Note that I have the software skills to create everything that I've listed here. This isn't pipe dreaming.
First, some basic distinguishing design goals, then more detail.
At first, I'll probably be the only moderator and will exercise moderator powers through the back end of the database, at the SQL level. I hope to quickly be able to appoint new moderators and code front-end moderator powers. (Then code the ability to appoint people with the power to appoint other moderators, etc.) Depending on how quickly users sign up, many of these ideas may require limiting the number of users until human capacity (number of moderators, etc.) allows more. I can live with that.
In the beginning, I (and then other moderators) hope to greet and mini-interview each user who posts. I make sure they respond, make sure their command of English matches that in their ad (otherwise I have to suspect they copied the ad, which is fraud.), etc. Then I can either suspend the ad (and user), leave it be, or mark it something to the effect of "basic moderator approval." Depending on the greeting conversation, I might give the ad some sort of gold star. That approval level will then be searchable.
One idea is to make IP ranges (individual addresses aren't needed) public, or optionally public. That will expose some Russian mail order brides. I'm not necessarily going to ban mail order brides, but if their location says US, that's fraud.
If I (we) catch affiliate marketing schemes or other clear fraud, I hope to have the time or delegate the task of chasing them as far as I (we) can. I've reported the sponsors of such operators to the BBB before, and this got results. As Craig's List has recently posted, "FUN FACT: Reporting a scammer's 'affiliate ID' to their affiliate marketing program often results in confiscation of the scammer's ill-gotten gains by the affiliate program." My personal experience is that there's some truth to this.
I hope to appoint quite a number of users (or have them appointed by other moderators) to the "purge"-power rank. That means that we trust them enough to use their judgement on when to flag off fraud.
I hope to build a feedback / rating system. If a user responds to an ad, the ad poster will have the ability to say, "Not right for me, but a good response" and thus build up the other user's "whuffie" (rating, trust level).
Eventually, the system can distinguish amongst who gave what user what "whuffie." This might lead to different camps following moderators (or other privileged user) who give "gold stars" to different kinds of users: the intellectual camp, the hippie camp, etc.
In the past, my responses can't all have been lame. I've sent some good responses to some unusual ads that not a whole lot of other people would have answered, and I heard zip. I want to limit that experience and help my users. To wit:
The system will show statistics on the number and timing of responses and responses to responses. I may implement a policy of suspending accounts if messages go unread too long.
The moderator reserves the right to be meddlesome. For example, in the case of a decent response to an unsusual ad, if the responding user comes to me and says "What the hell?" I will try to get to the bottom of that.
If anyone wants to be a pre-alpha user, a la the Cathedral and the Bazaar's technique of build early and often, let me know. My email address is on the status page.
Once the site gets big enough, the only way to do full text search will be to index individual words and phrases. This leads to the possiblility of sorting and analyzing ads by unusual words. Perhaps we can use exisiting research to rate ads. This also allows us to build a vocabulary of fiction references and whatnot--"Lord of the Rings" is by Tolkein in the fantasy genre, etc. We can use the exact phrase searches to help build this vocabularly.
Create a Facebook app that takes interests off of FB but no personal data--no names, no friend's name, no pictures, etc. (Inspired in part by codename "Drew the Cook.")
A feature that makes it optionally easy for ads to be Google indexed. Maybe we could do some "semantic web" tagging. (It should be assumed that ads will leak onto the web, but that's different from making it easy to post to the web.)
As I said, I have the coding ability to do anything I've written about. Some things I've written will take much longer than others, of course. I also have the time (and money) to spend weeks of many-hour days on this, for free.
To keep me going at first, words of encouragement will help. Offers to be a pre-alpha user will help. Offers of other help, well, will help. Offers to help code are welcome, for example.
Eventually, I'll probably ask for donations. My current position is that I can't do that until I have a good deal to show. Maybe I can't ask until we've produced our first good date! :)
If we get to money, first priority will be to spend on expenses rather than time / salaries. We'll need more expensive hosting at some point, for one.
citations of ideas I've mentioned
Although the feedback is mostly addressed, a note to self to re-read "female feedback" at some point.