Australia's decision on Tuesday to expel Iran's ambassador is more than a diplomatic headline; it’s a test of how democracies weigh evidence and act in the face of uncertainty.

Different parts of government operate with different standards of proof, including criminal courts, civil tribunals and national security intelligence agencies all apply their own thresholds, with distinct mathematical logics and risks. The challenge for us is about which standard is being applied, so that decisive action in defence of sovereignty does not come at the expense of public trust in justice.

Intelligence agencies often cannot afford to wait for courtroom-level proof before acting—especially in terrorism, where prevention demands intervention before an attack occurs. The absence of prosecutions, in this context, often reflects success, with threats disrupted early and quietly, long before harm materialises.

In the criminal justice system, the burden of proof is at its highest, ‘beyond a reasonable doubt.’ Courts avoid percentages, but the standard implies a confidence level of more than 90 percent, reflecting the principle that it‘s better to acquit 10 guilty people than convict one innocent. Civil and administrative law applies a lower threshold, the ‘balance of probabilities’, meaning a likelihood of 50 percent or more, because disputes over contracts, visas or entitlements cannot be held to the same evidentiary standard as crimes. Underpinning both approaches is the assumption that the world is generally transparent, punctuated only by moments of deception or concealment.

Intelligence work, however, inverts that presumption. Its starting point is that opacity is the norm, with states and non-state actors working to create uncertainty. Judgments in intelligence, therefore, begin from a very different basis, where moments of clarity are rare, fleeting and contested.

Analysts rarely have the luxury of certainty, let alone the proof needed for a criminal conviction. Instead, they deal in probabilities, not proofs. Their calibrated terms, such as ‘likely’, ‘highly probable’ and or ‘remote chance’, reflect a form of statistical reasoning (called Bayesian) that adjusts as fragments of data accumulate. A ‘highly likely’ judgment might equate to 70 to 80 percent confidence.

Over the past several decades, agencies and expert communities have grappled with the challenge that such non-mathematical terms have very different meanings to different audiences. That recognition has driven a deliberate move towards explicit numbers-based language and a more disciplined lexicon to reduce ambiguity and prevent misinterpretation.

However the assessed probability is expressed, it is unlikely to be high enough to persuade a jury, but it is more than enough to justify government action when the cost of inaction outweighs risky inaction.

The expulsion of the Iranian ambassador illustrates this calculus. If Iranian officials intimidate Australians, that is a crime. In court, prosecutors would have to prove it beyond a reasonable doubt. But in terms of protecting sovereignty and security, and signalling to Tehran, waiting for a courtroom verdict would be reckless. When intelligence assessments indicate a strong probability of state-directed intimidation, and the balance of probabilities supports that conclusion, decisive diplomatic action becomes not only justified but essential.

That contrast is clear when set against the Australian Federal Police’s recent use of criminal law to arrest and charge an individual with foreign interference. In that context, investigators will need to marshal admissible evidence—documents, communications, witness testimony—that can withstand the scrutiny of open court and persuade a jury beyond reasonable doubt. The higher bar is proper: the criminal law carries the gravest consequences, including deprivation of liberty.

By comparison, diplomatic expulsions or administrative sanctions rely on intelligence assessments and probabilities, enabling the government to act quickly to protect sovereignty even where courtroom-standard proof is not yet available.

Each of these standards embodies a value. The criminal law bar protects liberty. The civil law threshold strikes a balance between efficiency and fairness. Intelligence estimates acknowledge uncertainty to protect the nation from surprise and harm. Taken together, they form an ecosystem of decision-making that ensures both justice and security.

The Iranian case shows why this discipline matters. Democracies need to be able to act on intelligence-driven probabilities to defend sovereignty while still upholding the highest burden of proof in the courts. The maths is different, the language is different, but the end is the same: safeguarding both justice and security in an increasingly uncertain world.