[Felix21]

I DON'T agree with this conclusion.

You can easily just test both (with Neither as your control) and survey your users to find out which group is more satisfied with your UI.

When it comes to what customers want or will like, what you "think" is almost always wrong, so stop thinking and start testing.

[dredmorbius]

The test shouldn't be what users say they're more satisfied with, but which labeling results in better task performance. Now, if self-reported user satisfaction is your designated task, that's fine, but contexts such as multi-user support, shared system access, single-user use of multiple accounts, and centralized maintenance/access of many multi-user systems might also be considered.