The landmark case is behavioral economists Eric Johnson and Daniel Goldstein's 2003 study "Do Defaults Save Lives?" (Science). Countries with opt-in organ donation saw consent rates of 4–27%; countries with opt-out defaults (enrolled unless you withdraw) saw 85–100%. Same choice, different default, wildly different outcome.
People stick with defaults for layered reasons: it reads as an implicit recommendation, changing it costs effort (see: friction), and status-quo bias makes people avoid decisions they might later regret.
Opt-in defaults protect users (marketing consent off by default); opt-out defaults are only justified when the benefit clearly favors the user (automatic backups, security updates). Quietly defaulting something to the company's advantage is a dark pattern.
Below, A's checkboxes default off and mostly stay off; B's default on and mostly stay on.
When to use
Before setting a checkbox or toggle's default, ask who benefits from that default. Design assuming most users will never change it.