Will humans bid farewell to Internet auctions?
The explosive and unrelenting growth of the Internet has led to a huge increase in electronic trade. The popular e-commerce site eBay alone sells $1,400-worth of goods – every second. But computer scientists, such as Loughborough’s Dr Syeda Fatima, believe the potential of the Internet to truly transform commerce is still largely unrealised. Her innovative idea is to replace human traders with so-called ‘software agents’, that could bid and bargain on our behalf.
Alison Laing met Dr Fatima to discuss the motivation, objectives and key challenges of this unique research project.
A human buyer using popular e-commerce websites such as eBay or Amazon, may find they offer advantages over manual trading. They are quick, convenient, and save a trip to the shops. But these systems require considerable human intervention, during the trading process, to research information on the merchants and products, to make decisions on what product to buy, and who to buy it from, and to negotiate the terms and conditions of trade.
Fact file
- Total number of Internet users in the world is 1,668,870,408 (according to ‘Internet World Stats’).
- Almost a quarter of the entire world population - 24.7% - use the Internet (according to ‘Internet World Stats’).
- Between the years 2000 and 2009 the growth in the number of Internet users around the world was 362.3%.
- $1,400-worth of goods sold every second on eBay (provided by eBay research labs).
- iPods, iPhones and digital cameras are the most popular products sought after on eBay.
Dr Syeda Fatima of the Department of Computer Science believes having a human perform these tasks takes too much time and adds to the transaction costs. In a £250k, three-year project, funded by the Engineering and Physical Sciences Research Council (EPSRC), she wants to take humans ‘out of the loop’ and develop software agents instead, to automate several of the most time consuming stages of the trading process.
Dr Fatima said: “Unlike traditional software, these agents will be personalised, continuously running, and able to decide and act optimally on behalf of their users to find the best possible deal for them.
As electronic markets grow, it will become increasingly difficult for humans to devise effective negotiation strategies in order to get the best deals. Our research is aimed at overcoming this problem by automating the negotiation stage.
Simple agents already exist. If you look at eBay, for example, agents do the initial bidding and then monitor the auction over hours, or even days, on behalf of the buyer. Dr Fatima and her team want to develop more sophisticated versions, that can negotiate in a variety of settings and in a range of auction types. Agents that don’t just negotiate on price, but on other attributes such as quality of service, delivery time, or method of payment.
However, the problem becomes very complex as you add more attributes for agents to negotiate. Even moving from one to two attributes can be complicated. As further attributes are introduced, the space of possible deals increases exponentially.
Agents must also be equipped with the strategic ability to collaborate and compete with each other. Developing programs for the Internet is very different from conventional software development. The Internet is unique among all computer systems because it is built, operated, and used by a multitude of diverse economic interests, in varying relationships of collaboration and competition with each other.
On eBay, the bidders collaborate with each other by following a commonly agreed auction protocol, while at the same time they must compete with each other in devising a bidding strategy that will strike a deal at the best possible price.
Dr Fatima explained: “One of the key problems is to find optimal bidding strategies that maximise bidder profit. The challenges are that strategy space is huge and we need very fast methods to search for optimal strategies.
“Then, in addition to finding optimal bidding strategies in a variety of settings with several attributes to negotiate, we have the added complication that the scenario is very competitive. There are several agents, or bidders, in the auction each trying to maximise their profits and each bidder’s optimal strategy depends on the other bidders’ strategies.
“In this context, game theory provides an elegant mathematical formalism for modeling and analysing strategic interactions between profit maximising entities. We are using game theory as a guide in designing negotiating agents.”
Dr Fatima also wants these software agents to be able to dynamically create business partnerships, as and when they are needed. An individual could therefore give the agent a list of things to buy, and the then agent figures out the best business partners, and initiates negotiations with them.
An example could be asking an agent to organise a trip to a foreign destination to watch an international cricket match.
The agent would first plan three negotiations, to get flight tickets, a hotel room, and sports tickets. Next, the agent would find the appropriate service providers, such as a travel agent, hotel agent, and cricket agent. And finally, it would negotiate the best prices with each of them.
Dr Fatima said they were still in the early stages of exploring all these methods, however she’s convinced the research will prove of benefit to business and individuals, saving them money and time.
She said: “In the future, software agents will do the shopping for us. They will get the best possible deals for us and save us money and time that we could use to deal with other important things in life.”
Want to know more?
- Website: www.lboro.ac.uk/departments/co
- Contact: Dr Syeda Fatima
- Email: S.S.Fatima@lboro.ac.uk
Download this issue in Adobe PDF format