“…That said, empirical work on bias must be done and reviewed with caution as it is frequently difficult to demonstrate with precision the causal link between biases and the individuals and institutions purported to be biased. Nevertheless, such research can assist in providing evidence on the extent of a problem, even if a particular bias is unproven, and inform the strategies to adopt in reform processes (Giorgetti et al 2020;Puig 2019). For example, practices such as double hatting can be descriptively mapped even if issues of bias are harder to pin down; and that mapping can prove very valuable in terms of demonstrating the extent of the problem, but also how to target reforms that might best curtail the perceptions of bias that it attracts (Langford, Behn and Lie 2017).…”