Uncertainty and Probability: Lessons from a Lifetime of Public Engagement and Decision Making

Short Summary

This talk delves into the nature of uncertainty and how numerical probabilities frame our understanding of the world. Using accessible demonstrations, the speaker distinguishes aleatory uncertainty from epistemic uncertainty, and shows how our assumptions shape what we think we know. Through vivid anecdotes—from a rigged coin and the Bay of Pigs to the COVID-19 pandemic and public science communication—the talk argues for humility in statistics, the value of independent analyses, and the critical role of transparent uncertainty in high-stakes decisions. It also touches on luck, coincidences, and the art of communicating science to the public.

Introduction to Uncertainty

The speaker opens by recounting a long career in science and public engagement, emphasizing a shift from pursuing traditional research to communicating science to the public. The central theme is uncertainty, defined as a relationship between an observer and the outside world, and as something personally owned rather than an intrinsic property of the world itself. The talk stresses that even when the world is uncertain, our knowledge about it is mediated by our assumptions and models.

Aleatory vs Epistemic Uncertainty

Using a coin example, the speaker distinguishes aleatory uncertainty (unpredictable outcomes before observation) from epistemic uncertainty (lack of knowledge about which outcome will occur). The exercise demonstrates how probabilities depend on what is known or unknown, and how revealing information (uncovering a two-headed coin) can abruptly change one’s inferred probability.

The Dangers of Misleading Language

The Bay of Pigs episode is used to illustrate how words like likely, possible, and probable can be misinterpreted, leading to overconfidence or misinformed decisions. The UK intelligence community’s practice of defining terms such as likely (55-75% probability) is highlighted as a corrective measure for clearer communication of uncertainty.

Lessons from COVID-19 Modelling

The talk recounts the UK’s COVID-19 response, where eight modelling groups produced twelve models estimating the reproduction number R. The key lesson is that model assumptions drive the width of uncertainty and that presenting a single number can be dangerously misleading. A composite or ensemble approach, along with transparent sharing of individual models, provides a more robust representation of uncertainty.

The Value of Probabilities

Probabilities, even when subjective, can be valuable in decision making. The speaker details a classroom exercise where participants assign confidence levels to answers, illustrating how explicit probabilistic thinking can reveal ignorance and improve judgment. The activity also introduces the concept of proper scoring rules, such as squared-error loss, to incentivize honesty about certainty.

Luck, Coincidence, and Probability

Three types of luck are discussed: constitutive luck (who you are in life), circumstantial luck (where you are), and outcome luck (how things turn out). The speaker uses historical anecdotes, including a survivor of a fatal airplane crash and a famous illusionist’s coin trick, to show how luck emerges from patterns of chance and human interpretation. The birthday problem and the classic “two people share a birthday” paradox are used to illustrate how intuitive judgments about probability can be misleading, and how simple approximations via Poisson or exponential models can illuminate our reasoning about rare events.

Public Engagement and Science Policy

The talk closes by reflecting on public inquiries and mandated science scenarios, where scientists are sometimes required to provide answers despite low confidence. The speaker argues that expressing uncertainty, humility, and transparent confidence is a powerful and liberating practice in public discourse and policy formation. The overall message is that probabilities, even when imperfect, help structure reasoning, communicate risk, and support better collective decision making.