Agent Based Modelling

People sometimes ask me what I think about “agent based modeling”. As a mathematician it’s actually somewhat confusing that during my studies I learned about Markov processes, simulating queuing processes, Bayesian statistics. But it seems that the good old world of mathematics has caught up with the world of branding: genetic algorithms, neural networks, agent based modeling, etc.

So let’s put agent based modeling in the box it belongs. It’s nothing more (or less) than simulation. I’ve studies queuing processes by applying simulation models. People would come into the process, each would determine which queue to select and the time it takes to serve a person was based on some random distribution. Without me knowing it, I was applying agent based modeling.

For marketing purposes, simulation I do believe has great potential. We can research individual behavior and decision rules but a system to aggregate these individual characteristics and to allow for scenario planning based on such an aggregation would be far too complex to execute without a simulation model. But with a simulation we could have ‘agents’ seeing ads, using the internet, going shopping, etc. If current researched behavior leads to current brand outcomes, it seems like a credible step to change the environment of these agents and to see how they would react. It’s like creating a virtual environment where you can give virtual consumers stimuli and to see what and when they’ll purchase.

So what’s next? Consumers learn and change their behavior and decision rules based on historical stimuli. This means that in our simulation model we would need to allow our agents to learn. So the next big thing is “agent based artificial intelligence”! Just googled this and very disappointed to see this already gets 62,000 hits. At least I do believe I can claim to be the first to bring this idea to the marketing community.

By the way, even though I’ve studied queuing process, I too always elect the wrong queue!