments' sake that when a message is spam, 20% of users click the "This is spam" button, but if not, then only 1% of users click it. Out of the 100 users who see the message, if the number who click "This is spam" looks close to 1%, then since those 100 users were selected as a representative sample of the whole population, Hotmail concludes that the rest of the 100,000 messages are not spam, and moves them retroactively to users' inboxes. If the percentage of those 100 users who click "This is spam" is closer to 20%, then the rest of the 100,000 messages stay in Junk Mail. A spammer could only rig this system if they controlled a significant proportion of the 100,000 addresses on their list -- not impossible, but difficult, since you have to pass a Turing test to create each new Hotmail account.
The problem is, there's a huge difference between systems that implement this algorithm, and systems that implement something that looks superficially like this algorithm but actually isn't. Specifically, any site like HotOrNot, Digg, or Gather that lets users decide what to vote on, is vulnerable to the attack of using friends or phantom users to vote yourself up (or to vote someone else down). In a recent thread on Gather about a new contest that relied on peer ratings, many users lamented the fact that it was essentially rigged in favor of people with lots of friends who could give them a high score (or that ratings could be offset unfairly in the other direction by "revenge raters" giving you a 1 as payback for some low rating you gave them). I assume that the reason such sites were designed that way is that it just seemed natural that if your site is driven by user ratings, and if people can see a specific piece of content by visiting a URL, they should have the option on that page to vote on that content. But this unfortunately makes the system vulnerable to the phantom-users attack.
(Spam filters on sites like Hotmail also probably have the same problem. We don't know for sure what happens when the user clicks "This is spam" on a piece of mail, but it's likely that if a high enough percentage of users click "This is spam" for mail coming from a particular IP address, then future mails from that IP are blocked as spam. This means you could get your arch-rival Joe's newsletter blacklisted, by creating multiple accounts, signing them up for Joe's newsletter, and clicking "This is spam" when his newsletters come in. This is an example of the same basic flaw -- letting users choose what they want to vote on.)
So if the Wikia search site uses something like this "automeritocracy" algorithm to guard the integrity of its results, it's imperative not to use an algorithm vulnerable to the hordes-of-phantom-users attack. Some variation of selecting random voters from a large population of users would be one way to handle that.
Finally, there is a reason why it's important to pay attention to getting the algorithm right, rather than hoping that the best algorithm will just naturally "emerge" from the "marketplace of ideas" that results from different wiki-driven search sites competing with each other. The problem is that competition between such sites is itself highly inefficient -- a given user may take a long time to discover which site provides better search results on average, and in any case, it may be that Wiki-Search Site "B" has a better design but Wiki-Search Site "A" had first-mover advantage and got a larger number of registered users. When I wrote earlier about why I thought the Citizendium model was better than Wikipedia, several users pointed out that it may be a moot point, for two main reasons. First, most users will not switch to a better alternative if it never occurs to them. Second, for sites that are powered by a user community, it's very hard for a new competitor to gain ground, even with a superior design, if the success of your community depends on lots of people starting to use it all at once. You could write a better eBay or a better Match.com, but who would use it? Your target market will go to the others because that's where everybody else is. Citizendium is, I think, a special case, since they can fork articles that started life on Wikipedia, so Wikipedia doesn't have as huge of an advantage over them as they would if Citizendium had to start from scratch. But the general rule about imperfect competition still applies.
It's a chicken-and-egg problem: You can have Site A that works as a pure meritocracy, and Site B that works as an almost-meritocracy but can be gamed with some effort. But Site B may still win because the larger environment in which they compete with each other, is not itself a meritocracy. So we just have to cross our fingers and hope that Search Wikia gets it right, because if they don't, there's no guarantee that a better alternative will rise to take its place. But if they get it right, I can hardly wait to see what changes it would bring about.