Where do you get your ideas?

by John Quiggin on August 28, 2020

The most memorable answer to this question came from science fiction writer Harlan Ellison, who said “Poughkeepsie” (on checking Wikipedia, I learn that he died a couple of years ago).

But in the context of discussions about remote work, I’m interested in the claim that random physical meetings (the archetypal example being corridor or water-cooler encounters with colleagues) are an important source of ideas, and therefore a reason for not working remotely.

This seems to be the kind of topic for which the data will consist mostly of anecdotes and introspection. A marginal improvement is too look over my own list of publications to see if I can identify any where the source arose from some particular interaction.

Looking at my 100 most-cited papers in Google Scholar, most collaborations are the result of planning rather than chance. In pre-Internet days, most of my collaborations started from seminars and conferences I spoke at or attended because the topic was of interest, or else from direct approaches by a colleague, usually in the same department. From the early 1990s onwards, direct approaches mostly came by email, and work has often been done the same way. In several cases, I have written joint papers before ever meeting my co-author(s), though in other cases in-person collaboration with one or two co-authors works better.

More interesting to me, are the cases where the idea has come from blogging. Some notable examples

  • My Zombie Economics book. Starting with blog discussions, the idea for a book came from blog commenter Max Sawicky, and was picked up by Seth Ditchik at Princeton UP, who also commissioned Economics in Two Lessons and my current book-in-progress Economic Consequences of the Pandemic
  • Cross-disciplinary collaborations with Henry Farrell and LA Paul both arising from my involvement with Crooked Timber
  • This paper, which started with a comment on a blog post to the effect that “future generations” are in fact already alive (At least I think that’s how it happened. I could never locate the comment to acknowledge the source.)

It seems to me that that these are much more like the kind of serendipitous links that are supposed to be generated by water coolers.

Of course, academic research is a special kind of work, and I’m much more involved with the Internet than most of my colleagues (or, at least, a few years ahead of the general adoption trend). So, I’d be interested in anecdotes from others and links to actual research, if there is any.



Anon 08.28.20 at 10:06 am

There is a huge literature on this, usually called “microgeography”. Even small changes in physical distance between potential coauthors (like, on different floors of the same building) cause huge changes in the probability they work together (e.g., https://pubsonline.informs.org/doi/pdf/10.1287/mnsc.2017.2798 and the lit cited therein). Bell Labs famously designed their building to ensure people from different groups were forced to bump into each other on the way to lunch or the parking lot, a decision that paid many Nobels as a dividend. In the patent and scientometrics lit, it is very clear that physical co-location is incredibly important for knowledge transfer and collaboration within firms. Within larger geographies, “disembodied” knowledge spillovers from one lab in a city to a different lab or firm are much more likely to happen if the latter is in the same city.


John Quiggin 08.28.20 at 11:27 am

Thanks! Fascinating.


Ebenezer Scrooge 08.28.20 at 12:13 pm

I think most of my ideas were solitary, but most of them were bad. I could shoot down a lot of the bad ones myself, but I often needed an outside executioner. (My darlings!) There was nothing as effective as frog marching the idea down the corridor and exposing it to a colleague.


Bill Benzon 08.28.20 at 12:28 pm

A twitter conversation with Hollis Robbins (Dean of Arts & Sciences at Sonoma State in northern California) led to an interview with her about her current book, which I published in 3 Quarks Daily: An Electric Conversation with Hollis Robbins on the Black Sonnet Tradition, Progress, and AI, with Guest Appearances by Marcus Christian and GPT-3. That in turn led me to think serious about GPT-3, the hot new AI engine that all the kool AI kids are talking about.

While I’m wary of the hype, I’ve got a long-standing interest in computational linguistics and natural language processing (I was trained in computational semantics by David Hays, a first-generation researcher in machine translation and the man who coined the term “computational linguistics”). So I began thinking a making comments.

John Lawler on Facebook pointed me to this blog post, where I found a provocative comment by Graham Neubig and that, in turn, has so far led to a bunch of blog posts and two working papers: GPT-3: Waterloo or Rubicon? Here be Dragons, Working Paper, Version 2, August 20, 2020, and What economic growth and statistical semantics tell us about the structure of the world.

Now, will that work go beyond the self-published working paper stage? I don’t know. While I do continue to publish in the formal literature every now and then, I find that, for better or worse, the work of shaping my ideas to fit the demands of those journals, and the problematic nature of the review process imposes, an opportunity cost that’s a bit steep for me. I’d rather do more thinking and self-publishing. At least that gets the ideas where others can take a crack at them.


Bill Benzon 08.28.20 at 12:31 pm

Whoops! That’s Dean of Arts and Humanities, not Arts and Sciences.


Sumana Harihareswara 08.28.20 at 2:41 pm

Heidi Waterhouse notes: “The important part about co-location is not actually sharing ideas, it’s getting to know each other so that lossy communication didn’t get interpreted badly.” This is less about pure research and more about industry, but might still be applicable.


