Why the best way to use data modelling is to understand its limits
Mathematical models shape our reaction to everything from global health to climate change. But as Erica Thompson, author of Escape from Model Land, tells Philippa Nuttall, the problem is they don’t always lead to the best real-world decisions
ESCAPE FROM MODEL LAND: HOW MATHEMATICAL MODELS CAN LEAD US ASTRAY AND WHAT WE CAN DO ABOUT IT
Author: Erica Thompson Publisher: Basic Books
The outcome of COP27 last month was criticised by many, including European Union Climate Commissioner Frans Timmermans, as too weak to keep global heating below dangerous levels. This conclusion was based on our understanding of mathematical models, which give us an insight into how the future is likely to pan out.
These models can, however, lead us astray, argues Erica Thompson in Escape from
Model Land. Her eminently readable and surprisingly humorous book highlights the failures of modelling today and how she believes models can be improved to enable policymakers to make the best real-world decisions.
Data is the beating heart of the modern world. Everywhere we go and whatever we do, we are bombarded with statistics or requests that can be turned into data. Did you enjoy your train ride? How was your hotel room? Did we deliver your post well?
Models “add relationships between data”, writes Thompson,
“Computer models can’t take into account our values or changes in the way people think”
a senior policy fellow in the London School of Economics’ Data Science Institute. “The purpose of modelling relationships between data is to try to predict how we can take more effective actions in the future in support of some overall goal.”
Models are not, however, a magic tool free from subjective, even unconsciously subjective, pressures. They are influenced by the humans who created them.
Models “encapsulate our imagination about the future”, Thompson tells The Parliament over Zoom. “There are questions about the degree to which our imagination limits the models and the degree to which the models limit our imagination.” This relationship “feeds in both directions”.
With this in mind, policymakers shouldn’t allow their thinking to be limited by what comes out of these models, says Thompson: “Computer models can’t take into account our values or changes in the way people think.”
Politicians should be bolder and “push in a particular direction if they want to”, she adds. “They don’t have to sit around and wait for a prediction about 2100 to come true. By acting in the world and representing different things in models, we create the possibility of different outcomes.”
This need for action and engagement, rather than a passive acceptance of a certain future, is central to successful policymaking around climate
change. Thompson agrees with the oft-repeated analysis that humanity’s failure to manage climate change is largely a crisis of imagination.
“Simply swapping fossil fuels out of the electricity system is deeply unimaginative,” says Thompson. “Surely we can do better than that?” If we are simply producing models based on business-as-usual where solar and wind have replaced gas and oil, “this is a political statement; a political endorsement of the status quo”.
Escape from Model Land sets out “five principles for responsible modelling”, key to which is the call to involve a wider range of people. “To make models that genuinely represent a range of possible outcomes, we need a range of perspectives,” Thompson explains. “The best way to do that is to have a diverse range of people working on them.”
Six models can’t be considered as “six independent throws at a dartboard because they have all probably been created by white middle-class men who went to Oxbridge and did a science degree”, she says. “You can’t call that a statistically independent sample.”
Ensuring that models on the same subject are carried out in different ways and by people from different backgrounds and of different
“To make models that genuinely represent a range of possible outcomes, we need a range of perspectives”
gender and age would also raise some “really interesting questions” about who is considered an expert, whose opinions are allowed to matter and what qualifications modellers should have, says Thompson.
To work on climate modelling, “do you have to be an expert in atmospheric physics or energy systems?” she asks. “This is a really fruitful debate and gets to the heart of gatekeeping and what it looks like in terms of science and the construction of models.”
Clearly defining who is an expert is particularly pertinent for climate change, where things are changing rapidly and “the data we gathered in the past are not necessarily a useful guide to the future”, Thompson explains. In such cases, “we have to run models and ask, ‘Does it look plausible to me as an expert?’ and, if not, go back and recalibrate”.
“Models help us think through the consequences of our assumptions,” concludes Thompson. “They act like a prosthesis of our brain, but we have to be careful not to ascribe assumptions to them and not to think that, because they are written in mathematical language, they must be the truth. This is still your opinion.”
As our conversation winds down, Thompson reiterates the need for policymakers to focus less on models and predictions, and more on “human values and how to integrate them into decision-making”.
Climate models are clear that if we don’t reduce emissions now, the future looks grim. Instead of pressing this point, politicians should be making decisions to avoid the worst impacts of climate change, says Thompson. “There is no point just waiting for it to happen. The best way to predict the future is to create it.”