Thread by @greg_m_priest

Next, Alisa Bokulich on the model-ladenness of data. Starts with Paul Edwards idea of model-data symbiosis, particularly the idea that data is model-filtered. Focus on geosciences. #hss18 #psa2018

A taxonomy of the model-ladenness of data. 1. Data conversion: Measured quantity is a proxy for phenomenon of interest. 2. Data correction: data gathered in field needs to be cleaned of noise or confounding factors to yield signal. #hss18 #psa2018

3. Data interpolation: often data has spatial or temporal gaps that need to be filled in in ways often mediated by the model. 4. Data scaling: you may gather data at scales different from the scale of interest, so it has to be rescaled. #hss18 #psa2018

5. Data integration: need to combine different sources of information about a phenomenon of interest into a skinfle integrated dataset. 6. Data assimilation: combining information from measurements with data from dynamical models. #hss18 #psa2018

7. Synthetic data: artificial data created by a simulation model, often used to test the various data processing methods used with the previous 6 categories. In all 7 cases, models are deeply implicated in geoscience datasets. #hss18 #psa2018

This is not to be understood as a corruption of the data. It is how the data becomes useful, even intelligible. #hss18 #psa2018

@threadreaderapp unroll plz