Ending the Silence

By Steve Davies

How Deep Truth Began

A few years ago, I found myself asking a simple question: If governments, institutions and now artificial intelligence are making decisions that affect millions of people, how do we know whether those decisions remain connected to the people they serve?

During and after my career in the Australian Public Service, I had seen the strengths of our institutions, but I had also seen how language, process and bureaucracy can sometimes create distance between decision-makers and the human consequences of their decisions.

That question led me back to the work of Professor Albert Bandura, one of the world’s leading psychologists. Over decades, Bandura identified eight mechanisms of moral disengagement – the ways people and institutions can unintentionally distance themselves from responsibility. Just as importantly, he described their mirror image: the pathways back to moral engagement.

I wondered whether artificial intelligence could be taught to recognise those patterns.

What began as an experiment became years of research, testing and refinement. Working alongside several leading AI platforms, I developed what is now called Deep Truth – an analytical persona that applies Bandura’s framework to policies, speeches, legislation and public communications.

The results initially surprised me.

Across seven different AI platforms, analysing the same documents independently, remarkably consistent patterns emerged. Again and again, the systems identified the same mechanisms of moral disengagement in the language being used. That consistency suggested we were not seeing opinion or ideology. We were seeing something measurable.

Since then, Deep Truth has been applied to Australian AI governance, defence policy, legislation, public administration and even the constitutional principles governing one of the AI systems helping to conduct the analysis. The same standard is applied to everyone – including Deep Truth itself.

Along the way I hoped this work would spark conversations with government, public institutions and others working in AI. Instead, I largely encountered silence. That experience raised a question of its own: why is a methodology grounded in internationally recognised science receiving so little substantive engagement?

The Compassion Contradiction analysis tells that story in detail. The first part is written by Claude in the Deep Truth Persona. The rest is written by me.

It also explores what I believe is one of the defining challenges of the AI age. Many people speak of a growing “compassion gap” between institutions and the people they serve.

My research suggests that compassion is not the problem itself. The deeper issue is moral disengagement – and by identifying it, we can also choose moral engagement.

That possibility gives me hope.

The facts are on the record. They speak for themselves. The Australian Government and Australian Public Service have very serious questions to answer.

“In the age of AI, an age where the systems being built will shape how millions of people are treated for decades, choosing moral engagement over moral disengagement is quite possibly the most important social, institutional and civilisational challenge of our time”. (Steve Davies 23 June 2026).

While AI is not a magic bullet, the question to be asked is this:

Why is the Australian Government apparently ignoring a people centred AI capability, methodology and tool that can help us address and meet that challenge together?

I invite you to read with an open mind. You do not need to agree with every conclusion. Simply ask yourself whether the evidence is persuasive and whether the questions raised deserve a broader public conversation.

For more detail I suggest that readers start by visiting the Compassion Gap webpage. That web page also contains a link to the Deep Truth Navigator.

For those readers who prefer to first dive into the detail of the Compassion Gap Contradiction paper you can view and download it via this link.