Trading

Amongst many procrastinations, trips to the pub and baths I've been working surprisingly hard on my Final Year Project of late; a Christmas idly spent will encourage guilt that way. I've changed from a civil violence scenario investigating artificial intelligence, through a multi-agent approach to an investigation into trust and reputation in an e-commerce scenario.

Don't Panic.

Let's explain..

Imagine, eBay. You want to buy yourself some electronic gadget. Everyone, co-incidentally, happens to be charging the exact same amount for this gadget - but you still need to decide who to trade with. Here you'll be looking at the reputation of users - as shown through eBay's central reputation system.

Now imagine Craigslist, GumTree, et al. There's no central rating system, instead you need to get at someone's reputation through a distributed set of other users. You'll likely ask people how trustworthy another user is.

My investigation is into which model works better, based on a variety of factors. For example, changing behaviours, communication trustworthiness (see below) and 'noise' in the system (making it hard to see what's actually happening).

A problem that crops up with both of these scenarios is that of trust. Note that trust is your perception of how likely someone is to be good/bad at a specific task. How much can I trust eBay's central server? What if people have been feeding it bad information? What if I can't trust the people who are telling me about other users on Craigslist?

--

I'm doing research into all of this, and it's getting pretty fun. The maths, however, is giving me some headaches - I don't like to see equations taking up half a page and filled with Greek. I think I'm over most of those hurdles, and I've been offered the chance to get some funding to stay on at Uni over the summer and put this together into a proper research paper, for publication in one of the fancier journals in this area of Computer Science.

Exciting eh? IM me if you'd like more detail on any of it :)