Kenny Easwaran 08.28.20 at 5:12 pm

Looking at my own co-authored work:

I have multiple papers published and in various stages of progress with one co-author, which started with a conversation at a bar at a conference. I haven’t yet had similar success with Zoom happy hours after online conferences, but it’s also true that no other physically embodied interaction has turned out to be as productive.

I have two other co-authorships in progress, one of which arose from a serendipitous conversation at a physical conference (we hadn’t seen each other since undergrad and had gone on to academic careers in different disciplines, but realized we had something similar to say about the Axiom of Choice) and one of which arose from an e-mail (and while I had read some of his work before, I actually can’t recall if we have ever met in person).

I have a four-author paper that arose from a physical reading group that I joined because it was at another university in the same city as me. I haven’t yet had similar co-authorship arise from virtual reading groups, but again, I’ve been in a lot of physical reading groups that didn’t lead to this sort of collaboration either.

I have three co-authored papers with my PhD advisor, all written years after I graduated (one of which involves two other authors as well, one who had been in my graduate program at the same time and the other of whom had been a regular at the same conferences as us). I assume these collaborations would have worked fairly similarly if my graduate education had occurred mostly through online means, though it’s hard to say whether relationships with fellow grad students and fellow conference-goers would be the same.

I also have two early co-authorships that arose entirely because of a blogpost or e-mail I wrote and were conducted entirely online (one of those co-authors I believe I have only met in person once, very briefly).

So in my own experience, it seems that physical proximity was important for a few of the collaborations, but not quite as high a proportion as I would have expected!

It’s a bit harder to say about all the smaller interactions that improve a paper but don’t promote someone to co-author. I’m fairly confident that a large fraction of those are from physical proximity (membership in local reading groups that are only partly based on research interests and more strongly based on being in commute distance of each other), though another large fraction are from questions in official conference or seminar talks that would have happened similarly through online means. I expect that most of the professional relationships that have strongly influenced my work were originally nurtured through repeated informal physical interactions either in the same department, or at physical events in the same city, or at para-conference events like dinner, drinks, and the like.

But there definitely are also a growing number that are initially nurtured through the kind of serendipitous repeated small interactions on social media.

(I’m very much of the school that thinks that urban geography and building microgeography are major factors, but I’m also interested in the ways that virtual interactions create their own geography-like conditions, and appreciate this excuse to introspect about my own past.)


Robert Vienneau 08.28.20 at 5:20 pm


Last week, I stumbled on a bootleg PDF somewhere on the web of Economics in Two Lessons. I cannot recall where, but I thought you should know that that is out there.


John Quiggin 08.29.20 at 12:16 am

@8 Thanks for the alert, Robert. I’ll mention it to the publishers. But, if anyone is keen enough to track down a bootleg version and read it, I guess I’m not too worried about missing out on royalties.


carl d 08.29.20 at 1:26 pm

In most cases of co-authorship you have met the person at least once in reali life first; it makes it easier to e-mail them I think


Dwight L. Cramer 08.29.20 at 9:39 pm

It would be very interesting, in about ten years, to compare the genesis of academic and professional collaboration from, say, 2015-2019, with the genesis of that collaboration from 2020 t0 2024. I think the response to the pandemic is creating a discontinuity that might provide some valuable insight into the microgeography problem.

As things currently stand, my sense is that a great deal of confirmation bias is at work in people’s arguments. But if you saw a marked fall off in research output (however measured) from jurisdictions that failed to control the pandemic as compared to those with greater success, that would be powerful.


Anders Widebrant 08.30.20 at 12:24 am

From my experience in larger corporate environments, I think there is a form of plausible deniability inherent to in-person conversations that lowers the treshold of communication and is hard to reproduce in other mediums.

That’s to say, if I reach out to you by email, or by some other directed form of communication, I’m making a specific effort, which implies you should reciprocate, which in turn makes me wary of wasting your time, and so on. If we’re just talking, then we’re just talking.

A plausible way to mitigate this is by regular scheduled small group calls with as little of an agenda as possible.


Paul 08.30.20 at 11:26 pm

I suggest taking a look at AnnaLee Saxenian’s Regional Advantage (1996), which argues that the shift of digital innovation from Route 128 (around Boston) to Silicon Valley came in good part because the hierarchical East Coast organizations didn’t support social interactions among people in different companies, which turned out to be vital for the spread of ideas in Silicon Valley. This account of the importance of human connection to develop and spread ideas still helps explain why Silicon Valley companies put so much money into their campuses and bringing people together despite their claims that the technology they produce makes distance irrelevant.


notGoodenough 08.31.20 at 8:07 am

John Quiggin @ OP

Apologies for this being off topic, so feel free not to post.

I wanted to thank you for the link to your work on jobs in the coal is killing us thread. As someone who has to spend time writing proposals and is active in environmental activism (professionally and personally) it is useful to have this sort of data to hand for those of us with only a limited understanding of economics.

Also, this thread has reminded me I should buy a copy of economics in two easy lessons, so thank you doubly!


John Quiggin 08.31.20 at 10:21 pm

@14 thanks!

Comments on this entry are closed.