Science is, in theory, “nice and principled”. We have the scientific method to guide experimentation, we have null hypothesis testing, so one could say nothing can really go wrong. What we get is what is “out there”. That is in theory, of course…

Coming back to good-old reality, things are not exactly perfect. First of all, we have the omnipresent sampling effect – our certainty over particular experimental results is always limited by measurement errors. Secondly, the way data is aggregated and conclusions drawn, may be biased: assumptions taken for granted, methods that are not really applicable to the given context, etc. Transparency is an issue. Finally, one of the biggest fish in the pond is the issue of scaling: drawing conclusions for large-scale systems, from small-scale observations.

From all this cloud of dust, argumentation techniques come to the rescue. They have been around for some time, especially in the domain of Safety Critical Systems; now it’s time to put them to good scientific use.

Within the CoSMoS case-study on plant ecology, efforts are currently directed towards developing structured arguments in support of scientific claims. We are particularly interested in using the Goal Structuring Notation (GSN) as a way of structuring and visually expressing arguments. There are many benefits to using this approach, ranging from the implicit fact that “a picture is worth a thousand words” to the fact that GSN has an explicit syntax and semantics. The doors towards clarity and soundness stand open.

All said and done, still this post cannot end without a tasty sample of what GSN can do. The following image represents a GSN argument obtained from Bown et al’s paper I have been studying. More information on the GSN notation and the case-study can be found here. In a few words, the diagram shows how the authors argued that their agent-based model yields results that are consistent with field observations. Starting from this main claim (G1), the authors eventually base their arguments on simulation results (Sn01) and existing literature (Sn02 and Crawley and Harral 2001, in particular).

GSN argument derived from Bown et al's paper

If you have any comments or are in doubt over what that “Crawley and Harral, 2001″ box actually contains, do let me know. We could actually stumbled upon a Pandora’s box .