Abstract
It is well known in impact assessment that predictive model outputs will be as credible as their inputs and that model assumptions will drive outputs. What is less well known is how the practice of integrated impact assessment with its pervasive use of predictive computer models and multiple teams of consultants can influence evidence relied upon in deliberations over the impacts and benefits of major projects. This paper draws on an integrated impact assessment of a major energy infrastructure project in Australia known as Basslink to examine the epistemic implications of current practice. It will be argued that what has become standard procedure can serve to diminish the disclosure of prediction uncertainty.