Make your own weather forecast

Most of us don’t have to make crucial decisions based on the BBC’s weather forecast.

On 23 August, the Met Office confirmed that it had lost its forecasting contract with the BBC – but will you miss it? Less than 10 per cent of us consider its projections on the weather to be “very accurate”. About 70 per cent of us think that they are “fairly accurate”. Does that make them useful? According to the Met Office’s surveys, 56 per cent of people think that its forecasts are “fairly useful”, while only 29 per cent find them “very useful”. Make of that what you will.

The problem is that we don’t know how best to interpret numbers. We live in an age of big data, in which information circulates in huge quantities. It turns out, however, that numbers have far less meaning to us than we like to admit. They may do little more than muddy life’s waters, bigger data bringing bigger doubts.

Most of us don’t have to make crucial decisions based on the BBC’s weather forecast. It often comes down to: “Will I need an umbrella?” or “Can I risk wearing my suede shoes?” In other areas of life, decisions informed by complex data have larger consequences – which pension fund to invest in, for instance, or how to proceed after a cancer diagnosis.

In these areas, the individual concerned usually ends up gambling on an expert’s ability to communicate relevant information clearly and without bias. That can be a bad bet. Our brains are not good at giving or receiving such information in spoken or written form. We know from myriad experiments that people make starkly different decisions when statistical information is presented in different ways. How best to proceed?

A paper recently published in the Journal of Business Research suggests a new approach to such dilemmas. Robin M Hogarth and Emre Soyer have carried out studies that show how our interpretation of data improves significantly when we live through the possible outcomes of a situation.

They started by asking 257 economists to make judgements and predictions based on a simple set of figures. The economists found the task difficult and performed poorly. Then Hogarth and Soyer gave them easy-to-use software that displayed the scenarios that might emerge from different actions. The “simulated” experience improved the economists’ performance.

According to Hogarth and Soyer, a simulated experience is useful whether or not you understand probability. Medicine could benefit from similar approaches, with software that allows patients to play with a variety of treatments, watching the outcomes and gaining a sense of what they want to pursue.

It may also be the way ahead for weather forecasting. The Met Office has been doing the hard work for us, making forecasts by interpreting an “ensemble” of scenarios. Perhaps it is missing a trick. We might prefer it if we could do some of the interpreting ourselves. A weather app could let us handle the data, allowing us to change the atmospheric pressure projections, say, at the start of the day, within the limits of what could occur, and then fast-forward through the scenarios that may follow. By immersing ourselves in the possibilities we could gain an intuitive understanding of how the day might play out and make better decisions than by listening to what a forecaster tells you about the numbers.

By doing less, rather than more, the Met Office might even improve those customer satisfaction figures. The true impact of big data might lie not in giving us locked-down certainties, but in equipping us to go with our gut.