NSF CAREER Award

Jeffrey P. Bigham

@jeffbigham

4/10/2019

A couple of days ago, the National Science Foundation (NSF) held a workshop on writing CAREER proposals. The CAREER program is a special solicitation from the NSF that is targeted at junior faculty. I was fortunate enough to win one of these in 2012, and shared my proposal with Ehsan Hoque to help him prepare for writing his own. He received the award last year, and spoke at the workshop about his process.

After I tweeted this, Gillian Hayes said that I should really post my proposal publicly, which she regularly does, so that any benefits that one would get from reading it would be experienced widely. So, I am taking her up on her advice, and not only posting the successful one, but also the unsuccessful one, and a bit of the story about how and when I wrote them!

===

I first applied for the CAREER award in June 2011. I remember writing that proposal largely on trains in Europe. I had flown to Prague, took a train to Vienna for the ICCHP conference, and then flew home from Munich. It was a fun trip, which Jen was able to join me on. It was a first trip post the birth of our daughter, who was about  6 months old at the time. Go-go amazing grandparents for allowing us to make this trip!

The title of my first proposal was called, “Human-Backed Access Technology: Enabling Accessibility Tools to Fall Back to Humans.” In it, I proposed many of the ideas that I’d work on for the next few years (continuing through today) around using on-demand crowd workers to fill in for where AI doesn’t work well enough in accessibility applications. Most of this work was a fairly straightforward extension of the approach I had used a year or two earlier for VizWiz.

Unfortunately, this proposal was rejected. I remember finding out about that just before Christmas, where I had just arrived at my parents’ house in rural Ohio for the holidays. It received okay ratings. I remember reading the feedback, and the primary concern was that crowd workers weren’t reliable, crowdsourcing might go away, and that crowds couldn’t be recruited fast enough for interactive applications. I’m sure there were other concerns, and I’m sure part of it was also just that there were many great proposals and not all could be funded. I also wrote this proposal in Word (my first and last not written in Latex), and so I also blame that poor decision.

Regardless, the rejection clearly mattered to me since 8 years later I still remember where I was when I got the news!  I also still take some glee in the fact that crowdsourcing now underlies nearly all of AI, and real-time crowd work continues to flourish (under various names and products).

I resubmitted again the following year (2012). I thought about not submitting, since there is a limit on the number of times you can submit (used to be 3 times pre-tenure, I think). This time I largely wrote and submitted it while at Towson University, where I was helping to run a program at the NFB Youth Slam, one of the outreach programs I pledged to support in my rejected proposal from the year before. During the day, I helped to run a programming course for blind high school students. At night, I sat in the hotel bar and wrote the revision of my CAREER proposal^[1].

The content remained similar, although I updated the title to reflect a UIST paper I had recently submitted:  “Closed-Loop Crowd Support for People with Disabilities.” Incidentally, the UIST reviewers didn’t like the “Closed-Loop” part of that paper, so we renamed the paper, “Real-time Crowd Control of Existing Interfaces.” This proposal also differed from the prior one because I had more preliminary results to report. By preliminary results, I mean we had essentially already built prototypes of everything that I was proposing to do. Another big difference was that I proposed various workflows that made sense for working collaboratively in real-time that made sense for important accessibility problems. Huge thanks to Erin Brady, Sam White, Kyle Murray, and Yu Zhong, in particular, for doing all this “preliminary” work. I don’t want this to be mischaracterized, there was plenty left to do… to some degree, I’m still working on the ideas I proposed here. But, the proposal was definitely stronger this time around.

I also remember where I was when I heard that this proposal had been accepted. I think it was in November 2012, I was in Pittsford, NY (Philip Guo’s one-time haunt, near Rochester), out to eat with my family. I got the email on my phone. I remember going back home and reading it in detail and being super-excited.

That’s the story of my CAREER award. I didn’t do most of the things that people say to do. I looked at one example (from Jake Wobbrock), I didn’t have anyone read my proposal, and I wrote them both at the last minute. Despite my Twitter presence, and large group, I have to actively push back again my tendency to be a research dreamer loner. For me the real problem is that I make a ton of mental progress rolling ideas over and over in my head, and I find that feedback from others disrupts that process. I want to go down weird directions, push ideas that others won’t get because they’re too ill-formed or that others think won’t work. If I tell you about an idea, and I don’t know you well, it’s probably a few months to a few years old in my head. Writing proposals is necessary but not my favorite thing to do, so I usually wait until the last minute so I only spend a limited amount of time doing that, freeing up time for parts of my faculty job that I enjoy (research, students, tweeting). So, shrug :)

Here are my two proposals, the unfunded one and the funded one:

“Human-Backed Access Technology” (submitted and rejected in 2011)

“Closed-Loop Crowd Support for People with Disabilities” (submitted and accepted in 2012)

I’ve since been on CAREER panels at NSF, and I can say that by far what works best for me as a reviewer is when people write something bold, interesting, and unexpected. Make them interested in the work, and excited to see you do it! Of course, other people look at other things (how feasible is it, how well did you describe your studies, etc.), but, that’s reviews. Personally, I find the interesting stuff more interesting, and try to write that way. Both of my proposals were pretty good (I think), but the second one was definitely more interesting.

Good luck!

This page and contents are copyright Jeffrey P. Bigham except where noted.
Blog posts are not intended to be final products, but rather a reflection of current thinking and/or catalysts for discussion, like tweets but longer.