Reshaping risk: how causal AI is shaking up risk modelling
Artificial intelligence and data analytics are reshaping how actuaries understand, quantify and manage risk – yet despite enormous predictive modelling advances, many systems still rely on correlations rather than true causation. This means actuaries often know what happens but not why.
The arrival of causal AI, based on computer scientist Judea Pearl’s theory of causal inference, heralds a profound shift. Instead of analysing statistical relationships passively, it enables systems to reason about cause and effect, simulate counterfactuals and make decisions that are grounded in understanding, not just prediction. It is a step towards more transparent, adaptive and trustworthy actuarial models.
Traditional actuarial models are built on large datasets and correlation-based methods – powerful for trend detection but limited when it comes to dynamic human behaviour, complex feedback loops or interventions that alter outcomes. For example, a model might identify that non-smokers live longer but it cannot determine whether encouraging a smoker to quit will reduce long-term risk for a given cohort. In insurance pricing, such dependence on historical data leads to static rates, limited behavioural insight and underestimation of risk elasticity.
Causal reasoning, in contrast, can identify which factors truly drive outcomes and simulate what would happen if those factors changed. This is a paradigm shift for actuaries: from relying on risk correlations to understanding the causal drivers that generate risk.
[....]
