They may not understand the underlying architecture, but they definitely do understand it from a user perspective - they can send an email regardless of provider.
Telephones are the same way, too. You get a phone number from whatever company you want and can use it to call people no matter what provider they’re using.
I’m not convinced that open/federated protocols are inherently more difficult for people in general to understand.
I think you are overestimating the level of understanding a lot of people have. For lots of people, IE == the internet. Start Firefox: "My internet looks different". Same for Outlook vs GMail.
The average office worker (and home computer user?) has eg. Outlook installed and set up by the IT department.
But yes, the analogy with cell phone providers is useful. With this analogy, who would still tie their social media profile to a single provider?
It's not the idea that you can chat with people on other providers it's that you still have to choose a provider with no real guide on how you would make an informed decision among them.
You choose your phone provider by looking at the top 3 or so established providers and the choice is usually made on price or other signing incentives. With your ISP you just pick from a couple ISPs that serve your area. With email you just pick GMail or whatever your work gives you.
All of these decentralized chat systems just need to act centralized. Have Matrix.org just be the matrix server that people sign up on and let the fact that other providers exist just be something that comes naturally. Have Mastadon be the StatusNet server but casually interface with other servers.
What seems to kill these projects is that the group positioned to own the market actively shoots themselves in the foot in an attempt to not become too big.
Let the protocol be an implementation detail. Sell the provider as a service.
