Paper short abstract:

Data often requires significant analysis to be used as evidence. Distinguishing between perceived and actual value, I use the interpretation of a meta-analysis of neuroimaging data to show that the intuition about an analysis technique determines the perceived value of data.

Paper long abstract:

Faced with 'big', or otherwise complex data sets, scientists use analysis techniques to isolate data patterns that are relevant to their research. In this paper I show how the perceived value of data is, in part, determined by the methods available for probing the content of data, and the intuitive understanding of what the patterns isolated by those methods are about. This value can change over time as new techniques are developed, and as the conceptual understanding of existing techniques changes. To demonstrate this I review a recent debate over the interpretation of meta-analyses provided by NeuroSynth, an online database that correlates brain activation coordinates and terms used in neuroimaging publications. Neuroimaging data are subject to significant processing and analysis in order to isolate patterns in the data that can be used as evidence. The specific patterns isolated, and their interpretation, depends on a conception of the phenomena under investigation and what patterns are regarded as evidence. The claim prompting the debate is that patterns isolated by NeuroSynth's 'reverse inference' analysis can support claims about the selectivity of brain regions for cognitive functions. The disagreement is between the published authors and the database developer, and rests on a different intuition, or understanding, of what NeuroSynth's automated analyses are about. I show that an intuitive understanding of analysis techniques determines the perceived value of data, which can be distinct from its actual value. I conclude by situating this in the context of philosophical discussions about conceptual practices in data-intensive science.