« Resigning Myself to Vista | Home | The Role of Universities - Knowledge, Skills, Citizenship »

Behaviour Modelling: Did it Go Out With A Whimper?

By crooky | June 25, 2008

An interesting article in Wired this month talks about how humanity is entering into the “Petabyte Age”. They describe the Petabyte Age as a time where there are sensors everywhere, data pouring in from myriad sources, unlimited data storage capacity and the processing power necessary to fill in the gaps. This article goes on to proclaim the “scientific method obsolete”. I want to argue for and against this idea.

Modelling, like Economics, assumes predictable, rational behaviour in all aspects of our universe. I’ve always had a beef with this supposition - especially on the economics side of the fence. If economic theory and models really worked, we’d have a utopia based on capitalism and the free market. These kinds of models and the political systems based on thse models fall down when they meet with reality because people do not behave rationally. In my experience, people behave irrationally. By extension, patterns in nature are equally as unpredictable.

On the other side of the argument is the fact that a model is only as good as the observations that form the basis of that model. If the observations are fouled, the model is fouled. Given an unlimited number of observations and validation points, you could build a model that could predict any behaviour. At that point - are we talking about a model anymore or omniscience?

That is what the Wired article is suggesting - that our modern ability to collect, store and analyze data about the world around us has made models irrelevant. We can measure and describe almost complete systems now. When you have a comprehensive description of a system, you don’t need a model because you have all of the data.

Models were originally designed to fill in holes in our ability to measure and analyze. Statistics helped us validate the accuracy of our models. The scientific method works as follows:

1. Define the question
2. Gather information and resources
3. Form a hypothesis
4. Collect the data
5. Analyze the data
6. Draw your conclusions and answer your question based on your analyses

The Petabyte Age allows researchers to skip step 3 - the point where models are developed. You can now just observe the data as it is and analyze the system that is in place. Modelling and hypotheses are important when you have limited resources and ability to collect data. When these limitations are taken out of the picture - are hypotheses really necessary? Wired says no.

My conclusion is that you still need to have some concept of what drives a system or motivations individuals in order to analyze even a comprehensive set of data. Otherwise, there is no patterns to guide our decision-making process. You could just say “well, this is how group A behaves” and when someone asks you to guess how Group B is going to behave, you say “we have to analyze their metrics before I can tell you that”.

Models and hypotheses form the foundation of wisdom - insight that can be applied to like situations without extensive data collection and analysis. Without wisdom, we’re all just fumbling around in the dark, hoping to find inspiration. Innovation in the research community comes from drawing on the wisdom of others - wisdom that is often embodied in models and hypotheses.

*********************
Aaron “Crooky” Cruikshank is the Principal and Founder of Friuch Consulting. He has written professionally about science and technology for ten years.

Share/Save/Bookmark

Topics: Research Methodologies |

Comments