KEYNOTE ADDRESS

Uncertainty cannot be eliminated but only managed with humility about what we don’t know and imagination for what we can’t yet see.

We live with two main kinds of uncertainty: chance- and knowledge-based. A fair coin before it is flipped is chance; once it is flipped but covered, it becomes a question of knowledge, since the outcome is already decided. In both cases, one might assume that the chance of heads is 50-50, yet assumptions matter: even simple situations rely on judgement, as shown when the “coin” turned out to be a unique coin with tails on both sides.

Words blur risk; numbers clarify it – particularly when mapped transparently. Spiegelhalter warned that phrases such as “a fair chance of success” can mislead decision-makers. Defining probabilities numerically turns vague confidence into measurable judgment, prevents overconfidence and anchors trust in evidence. For example, intelligence services now use numerical ranges for verbal terms (e.g., “likely” refers to 55-75%), ensuring consistency and accountability.

Today’s uncertainty is amplified by information ecosystems (social media/AI). The collapse of traditional news and the rise of algorithm-driven feeds have made information unpredictable and fragmented. Large language models and social media dynamics spread confident claims without conveying uncertainty, becoming just an extra noise.

Better decisions come from humility, diversity of views, and resilience, not optimisation.

Disagreements should be exposed rather than enforcing consensus. When assessing the Abbottabad raid, Obama was provided a full range of estimates by various intelligence heads rather than a single figure. Decision-makers need to see this spread to grasp the real uncertainty, recognise model risk, and counter groupthink: something best reinforced through deliberate use of red teams.

Calibration should be trained while overconfidence should be punished. Using Brier-score logic (squared-error loss) helps discipline judgement by heavily penalising confident mistakes while rewarding accurate, well-calibrated forecasts. Over time, this approach encourages analysts to express probabilities honestly and resist the temptation to sound more certain than the evidence allows.

Systems should be designed for resilience in face of “expected surprises”. In conditions of deep uncertainty where even the range of outcomes is unclear, highly optimised systems can easily fail. Real preparedness comes from maintaining flexibility and redundancy, encouraging dissenting views to test assumptions, and using imaginative scenario planning. The UK Ministry of Defence, for example, employs science-fiction writers to help broaden strategic thinking.

Risk should be communicated with trustworthiness, balance, and provisionality.

Trust must be earned, not asked for. Institutions should demonstrate trustworthiness – informing rather than persuading. This involves presenting balanced upsides and downsides, disclosing uncertainty and evidence’s quality, and pre-bunking likely misunderstandings.

Communication must fit the situation and be clear about change. In a crisis, John Krebs’ five-step approach works best: 1) explain what is known, 2) what remains unknown, 3) what actions are being taken, 4) what people can do in the meantime, and 5) make it clear that advice will evolve as new information emerges.

Uncertainty should be quantified whenever possible, but one must also acknowledge what cannot be modelled. Tools such as fan charts (which show central probability ranges while leaving a residual “other” category) help communicate limits honestly and remind decision-makers to plan for unknown extremes, which formal risk registers often overlook.

Quotes

“We’re into the era of unknown unknowns where we don’t even know what we don’t know. What do we do? We need humility to admit we don’t know things and imagination to try to think of what might happen. We’d love to try to convert the unknown unknowns into the known unknowns.”

“We live in a world of uncertainty. We do like it, and we embrace it. That’s what makes us human… Can you imagine living in a world that was certain? It’d be like the worst sort of prison”

“Question your assumptions all the time – somebody might be trying to fool you.”