Can you trust the numbers?

Back in June, I wrote about being appointed as one of the Royal Statistical Society’s William Guy Lecturers for 2026/27. My lecture, aimed particularly at students aged 16 and over, asks a deliberately simple question: can you trust the numbers?

The full title — Can You Trust the Numbers? Statistics, Misinformation and the Information Ecosystem — reflects a set of questions that have come to sit across much of my work in recent years. We are surrounded by data, metrics, dashboards, rankings, surveys and forecasts, while also navigating an information environment in which claims travel quickly, context is easily lost, and confidence or apparent precision can be mistaken for accuracy. As I noted when the lectureship was announced, statistical literacy matters beyond its role as a technical skill: it is part of how we engage with public debate, public policy and democratic life.

As I developed and recorded the lecture, however, I found the argument moving beyond the original question. The deeper issue is how we make good judgements under uncertainty, how trustworthy information is produced, communicated and maintained, and what kind of institutions, technologies and civic capabilities we need if people are to participate confidently in a digital society.

From numbers to judgement

Statistics is sometimes understood primarily as a collection of mathematical techniques: ways of gathering data, estimating probabilities, identifying patterns and testing hypotheses. Those techniques matter enormously, but the wider contribution of statistical thinking is also about judgement. Statistics helps us reason under uncertainty.

Numbers do not make decisions by themselves. People decide what to count, how to measure it, which comparisons matter, how a result should be presented and what conclusions it can reasonably support. The same applies when we encounter an opinion poll, a graph, a news headline or an answer generated by an AI system: the important questions concern provenance, evidence, method, context and uncertainty, rather than simply whether the claim looks convincing.

This is why the answer to a difficult information environment cannot simply be to distrust everything. Cynicism is no more useful than credulity. A healthy democratic society needs capable and confident citizens who can assess competing claims, recognise uncertainty and ask what the available evidence does — and does not — support.

Seen this way, statistical thinking becomes a form of civic capability. This is close to what work on civic statistics describes: the knowledge, skills and dispositions needed to understand and critically engage with statistical evidence about important societal issues. Statistical thinking gives us shared ways of asking questions, testing claims, expressing uncertainty, disagreeing constructively and changing our minds when the evidence changes. These are not only useful statistical habits; they are important democratic ones.

Information as ecosystem and infrastructure

One idea that became central while developing the lecture was to think about our information environment as a complex socio-technical system. Information does not travel neatly from an authoritative source to a passive audience. It circulates through an ecosystem involving researchers, statisticians, governments, journalists, businesses, civil society organisations, schools and universities, technology platforms and citizens themselves. These actors influence one another, while their behaviour is shaped by institutions, incentives, technologies, professional practices, social relationships and wider cultural norms.

The quality of the information environment therefore cannot be located in any single component of that system. Deliberate fabrication is only one way information can go wrong, just as no single technical intervention can secure a trustworthy information environment. Wardle and Derakhshan’s (2017) framing of information disorder is useful here: information can mislead through poor measurement, selective reporting, missing context, inappropriate comparisons, outdated evidence, unacknowledged uncertainty or simple error; it can also change meaning as it is summarised, reframed, amplified and recombined across different systems.

AI makes some of these challenges much more visible because it changes the speed, scale and accessibility of information generation, but it did not create the underlying problem. The deeper question concerns how people exercise judgement within systems in which technological, institutional and social processes are deeply intertwined — including questions of automation and the limits of delegation.

It also connects to a question I find myself asking often: what makes a good digital society — and, perhaps more importantly, who decides? A good digital society is more than one with advanced technologies or efficient digital services; it also concerns agency, participation, accountability, public value, inclusion and the quality of the institutions and relationships through which people understand and shape the world around them.

Two metaphors became particularly important as the lecture took shape: thinking about information as an ecosystem helps us understand its complexity; thinking about trustworthy information as infrastructure helps us understand our responsibility for maintaining it. We readily recognise roads, bridges, energy networks, telecommunications and water as infrastructure. Much of this is highly regulated, with some of its most essential elements designated as part of the UK’s critical national infrastructure. These systems underpin activity across society, require investment, standards, assurance and maintenance, and often become most visible to us when something goes wrong.

Trustworthy information has similar characteristics. Official statistics allow societies to better understand themselves. The UK statistical system reflects this need for trustworthiness by design: the Office for National Statistics is the UK’s largest producer of official statistics, the independent UK Statistics Authority provides statutory oversight, and the Office for Statistics Regulation provides independent regulation through the Code of Practice for Statistics. Professional leadership is distributed across the system: the National Statistician heads the Government Statistical Service, working alongside the Chief Statisticians for Wales, Scotland and Northern Ireland and statistical Heads of Profession across government. Learned and professional bodies, such as the RSS, also provide expertise, challenge and public engagement from outside that statutory system.

