The average car trip in America has 1.2 riders, including the driver. The first time I ran smack into a full-blown I35 traffic jam, and looked at the lanes of cars, minivans and SUVs filled with one person each, I thought to myself wow, if only we were all carrying three or four people, this nonsense wouldn't be happening. The trick though is that carpooling is always a bit difficult to organize with any spontaneity, and so it is an often unpopular tactic in transit.
Dynamic ridesharing systems intend to change this. Using the fairly ubiquitous technology available to users (cell phones and web services) these nascent programs enable users to call for a ride wherever they happen to be. A web service matches the user up with a driver already going in their direction. The service then gives users or riders some financial rewards, in the form of money or business discounts.
These are still in the developmental phase, but they hold a great deal of promise for facilitating a peer-to-peer transportation network, a meshed public transit system.
DynamicRidesharing.org serves as a repository for information about dynamic ridesharing and past and current projects.
Nokia also has an interesting study about the potential for dynamic ridesharing.