Famed for the beauty of his economic models, the Nobel Prize-winning economist Paul Krugman once reflected that “there’s a pretty good case to be made that the stuff that I stressed in the models is a less important story than the things I left out because I couldn’t model them”.

It’s a casually explosive comment, because we use models all the time. Designed to reduce the world’s complexity to a manageable state, business models, economic models, scientific models are tools with which we test out our hypotheses and decisions.

But their simplification and utility is a trap. Because they must leave out so much – otherwise the model would be unwieldy – we’re vulnerable when we mistake them for reality.

Still, the rhetorical power of models is persistent, because they imbue statements about the future with the aura of inevitability. In an age of uncertainty, they seem to promise certainty.

Nor are they as objective as their numbers imply. Chair of the US Federal Reserve, Alan Greenspan, acknowledged as much. When testifying before Congress about why he had failed to predict the 2008 banking crisis, he called his conceptual model an ideology. “Everyone has one”, he said. “You have to. To exist, you need an ideology”.

His own ideology had assumed unregulated markets to be the safest, something he now saw as “a flaw”. But that flaw – and the economic crisis that followed – inadvertently demonstrated just how easily models give authority to bias and belief.

Taking history as a model presents similar dangers. The belief that history repeats itself is widespread, though rarely shared by professional historians. Mostly, it is our own history that we see being repeated – not anyone else’s.

When the Arab Spring unfolded, the Russians saw Russian history, with the politician Dmitry Medvedev fearing that, like the fall of the Berlin Wall, these demonstrations would prove destabilising for Russia.

Meanwhile, President Obama likened uprisings in Tunisia and Egypt to the Boston Tea Party and the beginning of America’s war for independence, drawing comparison too with the civil rights protest of Rosa Parks. Such analogies blinded both leaders to the dangerous contingencies and complexities of unfolding events.

The difficulty of reading the future in the past presents special difficulties in the context of artificial intelligence. Working from datasets – necessarily from the past, and often incomplete and biased – can produce outcomes that purport to be objective but which, when challenged, may prove to be seriously flawed.

In Pennsylvania, when the Allegheny Child Protective Services tried to cut costs by using AI to evaluate children most at risk, it used its own data. But there were two problems with this approach.

First, the dataset didn’t include families the ACPS had never had contact with: predominantly middle or upper class white families. It also both over-weighted the risk to older children who’d been referred often and underweighted infants and toddlers they’d never heard of before. A baby left in the snow would appear less at risk than a teenager out at night. The scale at which the technology was applied meant that thousands of such errors emerged.

We see this today as commentators argue that Putin is already defeated. Because the past shows that it is harder to hold a country than to invade it, Russia has already lost its war.

Wilful blindness or wishful thinking, at a time when we desperately struggle for political imagination and creativity, such palliative certainties provide neither inspiration nor insight. They simply reassure us that we need merely wait.

Even the harshest critics of models acknowledge their usefulness; we would lack insight into the climate crisis without them. But we would do better to approach them as possibilities – provocations that challenge us to envisage alternatives, rather than giving us permission to surrender to their narrative power.

We don’t know the future because it hasn’t happened yet – and in that uncertainty, however uncomfortable, lies opportunity.

The opinions expressed in this blog are those of the author. For more information, view our full statement on external contributors.