Research provides ways to test competing explanations. Journalism makes evidence visible and supports democratic accountability. Education develops the capability to interpret claims. Civil society provides knowledge, relationships and trusted connections within communities. Digital technologies increasingly mediate all of these activities. None of this infrastructure maintains itself.

Trustworthiness, resilience and stewardship

The infrastructure metaphor also sharpens the distinction between trust and trustworthiness, drawing in part on the work of British philosopher Onora O’Neill. In her 2002 Reith Lectures, A Question of Trust, O’Neill argued that the response to concerns about trust should not simply be to demand that people place more of it, but to give them better grounds on which to judge whether claims, people and institutions are worthy of trust. In the language I use in the lecture: Trust is an outcome. Trustworthiness is a design property.

David Hand (2022) makes a related statistical argument in his work on the trustworthiness of statistical inference: confidence in statistical conclusions ultimately depends on the data, methods and inferential processes from which they are derived. Public trust is therefore not something institutions can simply request, still less manufacture through more persuasive communication. Trustworthiness has to be demonstrated through evidence, competence, transparency, accountability, challenge, assurance and a willingness to identify and correct mistakes. It is not a status achieved once and then retained indefinitely; it has to be maintained as systems, institutions and circumstances change.

The infrastructure framing also raises a question of resilience. A recent RUSI commentary on the evolving information-resilience landscape makes a useful distinction between responses focused on specific threats and those concerned with the wider system: the conditions that allow people to access, assess and establish trust in information and its sources despite both threats and systemic challenges. It argues that issues such as media literacy, platform design, AI, access to trustworthy information and the institutions surrounding them cannot sensibly be treated as entirely separate problems.

Deepfakes, malign influence and deliberately false or manipulated information clearly require attention. Yet information resilience also depends on the health and plurality of journalism, the capability and credibility of institutions, the strength of civil society, the availability of reliable information, and citizens having sufficient confidence and capability to navigate competing claims.

This sits alongside the UK Government’s whole-of-society approach to resilience. The Resilience Action Plan implementation report (2026) emphasises that resilience cannot be delivered by government alone, and that households, communities, businesses, civil society and public institutions all have roles in preparing for and responding to disruption. It also describes the Home Defence Programme as bringing military and civilian efforts more closely together through a whole-of-government and whole-of-society approach. The Strategic Defence Review (2025) makes a similar point, framing home defence and resilience explicitly as a whole-of-society endeavour involving industry, civil society, academia, education and communities alongside government and defence.

Information resilience, democratic resilience and home defence are distinct policy problems, but they expose a common systems insight: resilient societies depend on capabilities, relationships and trusted networks that have to exist before they are needed. For information, the timing matters. The COVID-19 pandemic provided a stark illustration of this: critical capabilities, relationships and trusted networks cannot simply be created once a crisis is under way. The ability to communicate credible evidence, maintain public confidence and enable people to make informed decisions requires foundations that are already in place. Social, cultural and information infrastructures — including strong civil society, trusted intermediaries and community networks — have to be built and sustained over time.

The DCMS Areas of Research Interest frame many of these as explicit research questions. As I wrote when they were published in May — before the July 2026 machinery of government changes — the refreshed ARIs place a stronger emphasis on understanding DCMS sectors as part of interconnected economic, cultural, technological and civic systems. They identify trust and resilience as cross-cutting themes and recognise the role of social, cultural and information infrastructures in supporting national cohesion and community capacity during periods of disruption or crisis. They also ask directly what supports a healthy and trustworthy information ecosystem and connect this to public trust, democratic participation and the resilience of the UK’s information environment.

As Chief Scientific Adviser at DCMS, these questions also connect to my interest in strengthening the wider evidence ecosystem through which government can draw on research, data and expertise from academia, industry, civil society and citizens. The ARIs themselves frame this as a shared endeavour to generate and apply knowledge for public benefit.

The same framing is a reminder that no single department, institution, profession or technology gets to decide what a good digital society looks like, nor can any one of them maintain the information infrastructure on which it depends. This is where stewardship matters: the shared responsibility to build and maintain the institutions, capabilities and relationships through which trustworthy information can be produced, challenged, corrected and used.

Ask one more question

The lecture comes back from these systems and institutions to something much more immediate. We cannot expect every citizen to investigate every statistical claim from first principles, nor should people require specialist expertise simply to participate confidently in public life. What we can encourage is a habit of disciplined curiosity: who produced this? What evidence is it based on? How do they know? What might be missing? How certain should we be? What don’t we know yet?

Their purpose is to help us become better at judging what deserves our confidence, while retaining our own agency in an environment where information is becoming ever easier to generate, manipulate, summarise and distribute. The easier it becomes to generate answers, the more important it becomes to ask good questions.

Be curious before you’re certain. Always ask one more question.

The full William Guy Lecture is below: