The institution of email

I think our norms around email could be based to some extent on very old norms from ancient tribal culture - there you had to respond, and not just weeks later, if someone asked you about something. Therefore one would have to target one of the things you mentioned (length, cost, permancence).

What if we had an email-tax, thereby changing the cost of the form? (Not so easy to implement in practice, of course). That would not change the norms per se but it would reduce the burden of the norms that result from a mismatch between the evolutionary source of the norms and the current technical environment. What email-tax level would lead to the highest welfare (happiness, productivity) could be an empirical question then (and I think the optimal tax would be higher than zero).

And the rarer strategy of actually dealing with all of one’s emails promptly doesn’t even seem obviously better.

Why not?

I get lost there, because my model is 'emails are important for communication, they have the problems you describe here, and the solution is to answer as quickly as possible'.

Sometimes, responding takes too much effort for that to be reasonable, but it often doesn't.

I think the norms follow from quite ancient communication expectations.  If someone in your village sends their child with a request, you can say no, but you can't just ignore it.  Acknowledgement is a near-requirement for communication to have actually happened.

What has changed is the number of people who expect to communicate with you, and the ease of doing so.   Which leads to a lot of things that look like communication, but aren't.  e-mails run the gamut from personal messages from a close friend or relative (which you may damage the relationship by failing to respond appropriately, but which you can negotiate expectations over time) to automated spam which you probably automatically discard.  

The ones in-between, from strangers or distant acquaintances are the tricky ones.  You don't necessarily owe them anything, but they might have something useful to say, and a few of them might turn into closer friends.  So you have to decide how much to invest, from a boilerplate "thanks for the e-mail, I probably won't have time to read it thoroughly." to an actual thoughtful engagement.

Celebrities have had to deal with this by mail (and telegram) for a century.  Some have hired people to respond for them.  Some are famous for personal responses.  Some are black holes.  There is reputational impact to those choices, but they're all acceptable.

What about a norm that emails should only require one reply, and almost never end up as long email chains? I think most emails that I receive are parts of long threads, such as threads to schedule a date and time for a meeting, or threads to discuss technical information. Shortening those threads by being more direct or switching the conversation to a better medium (like a call) seems like a worthwhile goal.

Cal Newport calls this a process-centric approach, in which one spends time up front writing emails and replies that do not require much or any follow-up. For example, instead of discussing a complex idea over email, Newport would argue that one should just schedule a call. Likewise, Newport has a template that allows for meetings to get scheduled with just two replies (shown in the linked source). Newport argues that although this approach reduces the number of times one must check one's inbox per day, it does not reduce the total time spent reading and writing emails. However, I suspect that if both the sender and receiver held this approach as a norm, the total time spent on email might decrease, because one would receive emails that are already easy to resolve.

It seems like this approach could be a norm that could plausibly spread because it confers a clear benefit but only requires one to write emails a bit differently. If enough people adopt it, the number of emails sent and received would decrease. In contrast, strategies like charging people to send one emails or not responding to emails by default seem tenable only for those who work independently of any typical employer, so they are unlikely to spread.