The Followup: Back to Work e43

blog note: this is the first installment in what i plan to be a regular feature, that i'm tentatively calling The Followup. the theme is simple: i listen to podcasts, and as i'm listening i tend to have lots of thoughts. a lot of times i wish i could get right into the conversation. i don't have the luxury (or desire) to listen to live recordings of podcasts and provide feedback that way, and half of the benefit of the format is the ability to time-shift. hence, i have to write things up after the fact, or let them pass forgotten. even though i time-shift, i'll try to keep these fresh…at least to the extent that i try to not let my listening backlog get too long. this post is based on episode 43 of the fantastic Back to Work podcast by Merlin Mann and Dan Benjamin, which was recorded two weeks ago.

on with the show.

discussing questions

at right around the 15:00 mark of the episode, Merlin said this and my ears perked up:

i want to talk about the potential role of questions in mostly helping you figure out what the next question should be. and that's not to say that a question cannot have an answer, but i'm interested in the idea of questions as part of an iterative process.

i got excited because even though Merlin meant this in a very real-world, productivity way, this sounds a lot like some key concepts in the areas of linguistics that i've been doing research in lately. in particular i've been working with and adding to a framework that formalizes discourse, founded by the Roberts / Tonhauser / Simons / Beaver research group. in plainer terms, i deal with a logic or way of computing how the pieces of a conversation have to line up in order to make sense or be useful.

stack 'em up

one important piece of the theory is that every conversation has a Question Under Discussion, or in some versions, a stack of them. you can think of this stack either in the plain physical sense, like a stack of cards, or in the computer science sense. either way, there's always one question that's on top. the goal of whoever is participating in the conversation is to answer that question. that might be an easy job, or it may be a very difficult one.

kinds of answers

Merlin actually has a really good intuitive grasp of what sorts of questions have simple answers and which don't. semanticists typically say that the meaning of a question is the set of all its possible answers. one example he gave was "do you want lemonade with dinner?" – for this yes/no question, there are just two answers: {(yes) i want lemonade with dinner, (no) i don't want lemonade with dinner}. if you're having trouble picking one of those, it means you're just indecisive, not that there's anything faulty with the question.

on the other hand, Wh-questions (so called because they are introduced by words that, in English, mostly start with the letters wh) have lots and lots of answers. "where are my keys?" is represented by the set of every statement of the form "my keys are [in x place]”. when we discuss these answers, even in formal logic, we tend to make some sort of reasonable or sensible partition. {my keys are on the desk, …on the nightstand, …under a pile of papers, …at work, etc.} are the options we present, not {my keys are on Olympus Mons, my keys are at the bottom of the ocean, etc.} nor {my keys are 1 inch from the north edge of my desk, my keys are 1.1 inches from the north edge of my desk, …1.2 inches…, …1.3 inches…, etc.}, even though these are all possibilities.

once we have a sensible set of choices, the goal of the conversation is to find the right choice or choices. this could take a single step: if i saw the keys on the desk, i can say so, and the problem is resolved. if i'm looking at the desk and don't see them, i can say they're not there and, by doing so, refine the question. these are called complete and partial answers respectively. in discourse theory, whatever you say next has to give a partial or complete answer to whatever question is on top of the stack.  if you fail to do that, your response is considered irrelevant, and whoever you're talking to will probably look at you funny.

questioning questions

one of the interesting things about relevance (and one of the major new ideas in a paper i presented at WECOL in November and am drafting for publication now) is that you don't have to assert things — that is, make statements — to be relevant; you can also ask questions or give commands. here's where the linguistics meshes with Merlin's ideas: it's possible to iterate questions to ask better questions, but there are rules on how to do it.  the new question has to be relevant to the previous question. since questions are always defined as a set containing multiple statements, they can never give a complete answer, but they can go a long way towards giving a partial answer.

think of the two questions as a Venn diagram. each circle represents all the answers to a particular question, and the overlap are the statements that could answer both questions. this set of statements is the new, refined question. it takes a little bit of logical thought to get a good overlap, but that's just the process that Merlin is after. so there's a lot in common between sound, rational, productive work processes and the way every single conversation works. i'm guessing a lot of people haven't thought of it in those terms, but doing so might help, both in how we get things done and how we talk.