4 ms·
> So you're taking a high (100s-1000s) dimensional problem and projecting it onto a 2-10 dimensional space where your conclusions are based on a form of cluster
by mrow84 7y ago
> So you're taking a high (100s-1000s) dimensional problem and projecting it onto a 2-10 dimensional space where your conclusions are based on a form of clustering which itself is purely dependent upon your (probably ideologically driven) assumptions. In such drastic oversimplifications it is extremely east to paint whatever picture you want, even unintentionally.
This misrepresents the scientific process the researchers have undertaken - they did not proceed by simply drawing conclusions from observations of a (high-dimensional) system. Instead, they took a simple model of reality, constrained the parameters to those they believe capture the dynamics (the endowment structure), and then validated that constrained model using observations. Their assumptions are embedded in the model structure and parameter constraints.
- logicchains 7y agoIt's not a scientific process if they can't make testable predictions. I'm a quant, and anyone in this field will quickly realise it's super easy to build all kind of fancy models that look perfect when looking at the past but fail completely at predicting the future. Predictions that haven't been tested and survived falsification are no more scientific than astrology.
- mrow84 7y agoTheir testable prediction is "no matter how the Public Good game looks exactly, it’s always the case that if endowments are too unequal, you’ll never get cooperation", quoted from the article. They themselves tested that prediction with experiments, but other are of course free to replicate them.