Forecasting

Quantitative forecasting — estimating the odds of future events, and later checking those estimates against reality — has transformed fields like finance and sports. But in others, like national security or public health, major decisions are often guided by intuition or vague, qualitative judgments about the future. This makes it harder to learn from experience and to know whose predictions deserve trust. 

Explicit, probabilistic forecasts can help. They make assumptions clearer — no guessing whether “probably” means a 60% chance or 90% — and allow accuracy to be measured. Over time, they can be used to improve decision-making. 

At Coefficient Giving, we use forecasting internally: our grantmakers often make probabilistic predictions about the outcomes of grants to test and refine their judgment. We aim to expand this practice more broadly — especially in areas like evaluating risks from AI or biotechnology, where errors in judgment could lead to disastrous outcomes. We also want to ensure that high-quality forecasts are available to inform decision-makers in those critical areas.

Work we support includes:

1. Platform development and practical forecasting tools, including those that use AI to make better or cheaper predictions.
2. Research on forecasting methods, including on ways to improve their accuracy and relevance for hard-to-measure questions.
3. Forecasting tournaments that explore questions relevant to our other focus areas.
4. Projects promoting the use of forecasting for high-stakes decisions, particularly those related to global catastrophic risks.