Answer in brief
The Tokyo meeting taking place this week
The Bank of Japan hosts ECONDAT on 5–6 October 2026, bringing machine learning and nontraditional economic data into the same research programme. Banca d’Italia’s co-organiser announcement confirms the Tokyo dates. For readers learning AI, the setting is useful because an economic question must survive contact with evidence. A model’s fluency or technical complexity does not establish that its answer measures what a researcher intended.
Two AI questions that should stay separate
In his 5 October opening remarks, Shinichi Uchida distinguishes analytical uses of AI from its implications for productivity and labour markets. Our reading is that these require different tests. An analyst might automate the classification of company statements; a policymaker might ask whether firms’ adoption changes production or employment. Success in the first task does not quantify the second. Keeping the question explicit is the beginning of a credible analysis.
Turn the programme into a learning project
The published programme includes research on sentiment and expectations using language models. Try a small exercise with public documents: choose one definition of sentiment, label a short sample yourself, record the model and prompt, then examine disagreements. Keep the sample’s origin and dates visible. This is an editorial suggestion for studying measurement, not a claim that a classroom experiment reproduces a conference paper or can support a national economic conclusion.
What would count as a useful result
Ask whether the method works beyond the examples used to develop it, whether the source population is representative and whether a simpler baseline performs comparably. Write down one circumstance in which the approach could fail. Our view is that the most useful skill here is disciplined comparison, supported by clear documentation. The meeting’s timetable establishes what is scheduled; evaluating a particular finding requires reading its paper and evidence. It is too early to treat the programme itself as a set of completed, independently validated results.
Questions and answers
Is ECONDAT a monetary-policy decision?
It is a research conference. A programme or an opening speech should not be presented as a new policy decision.
What can a beginner learn from this meeting?
Start with how a dataset was collected, what the model measures and how its output is checked against an independent observation.