Engineering Science - Skills

Graduate Education at Stanford

2008-04-16T02:41:00Z

Mark Horowitz, Associate Vice Provost for Graduate Education at Stanford, gave a talk at Google a few weeks ago about some of the things Stanford is working on to enhance the quality of its graduate programs. After many years (a decade or more?) of having no senior leadership with responsibility for graduate education at the university level, Stanford has finally created an office of graduate education. Mark's talk was an overview of the initiatives coming out of the new office.

Stanford's new activity in graduate education is based on an internal report. The basic thrusts are

more interdisciplinary education,
greater diversity, and
leadership training.

These are all fairly standard ideas, but a few things struck me about Stanford's approach.

Stanford has a ridiculous amount of financial resources. It appears that the university is backing this initiative wholeheartedly. This is not just a few seminars cobbled together by an underfunded graduate or postdoc office - it looks like the real deal.
Stanford has amazing courses throughout the institution, but usually access to courses in one department is limited to students in that department. The new Office is doing the smart thing of trying to better leverage existing resources. To give a sense of how serious the university is about this, they are talking about things like moving the Law School from a semester to a quarter system (maybe the other way around? this is from memory) so that students outside the Law School can more easily take graduate courses and vice versa.
One problem people have with taking classes outside their department is that it's hard to find appropriate classes in other fields because there are so many classes out there. The catalog is paper based (!) and there is not much descriptive material on courses. Stanford has set up a web site called CourseRank, which is an exercise in using collaborative filtering to help students find relevant courses outside their departments. I suspect that CourseRank may be an exercise in trying to set up a social networking site for the sake of cool points, but the idea seems sound. One lower tech and much cheaper way to accomplish some of the same goals would be to do some basic mining of data from the registrar. Looking at past cross-disciplinary enrollments would be a simple way to provide students with suggestions for possible good courses to take outside their departments.
Stanford is setting up career development as formal classes during the summer and before fall quarter (the Stanford Graduate Summer Institute ). Students have to apply, and in some cases pay a fee. Having an application process is smart - it makes the course into a desirable thing that you have to compete for rather than a freebie extra thing that people have to be dragged to. The fee accomplishes the same purpose, plus provides resources to sustain the effort.
Career development workshops are being pitched as "leadership training". It's the same thing, but "leadership" is much sexier and probably easier to sell. I suspect that because of the word's additional usage as "leader in a field", "leadership" has positive connotations even for the most unreconstructed students-must-always-be-at-the-bench faculty.
Stanford is trying to set up mentors for students outside of the regular faculty advisors. The vision appears to be pairing students with area professionals, e.g. people at dot-coms, biotechs, etc. Definitely a good way to provide some guidance that many faculty are ill-equipped to do.

All in all a good effort. It's especially promising because places like Stanford tend to inspire places that want to be like Stanford. I'm looking forward to seeing the results and to seeing other universities try to follow in their footsteps.

Better Adaptation

2007-08-10T13:26:00Z

If the people with skills that are the most valuable outside of academia leave universities in disproportionate numbers (as I think they do), students will miss out on what they have to offer, and there the risk that academia will grow progressively more insular. How might one counteract this effect?

One straightforward remedy is to create more opportunities for interactions between academia and industry. There is already some work in this area in the form of industrial postdocs. I think these kinds of positions can be effective mechanisms for helping people transition to industry positions, but when used as such, their impact is limited. Unless the postdoc returns to academia (and I would guess that a relatively small fraction do), only one trainee really benefits.

It's moving people in the opposite direction, from industry to academia where they can teach, that could really make a difference. This kind of thing happens informally on occasion - a famous retired luminary from industry will get some office space in a department and show up from time to time - but to really maximize the benefits, one would need to recruit people still active in industrial research and figure out ways to have them impart their skills and knowledge (particularly in areas less prevalent in academia) to a broad range of students and faculty members.

I think such a program could be run on a small scale relatively inexpensively: a set of teaching fellowships that could serve the role of a sabbatical for people in industry. In addition to being able to offer some unique courses, departments would benefit by gaining relationships with corporate labs - collaborations could mean new funding sources as well as jobs for promising students. Companies would benefit from potential technology transfer and having their people learn new skills, so everybody wins.

