ECB looks to AI to better understand inflation after big misses

It is now exploring ways to process and analyse millions of data points

FRANKFURT (Reuters) - The European Central Bank is looking at how artificial intelligence could help improve its understanding of inflation after underestimating price pressures for years and delaying the start of the most aggressive policy tightening in its history.

Joining the masses of firms already using AI, the ECB is now exploring ways to process and analyse millions of data points, including public price data, corporate statistics, news articles and bank supervisory documents to produce better analysis for policy decisions.

"AI offers new ways for us to collect, clean, analyse and interpret this wealth of available data, so that the insights can feed into the work of areas like statistics, risk management, banking supervision and monetary policy analysis," a blog post published by the ECB on Thursday, said.

The ECB has for years underestimated inflation, and some policymakers have openly questioned its models and the viability of basing sound policy on numbers that required constant upward revisions.

Among several AI initiatives, the bank wants to deepen its understanding of price-setting behaviour and inflation dynamics, the blog said.

Using web scraping, the ECB can collect masses of real-time price data but the figures are unstructured and unsuitable for calculating inflation. So the ECB wants to harness AI to structure data and improve its analysis, it said.

Another initiative is to automate the classification process of data from tens of millions of firms, banks and public sector entities, so it had a better understanding of their financial state.

The ECB also hopes to use AI to simplify its communication, fighting back critics who say that its overly complex, technical language is impossible for ordinary people to grasp.

"A large language model can also help improve texts being written by staff members, making the ECB's communication easier to understand for the public," the bank said.

"Relatedly, we have used neural network machine translations for a while now to help us communicate with European citizens in their mother tongues."