As we exit February, we’re coming out of prediction season. How will 2026 differ to 2025? What AI developments will have the most impact? What changes will shape the year ahead?

Experts from all industries are currently publishing their answers to these questions. And more.

But how can those writing their predictions for the year ahead best identify where change will/won’t occur? And how can their audiences spot which predictions are worth taking notice of?

I predict that by giving this article 5 minutes of your time, you can find out. Why? Because we’re answering the question “how to make a good prediction?” This will allow you to spot a good one too!

Breaking down the big questions

Often, predictions are made in response to large questions. How will ‘Sector X’ change in the next year? How will AI reshape ‘Topic Y’?

But the best predictions use of a process called ‘fermizing.’ This is a method developed by the physician Enrico Fermi. It involves breaking down a problem into smaller, manageable problems that are easier to solve. And in doing so, makes the proposed answers (predictions) less general and simplistic.

Identify the players in a prediction

Game theory suggests that each prediction involves a series of players. These players could be a technology. A consumer group. A type of business.

Political scientist Bruce Bueno de Mesquita’s work proposes that good quality predictions hinge on 1) identifying the actors (“players”) involved. And 2) what the players want.

Bueno de Mesquita says that this model works because it assumes players act in their self-interest. Something which is predictable if quantified correctly. And this is supported by Morgan Housel’s idea that greed – alongside fear, love and hate – are timeless human motivations.

Thinking broadly and probabilistically

The Good Judgement Project (GJP) is a forecasting tournament. The GJP have identified that a consistent sign of a good forecaster/predictor is someone who makes probabilistic predictions.

A probabilistic prediction states that there’s a percentage chance or specific likelihood that something will happen. This is opposed to a deterministic prediction, which speaks with certainty.  

Probabilistic predictions are better because they require using a broad array of knowledge and thinking. And show open-mindedness in doing so. Whereas a deterministic prediction suggests the opposite. This concept is illustrated by Isaiah Berlin’s essay The Fox and the Hedgehog. In this, the fox is a metaphor for probabilistic thinking, and the hedgehog is deterministic.

In short, predict like a fox. But be wary of predictions from hedgehogs!

Search for signals of change

The best predictions come from fox-like thinking. But their content also details the signals of change.

Game designer and futurologist Jane McGonigal describes signals of change as vivid, detailed and specific examples of innovation, change, or invention. But NOT broad trends like AI or inflation.

The best predictions will need to gather details about signals of change. And then cite them in their predictions. The more specific the signal, the more certainty your prediction will imply.

Respect the reality of change

Making public predictions is a tough job. A prediction that suggests “things will be similar to last year” won’t grab any headlines. But equally a prediction that implies “everything will change” will get headlines but be criticised for over exaggeration.

The best predictions will suggest that change is likely. But always caveat it.

For example, Bruce Bueno de Mesquita distinguishes between irrational and rational actors. Irrational actors are people with unpredictable behaviour (e.g., persons without stable preferences). Conversely, rational actors are those whose behaviour is predictable and can be modelled.

Bueno de Mesquita believes that we shouldn’t predict change on behalf of irrational actors because they’re too unpredictable. So, reflect this thinking in your own predictions! 

Similarly, Morgan Housel cautions any thinking about change with the caveat that core human motivations are stable. And that the biggest changes come from what we don’t see coming.

Be honest about your predictive past!

If people are assessing how good your prediction is, the bluntest instrument available to them to assess this, is to see how accurate your past predictions have been.

If you are a professional predictor/forecaster, your (suspicious) audience can do this by finding your brier score. This is a score based on how accurate individual’s previous probabilistic predictions have been. The higher the score, the more faith you can have the individual’s predictions.

But assuming you don’t have a brier score, there’s a simpler way you can create some trust in your predictions. And that’s by sharing your history of past predictive successes and failures.

Yes, failures too! That’s because we know from the pratfall effect that admitting a fault can increase your likeability. Which in turn increases trust.

So, if you’re looking to make predictions about the year ahead:

  • Break down the prediction you’re making into more solvable chunks
  • Think broadly and probabilistically
  • Look for signals of change. And cite them when communicating your predictions
  • Don’t predict change from unpredictable actors
  • Share your past stories of predictive successes and failures

This article was originally published on Research Live.