The Importance of Communication Skills in Science

2007-07-27T16:58:00Z

The Poincare conjecture was until recently one of the great unsolved problems in mathematics. Stephen Smale proved the conjecture for 5 or more dimensions in 1961, Michael Freedman proved the conjecture for 4 dimensions in 1982, and Grigori Perelman proved the final case for 3 dimensions in 2003. All 3 mathematicians won Fields Medals for their work (though Perelman declined the award).

Smale's and Freedman's proofs were widely lauded, and both have since landed prestigious appointments: Smale was a professor at Columbia and then Berkeley; Freedman was a professor at UCSD and is now at Microsoft Research. Perelman's proof, in contrast, has generated considerable controversy; he is currently unemployed and lives with his mother in St. Petersburg, Russia.

A fascinating account of Perelman's saga recently appeared in The New Yorker. The gist of the story is that rather than publishing a formal proof of the Conjecture, Perelman posted a series of sketches to arXiv. Other authors subsequently refined the sketches into a full proof. Because these authors did not understand some parts of Perelman's sketch, they inserted some of their own methods in places, and as a result have claimed credit for the final proof. Perelman has renounced mathematics altogether over the dispute.

Authorship is particularly relevant in this case, as the Clay Mathematics Institute has offered a million-dollar prize for a proof. Because Perelman did not publish anything in a peer-reviewed journal, it is not clear that he is eligible.

I think that many researchers believe that if the science is good, everything else will follow. Perelman's case is a cautionary tale: he has clearly done brilliant work, but his mode of presenting his results has cost him dearly.

Effective Teaching

2007-06-04T16:32:00Z

There's an interesting piece in today's Chronicle of Higher Ed on the effectiveness of regular quizzes on learning. Basically, it appears that the act of recalling information (as for a quiz) reinforces the memory of that information. Quizzing people soon after they learn something produces pretty big improvements in their long term retention of that information.

That's useful to know, and it agrees with my own positive experiences using regular, short quizzes in the classroom. Quizzes have the additional benefit of forcing people to keep up with the material rather than trying to cram it all in at the end. Since class materials are typically cumulative, mastering material as one goes along means that one is able to better understand new material in the context of recently learned material.

Other simple things can make a big difference, too. I taught honors multivariate calculus most fall terms when I was at Dartmouth. There were 3 sections of the class, each taught by a different person. We all shared a syllabus, midterms, and finals, but apart from that were free to teach the course however we wanted. One thing I did that I thought was particularly effective was to spend about 2/3 of the class lecturing and then for the remainder of the class, pair up the students and have them work on an in-class exercise that made use of the concepts from the lecture. My sense was that it worked really well because the slower students ended up getting individualized attention from the faster students, and the faster students solidified their understandings by having to explain what they had just learned. In the end, my students invariably scored higher than those from the other two sections on the midterms and the finals, a fact that I attribute to the pairing exercises. Not by a few points, either - the difference in the class averages was often on the order of a letter grade. I'm sure this kind of thing is well-studied, but I was just making it up as I went along.

Here's the thing I'm wondering: there is a ton of research in education. Why do so few of the findings make it to the classroom? I don't know of any institutionalized mechanisms for extracting the most salient findings from the education literature and disseminating them to people who teach at the university level (sure, there are little seminars and special classes here and there; I'm talking something that goes to everybody). Why is that? Is there no reward structure in place for education researchers to disseminate their findings beyond education journals (where they will be read by a handful of other education researchers)? I suspect one could make tremendous improvements in S&E education simply by better publicizing things that are already known.

In fact, the whole enterprise could probably be done for maybe a few hundred thousand dollars given an enterprising faculty member with a bright master's student or two. Here's a great master's thesis idea:

Put together a review paper on what's known about the most effective ways of teaching college-level science / math / engineering.
Extract from that paper a 10-page executive summary.
Talk nicely to a foundation to get funds to print and widely disseminate copies to grad students, postdocs, and junior faculty.
Bask in fame.

