Summary and discussion Session 7: Applications in Electronic Commerce Michael Wellman University of Michigan Slides: PowerPoint (235K)

Relevant links

Summary

What is electronic commerce?
Really just computationally automated trade

Three main segments of electronic commerce:

• Find - how to find the good and vendor?
• Bid - how to agree on a price?
• Exchange - how to deliver the good and payment safely?

Authentication and adjudication become real problems in last two phases.

Exchange

• Payment initiated electronically, but most often goes outside the electronic world for both delivery (Fed Ex, UPS) and payment (Visa, AmEx).
• The problem with payment and delivery systems is that there is a network externality - the values of such systems are directly related to the number of people using them.
• Very difficult to come up with a universal payment method/media.
• UPAI at Stanford is a wrapper around many payment methods to create one abstraction. A problem in that you either have to use lowest common denominator or implement features on your own.
• NetBill is a system for buying electronic documents over the web. It validates transfer of goods through atomic transactions.

Auction systems

• Auction is a generic word for market.
• AuctionBot is an online system for web users to buy goods using auctions.
• Security concerns:
  1. Make sure your agent gets to bid - may want to submit programmed strategy.
  2. Regulate resource consumption in submitted agents.
  3. Make sure bids are honest.
  4. Make sure billing is correct.
  

The find phase

• Searching documents is a semi-structured problem.
  Most everyone uses HTML and many HTML forms have the same structure. Many things differ, however.
• Example - Jango finds reviews and other product information and prices from web documents.
  There are applications that run on top of Jango to process results. Jango runs with your identity to prevent vendors from discriminating. This makes competition tighter.
• How to specify fuzzy categories?
  Always a trade off between the number of sites visited and the amount of information gathered at any one site.
• Eventually agent programs could gain enough popularity that vendors would be forced to deal with them.
• How to make agents deal with bundling of goods?
• How to deal with privacy and protection of bargaining position?