Insurer Amlin and Oxford University have launched a three-year research project focusing on the systemic risks of modelling.

Amlin Group Chief Underwriting Officer Simon Beale says reliance on technology in recent years has increased risk, as demonstrated by the global financial crisis.

While the insurance market “largely escaped this contagion”, it is vulnerable to scenarios “where modelled expectations of risk do not reflect reality and/or modelled output is taken for granted”.

Mr Beale says the research is “an important first step in understanding better the risks we face and ensuring that our reliance on the increasing amalgamation of models does not prove catastrophic for the insurance industry”.

Andrew Snyder-Beattie, Academic Project Manager at Oxford University’s Future of Humanity Institute, told insuranceNEWS.com.au the team hopes to “achieve a more robust form of model verification”.

“We’ve finished some precursory research into how cognitive biases and heuristics interact with modelling techniques, specifically in the use of catastrophe models,” he said.

“One area we’ve examined is that of a ‘biased error search’. Depending on the model outcome on vendor settings, we can’t guarantee that the decision to tweak the model is made in an unbiased way.

“Rather, market forces or personal incentives might play a role in when to search a model for errors, what sorts of errors to look for and when to stop searching for the errors.”

Mr Snyder-Beattie says the team has also addressed the “autopilot problem”.

“When we rely too heavily on models to make decisions, our ability to deal with uncertainty or situations can decline,” he said. “One way around this might be to force people to make estimates without the model before seeing the model’s estimate.”