"No one opts in" aside, any opt-in metrics you do collect tend to be skewed towards how the people who opt in, use the product. Anyone serious about making product decisions using opt-in metrics should be aware of this bias.

This could be solved by a lot of transperancy about what collected telemetry is saying. A user can then check if the users that opted-in to telemetry are representative of his own use cases and thus make an informed decision if he should opt-in as well (if he's not well represented). Telemtry is a lot like voting.

The vast majority of people will not read about the collected telemetry, even fewer will read it and then make a decision to opt-in.The telemetry is optimizing for the vast majority, not the loud minority, hence opt-out works better in order to cater to a larger group of users. Your voting analogy is really bad.

With that said, I don't really like telemetry and will turn it off.

I think voting is a good analogy for telemetry. You submit your use case to help decide development direction.

So, `about:telemetry`, https://telemetry.mozilla.org

The difference between opt-in and out organ donors between otherwise similar countries is a staggering 80% — people are seriously lazy and will choose the default almost always. I think one should not be afraid to “exploit” this innate human quirk, if it is done for good reason. Unethical people will abuse it either way.

Almost nobody goes out of their way to enable telemetry because they want to help some project. Very few power users do (and they don't represent the majority in most cases) and I'd say zero regular users would care.

I don't mind it if it's opt-in with a user-facing setting to turn it on or off.