The way that we use it today for example. Many people have inbox full of emails that probably they would not read or attend properly. This is behavioural, not a technical issue but it is an example of why we need other collaborative models. There is a good article about it in the IEEE magazine. In the technical and security side, quite a lot of traffic between SMTP servers is in clear text. Identity is very easy to spoof. And so on ... But as I said, there isn't anything better yet.

an inbox full of emails that you do not read is a behavorial problem. Not one of email. No amount of twitter, facebook or "email 2.0" is going to fix that.

Even though the traffic between SMTP servers is in plaintext, encrypting that traffic does not solve the underlying problem. Trust on the Internet is a very very hard problem that is not going to be solved any time soon.

Trust on the Internet has a solution. SSL with proper CAs. Then the problem atleast shifts from Internet to real world.

The problem today are the cost of the infrastructure to support CAs (this may be a minor problem) and usability (I think this is the most difficult to address). I cannot imagine many of my FB friends using PGP (or GnuPG), public and private keys.

They wouldn't need to. In simple cases like facebook, they would use some CA signed certificate (which basically says that they are who they claim to be) to sign in to facebook via ssl/https (this could be easily automatized in browser).

Yes but then you need a certificate to store (your private key). If you are in the same computer it is not a problem, if you move to another one, then you need to take the certificate with you. I think the problem is not as easy if you want proper security mechanisms.