Tuesday, February 13, 2024
Context: In my role as division director of IIS, I’m sending out a short message to the IIS mailing list on the Second Tuesday Every Month (STEM). Here’s the installment for February 2024.
Hey, can we talk about annual reports for a minute? NSF requires that work you do on funded projects be documented through a report each year of a multiyear project and one at the end of the project. The details of what is needed in the report are available in the materials you are sent, so I won’t go through them here. There’s a lot you are asked about, but the process is quite a bit lighter-weight than what is used by many (most?) other funding agencies. NSF really wants to know (1) that the investments are proceeding as expected, (2) if there’s stuff you’re doing that is exciting and we can join in on that excitement, and (3) whether there are things we can point to that help NSF show that it is making good on its fundamental mission.
It's a bit of an interruption, but it’s not a huge lift and the communal benefits are significant.
Unfortunately, CISE, the Computer and Information Science and Engineering directorate we’re a part of, is an outlier at NSF in that our reports are considerably later than those in other directorates.
NSF has tried various things to encourage PIs to get their reports in on time. For example, NSF will not make any new awards or allow any extensions to a PI’s (or Co-PI’s) other awards while any report is due or under review. (In fact, the management system discourages me from making any decisions, even declines!, on proposals that share PIs or Co-PIs with projects with late reports.) However, this rule doesn’t seem to help much. Recently, NSF experimented with a pilot program in which a project’s funding is paused when a report is overdue and hasn’t been submitted and approved. That got people’s attention! But, I’m hoping we can get reports in more quickly so that heavy-handed interventions aren’t needed. Please help!
One point of confusion I’ve heard is that people get messages that use the terms “due” and “overdue” and they are not sure how to interpret the associated dates. It’s not super complicated. The idea is that you should report your activities on the anniversary of the start of the project. Instead of requiring that you do so precisely on the anniversary date, you get a “window” to submit the report. (As I’ve gotten older, I’ve been celebrating birthdays and even family holidays this way… any convenient time during a close-enough window. Hanukah is a particularly challenging one to schedule because it moves around and there isn’t always a long weekend that it corresponds to. Anyway…) The start of the window is when the report is “due” and the end of the window is when the report is “overdue”. Please get things in during that window, ideally on the early side so your cognizant program director has time to read through and approve it (or provide feedback). Thanks for your consideration!
One bit of feedback I got about my last message is that I described an NSF policy in my own words. It’s worth pointing out that my phrasings are not the official policy, so I always try to point you to the official document. People work really hard on these documents making sure they say what NSF wants them to say. Oh, relevant to that, a new NSF Proposal & Award Policies & Procedures Guide (PAPPG) has been released! It takes effect in May:
https://new.nsf.gov/policies/pappg/24-1/summary-changes
Also, two months ago, my message included a puzzle I created:
I have a piece of furniture consisting of n drawers (in a single stack). Each drawer holds one item of clothing. I wash all n items of clothing and then put them away in a random order. When I put away item i, which goes in drawer i, I open drawer i and close all the drawers above it (to make sure I have access to drawer i). When I’m done putting away all n items, what’s the expected number of drawers that are open?
I got some nice solutions from STEM readers, a few of whom tried to get some help from chatbots. (The chatbots were, as chatbots often are, weirdly astute and also strangely off the mark. I think all got the right answer but with explanations that didn’t actually make sense.) Anyhow, here’s my solution, if you are interested:
By linearity of expectation, the expected number of drawers open at the end is the sum of the probabilities of each individual drawer being open at the end. The very bottom drawer will definitely be open, since no other drawer opening will cause it to become closed. The drawer right above it will be open with probability ½, because it will get closed if it is selected before the bottom drawer, and it’s equally likely for either of those drawers to be chosen first. Similarly, the drawer that’s third from the bottom has a probability of ⅓ of being open at the end. This pattern continues all the way up. So, the expected number of drawers open at the end is 1 + ½ + ⅓ + … + 1/n, which is exactly the nth harmonic number, which grows like the natural log of n.
I thought that turned out rather well.
Until next time!
-Michael