Your Inbox Is a Bandit

Your Inbox is a Bandit 🔗

1 The Bandit 🔗

In probability and machine learning, there’s a notion of multi-arm bandit problems. Your mental image should of you sitting in front of a row of slot machines (known as a one-armed bandit: the “arm” being the lever, “bandit” because it takes your money), each of which may produce a payoff at any moment. Which one do you play next? Facing a machine, you have a choice between exploitation— pulling its lever— or exploration— choosing another machine. The literature is always cast in terms of maximizing your reward, though in the case of slot machines, and the situation we’ll discuss here, it’s more about minimizing your loss.

Many real-world problems can be cast as bandit problems. How do you allocate money to portfolios? Which restaurant do you go to? What ad do you show? And so on.

Very loosely speaking, your email inbox is a bandit. You’re given a bunch of message threads, and you have to decide whether to “exploit” a particular one— read it, respond to it, deal with replies in the thread, and so on— or “explore” another thread. And much like a slot machine, this bandit also leaves you broke: if not for money, then certainly for time and energy. Having to make choices is cognitively taxing, and having to constantly make decisions from a large number of options becomes numbing. We end up picking either sub-optimal strategies (like always choosing the top thread), clicking at random, or just rooted in indecision. Nobody enjoys this.

2 Combatting the Bandit 🔗

Many people have suggested strategies for dealing with this. One popular technique is Inbox Zero. The jokes about it suggests virtually nobody attains it, but I’m not even convinced it’s a virtue. I have many correspondents— such as my collaborators— who write detailed, thoughtful messages, and deserve detailed, thoughtful answers. (And if I don’t provide those, they’ll stop writing me, which will be to my detriment.) Mindlessly dashing off replies to get them out of my inbox is neither an option nor desirable.

Other have suggested strategies like Getting Things Done. These always strike me as heavyweight and solving problems I don’t have.

This article instead describes a strategy I’ve been trying to avoid this problem, and it’ll evolve as my strategies evolve.

3 My Context 🔗

... continue reading