CASA Conference 8th - 9th January 2009 - DAY 2

Day 2 of the conference concentrated on a number of different projects involved in S4 program, from Leeds University and CASA. This day was aiming to be a bit more of a workshop day. Unfortunately I only managed to catch the afternoon sessions.

There were some v. interesting papers - particularly to do with Multi-Agent Modelling. Alison Heppenstall discussed the creation of individual consumer agents to model a local communities shopping habits. She also briefly discussed Nick Malleson's crime work - which looks really useful. He is using the PECS model, which seems to be a pretty robust way of thinking about agent behaviour.

This paper was followed by one delivered by Vassilis Zacharidias from CASA on modelling queing, whilst it was a bit diffiuclt to understand at points (lots of maths!) the resultant model of the shambolic queing in Victoria Station looked pretty accurate - well at least everytime I'm there I walk round in circles next to the escalators!

The highlight of the day for me was Andrew Hudson-Smith's slightly tongue in cheek discussion on how to visualise ABM with 3D. It's well worth downloading his paper (in fact all of the presentations from the 2 days) here. Andrew's half-hour paper went through his journey from knowing nothing about ABM to NetLogo to reinventing the wheel in 3D StudioMax to finally coming back round to NetLogo - but linking the code (and graphs!) to both 3DSMax and Second Life. This gives you decent-looking 3D models which have a very strong grounding in a powerful ABM model builder which gives feedback and can be calibrated in lots of different ways.

This is something that I am very interested in - especially in the use of gaming-engines (with their in-built rendering and real-world physics models) to extend and expand the normal ABM world. If hundreds of people have already spent a lot of time sorting out making sure computer-based avatars/agents/soldiers/aliens/whatever don't walk into walls and obey the forces of mass and gravity then that's great and lets not reinvent more wheels.

One of the best points that Andrew made was that by doing this stuff in Second Life - the modelling process can become colloborative, with people running the models at different times etc. and everyone tweaking as they see fit. But what really made me stop and think was that suddenly we can now have our avatars (REAL people from anywhere in the world at anytime) logging into the models and directly interacting with the 'dumb' agents themselves. It really does bring an interesting dimension to how we should be using ABMs and also what they can tell us about how systems will react to outside influence and to those 'rogue' situations of human agency that are just so hard to recreate when you are using generic agents. Anyway, a lot of food for thought and definately someone I need to contact and discuss this all further with. He said that CASA are just at the tip of the iceberg with thinking through all of this, and I really think it has massive potential for my research.