I’m giving a talk at AEA 2019 in Minneapolis. The graphic summarizes the content. For the slides go here.
Deriving Value from Models: Appreciating Partial Truth, Simplification, and Multiple Uses
I’m giving a talk at AEA 2019 in Minneapolis. The graphic summarizes the content. For the slides go here.
Evaluation Uncertainty 
Kia ora Jonny
Thanks for sending me this. You make extremely important points that I hammer home at every chance I get.
My little speech is this :
A decade or so ago I attended a modelling session run by people from two of Europe’s most important modelling institutions – the University of Ghent in Belgium and the University of Sussex in the UK. They use all manner of modelling approaches from system dynamics to agent based modelling. They specialise in participatory modelling approaches – often of major public sector interventions. They stressed that there are two kinds of models. There are models for prediction and models for insight. Models for prediction are always complicated and models for insight are always simple. You will rarely have a simple model of complex situations that’s reliably predictive, and a complicated model will rarely give you insights. [With all the caveats about generally right, specifically wrong that you illustrate] Incidentally I prefer the word ‘insight’ to the word ‘explanation’.
There is considerable overlap in the concepts of system diagramming and system modelling. The Open University in the UK has a fantastic course on system diagramming, with a really helpful interactive tutorial here :
http://systems.open.ac.uk/materials/T552/
And the course material is excellent also (see attachment).
Both are freely available on the net
I was due to run a system modelling pre-conference workshop, but my injury prevents me from travelling. However, I’m filling in time by writing a workbook on system diagramming. I intend it to be a companion volume to the OU one.
>