Microsoft Research has an interesting project along these lines, called "Penny Black".
http://en.wikipedia.org/wiki/Penny_Black_(research_project) http://research.microsoft.com/en-us/projects/PennyBlack/
It doesn't sound particularly evil to me. What's the danger, in your mind?
This is LinkedIn's business model - search for InMail on their site.
This is an interesting service to consider from the perspective of both the sender and recipient of an email message.
The value that you create for the recipient is the spam filtering that you mention only as an aside. If someone is paying to send an email message, it's very unlikely that the message is spam. This form of micropayment makes email cost a little bit to senders, but it makes spam lose its cost-effectiveness. As a recipient, you can pretty much guarantee that email coming from this service is not spam. That makes going through your inbox a lot simpler.
On the other hand, value is created for the sender of email too. Because the recipient of the email is (mostly) assured that the message isn't spam, a response is much more likely. It sounds like you have a means (or idea) to guarantee a response too, which obviously amplifies this effect.
I think the service has potential. The first question that comes to my mind is who do you charge? You definitely charge the sender, but do you charge the recipient as well? Also, how do you make this work in a user's existing inbox? It seems like having users check another inbox will decrease the effectiveness of the service.
It's interesting that your solution is sort of an incremental solution to spam. To me, the best solution to spam would be to add authentication to SMTP or a new protocol with widespread adoption. However, your solution is more realistic.