Causal machine learning tools are beginning to see use in real-world policy evaluation tasks to flexibly estimate treatment effects. One issue with these methods is that the machine learning models used are generally black boxes, that is, there is no globally interpretable way to understand how a model makes estimates. This is a clear problem for governments who want to evaluate policy as it is difficult to understand whether such models are functioning in ways that are fair, based on the correct interpretation of evidence and transparent enough to allow for accountability if things go wrong. However, there has been little discussion of transparency problems in the causal machine learning literature and how these might be overcome. This article explores why transparency issues are a problem for causal machine learning in public policy evaluation applications and considers ways these problems might be addressed through explainable AI tools and by simplifying models in line with interpretable AI principles. It then applies these ideas to a case study using a causal forest model to estimate conditional average treatment effects for returns on education study. It shows that existing tools for understanding black-box predictive models are not as well suited to causal machine learning and that simplifying the model to make it interpretable leads to an unacceptable increase in error (in this application). It concludes that new tools are needed to properly understand causal machine learning models and the algorithms that fit them.

Relative to the rest of the population, Indigenous Australians are disadvantaged across a number of outcomes. Not only is income lower, but so is life expectancy. Furthermore, the incidence of a number of chronic conditions is much higher (especially after controlling for the younger age profile). There are a number of interrelated historic, social and economic reasons for these poorer outcomes, but one important factor is likely to be substantially lower education levels. That is, Indigenous Australians are much less likely to finish high school, and for those that do, less likely go on to university or other studies. This paper looks in an empirical way at the relationship between education and health for Indigenous Australians, comparing it to similar associations for the non-Indigenous population. Using probit model estimates on data from the 2001 National Health Survey, two traditional health outcomes (self-assessed health and chronic conditions) as well as a set of health risk-factors are examined. For all health measures used, those with lower education levels have significantly worse health outcomes. Interestingly, education is found to have a significantly different magnitude of association for some measures across Indigenous and non-Indigenous Australians.

In this article, we focus on data trust and data privacy, and how attitudes may be changing during the COVID-19 period. On balance, it appears that Australians are more trusting of organizations with regards to data privacy and less concerned about their own personal information and data than they were prior to the spread of COVID-19. The major determinant of this change in trust with regards to data was changes in general confidence in government institutions. Despite this improvement in trust with regards to data privacy, trust levels are still low.