WTW (Willis Towers Watson, NASDAQ: WTW) has announced the launch of the newest version of its leading-edge Radar pricing software.
Radar 4.15 overcomes a major barrier to wider adoption of machine learning by introducing a ‘marketfirst’ capability that makes it significantly easier for insurers to benefit from the full predictive power of machine learning without losing the ability to interpret accurately the outcomes of
increasingly complex models. By introducing a proprietary algorithm that resolves the previous inherent trade-off between accuracy and interpretability of models, the Radar upgrade delivers an exceptional level of predictiveness and transparency.
This will transform insurers’ ability to unlock the full potential of machine learning in order to optimise their underwriting and claims processing,
whilst adapting to a regulatory environment increasingly focused on fairness.
Serhat Guven, Managing Director at WTW, said: “The innovation underpinning Radar 4.15 is ground-breaking, for the first time giving insurers access to state-of-the-art, speed-toaccuracy performance that goes hand-in-hand with unrivalled levels of transparency. Technological advances will inevitably continue to reshape the insurance landscape, yet it will only be possible to harness the full potential of these technologies if customers and regulators trust they are being used by insurers responsibly.”
ELIMINATE BIAS
AI-based and machine learning tools offer higher predictive performance and accelerated speed to market, yet by their nature are more complex and less transparent, preventing the user from fully understanding the model they are creating. Their ‘black-box’ nature poses significant risks if deployed without due care, including amplifying bias risks that lead to discriminatory decisions, and regulatory tolerance is running out for pricing perceived to be unfair.
Serhat added: “Radar 4.15’s powerful new algorithm addresses the black-box limitations of machine learning by offering insurers a ‘best-of-both-worlds’ solution that combines the pure predictive power of the machine learning approach of Gradient Boosting with the exceptional
intepretability and transparency of Generalised Linear Models.”
Other updated features of Radar include enhancements to Radar Base and Radar Workbench.
More here at the WTW site.
Be the first to comment