One thing I wish more people in the UK understood:

“behavioural science” (as practiced by the Nudge Unit) is much, much closer to Digital Marketing than to any type of academic social science.

Social science is about developing theories with testable hypotheses and subjecting your work to peer review.

Behavioural science is mostly atheoretical A/B testing conducted for corporate clients, with no external replication and no theory-building.

Which is why behavioural science is notoriously prone to discovering "kooky" results that are great material for an NPR podcast but don't replicate.

It's also why some of the field's leading figures tend to get, uh, a bit defensive when people scrutinise their work...

Source for the now-deleted tweet:
https://statmodeling.stat.columbia.edu/2019/08/15/replication-police-methodological-terrorism-stasi-nudge-shoot-the-messenger-wtf/

And another of Gelman's blogs on Sunstein, worth a read:

https://statmodeling.stat.columbia.edu/2019/08/24/more-on-why-cass-sunstein-should-be-thanking-not-smearing-people-who-ask-for-replications/