When the data isn't real but the decisions are.

Synthetic data is having a moment—those artificially generated datasets that promise to solve our privacy concerns whilst maintaining statistical validity. It's seductive, isn't it? The idea that we can create perfect, privacy-compliant data that mirrors reality without exposing anyone's actual information.

Here's what keeps me up at night: when we remove the messiness of real human experience from our datasets, what stories do we lose? What nuance evaporates in the translation from lived experience to statistical approximation?

This tension between data protection and authentic understanding isn't just theoretical. In our recent evaluation of court dogs in Victoria's criminal justice system, we witnessed how the most profound impacts—a child finally able to testify, a vulnerable witness finding courage through a gentle presence—resist neat quantification. These moments matter precisely because they're irreducibly human.

As researchers and designers, we're constantly navigating these ethical territories. That's also why we've just shared our internal framework for deciding on research incentives—because transparency about our methods is part of how we hold ourselves accountable to the real people behind the data and decisions.

Let's not mistake computational convenience for human truth. The stories that matter most are often the ones that refuse to be tidied into perfect datasets.