Maarten, I forgot to mention this text, which is actually the most similar to the ideas you’ve explored in this piece of all the texts I recommended (particularly with respect to the notion of measuring representations of x rather than x itself):

Michael Power — Organized Uncertainty (2007)

It’s a dense read, but Michael’s main argument is very interesting. He contends risk management is nowadays not about actually managing or reducing risk and uncertainty but rather about managing the management of risk by creating and relying upon ‘objective’ RM protocols that have little to do with risk itself and everything to do with ‘checking boxes’, which gives the appearance of ‘doing something about the problem’.