"Mark abuse" feature may be subject to abuse itself
In your FAQ, you wrote "By clicking 'mark abuse', you are making sure thousands of other users will never see the spam you just marked."
This gives any Mollom user the power to suppress content that may or may not actually be spam -- not only on their own site but across the entire Mollom network.
Such abuse of the "mark abuse" feature could be intentional, but it could also be the unintentional effect of someone who simply doesn't distinguish between spam and other (non-spam) unwanted content such as unpopular political views.
How do your systems prevent this kind of censorship?
Will censored posters be informed that their content has been blocked? If so, will they have a means to rectify inappropriate censorship and blacklisting?
What if someone wants to have a legitimate discussion about Viagra or pornography? Those are words I don't feel free to use in emails because today's email spam filters would likely block them. In other words, email anti-spam systems have effectively undermined free speech -- I cannot express myself freely without wondering if certain words are going to trigger a filter and prevent my message from being delivered to the recipients' Inboxes.
The major email service providers have failed to provide adequate means to prevent inappropriate censorship ("false positives"), and I would like to know how Mollom is addressing that problem, particularly when your system is "sure" that a particular post is spam.
It will take more than one
It will take more than one person marking a comment as spam for blocking this specific post from all other sites. Mollom is based on statistical learning methods and on a very large sets of training data. So only if a comment is "generally assumed" as spam, it will be blocked. The more people marking the comment, the more sure Mollom will get it is spam.
Second, this blocking occurs in two steps. Only the hard-core spam (of which we know for sure that it is spam) will directly be blocked. All others posts will be marked as unsure, meaning that we cannot be 100% sure that it is spam. In those cases the user will have to fill in a CAPTCHA to authenticate being a real human before the comment can be accepted. More on this can be found in the How Mollom works section.
So if your users want to have a discussion about Viagra, they are free to do so, but might have to authenticate their human-ness. So Mollom fully allows free speech, between humans :)
I hope this clarifies your questions.
Can a site administrator have
Can a site administrator have a last word about a blocked content?
I mean - if Mollom blocks content, can I have access to it, to review it and maybe change the decision?
Blocked user content is not
Blocked user content is not accepted to the site and can not be accessed afterwards. Other spam blocking services require the use of a moderation queue in which these comments are stored, but since this moderation queue is actually filled with only spam, it has no real purpose since finding the lost ham post in it is practically impossible.
The Mollom approach is to never block human content by way of a CAPTCHA authentication. This way a moderation queue is not needed. In the unlikely event that a spam message gets through, you can simply hit the "delete" or "mark as abuse" link under the comment to notify Mollom of the misclassification.
So moderating mollom's
So moderating mollom's actions is actually done by the user himself. Excellent.