Bubble Benefits

2007-05-20T00:46:00Z

Slate's Daniel Gross has a new book out, Pop! Why Bubbles are Great for the Economy. It's an interesting argument: while bubbles are wasteful, oftentimes (1) there are a few enduring successes, and (2) even the failures can leave behind valuable infrastructure that then gets consolidated and reused by sounder post-bubble businesses.

EBay, Amazon, and Google all survived the recent dot-com boom. Moreover, there was a huge buildout of telecommunications capacity that is finally starting to be used effectively. Less obviously, bubble times can also result in "mental infrastructure." One form this takes is fundamental changes in consumer behavior. Huge amounts of advertising dollars were spent in the 90's convincing people to shop and invest online. Now that people accept that these are reasonable things to do, new businesses benefit.

A second form of mental infrastructure is training for service providers. The dot-com era generated new programming practices (e.g. agile methodologies, design patterns), a wealth of documentation, new publishers (O'Reilly, Pragmatic Studios), schools, and huge numbers of trained software engineers, designers, testers, project managers, and on and on.

It's this latter piece that's of potential interest for scientists and engineers. Academia is incredibly slow to react to changes in the market for a whole host of structural reasons, and as a result, there are substantial excesses of PhDs cranked out in many fields. Bubbles can offer an out for these newly minted graduates by providing for some a relatively easy direction in which to make a lateral career move.

Right now there are at least 2 areas in which there is bubble-like activity: Web 2.0 companies and quantitative finance. I'm currently in Portland at RailsConf, and it's a bit of a Web 2.0 developer frenzy. Dozens of new little companies are hiring, and there is buzz about all sorts of interesting projects going on. If you're smart, Ruby on Rails is easy to pick up, and it's a pretty good plan B. As for quantitative finance, over on jobs.phds.org there are a couple dozen new ads for quantitative analysts every week. A fair number of the ads are for mathematicians / physicists with no prior finance experience, and the pay well into 6 figures, even for the entry level positions.

Now neither of these areas is going to give you tenure-level job stability. The bottom is going to fall out sooner or later in both; probably sooner with the hedge funds. However, in both cases the pay is substantially greater than what you'd get as a postdoc (by a factor of 3 to 5), and you'll acquire skills that you can use in whatever you do next. Which makes more sense economically? 3-5 years as a postdoc earning $35K and preparing for an incredibly scarce tenure track job? Or the same amount of time earning far more as a developer or a quant? There's considerable uncertainty about next steps in either case, but at least in one you're getting paid.

I think the real challenge for PhDs is more of an attitude adjustment, namely accepting that:

your education is a sunk cost,
research is not necessarily the One True Path,
you are still a fine person even if you don't write papers for a living, and
you always have to learn and adapt your skills.

The Scientific Communications Act of 2007

2007-04-09T19:36:00Z

There's an interesting new bill working its way through the House: The Scientific Communications Act of 2007. (A tip of the hat to the fine folks over at ArsTechnica for cluing me in to its existence).

The bill allocates $50 million over 5 years to the NSF to improve the communication skills of S&E graduate students. The rationale offered in the bill is pretty straightforward:

Grad students often don't get training in communication skills, and these skills would help scientists talk to:

the public (i.e. increase public support for science)
policy makers (i.e. help scientific findings be incorporated into public policy), and
business leaders (i.e. help businesses use science to create new / better products)

All very sensible stuff, it seems to me. I am left wondering a few things:

First and foremost, how did this bill come to be? The lousy quality of graduate training in communication skills doesn't seem to be the kind of thing that most members of Congress spend their time worrying about. I wonder if this is the work of a AAAS fellow? Is there a good conduit for similar ideas to make their way up to the Hill?
Is this level of micromanagement unusual? Or does Congress go in and make NSF do little projects like this all the time? If the former, then I have rather mixed feelings about the bill. Despite the fact that I think communication skills should be taught, I'd be concerned that a different Congress might impose other, less desirable missions on the NSF, say, a set of grants to shore up the foundations for Creation Science.
Is this likely to go anywhere? Or is this just some kind of gesture?
Finally, why doesn't the NSF do this already? This seems like a no-brainer, and it's entirely in the self-interest both of the NSF as an institution and of scientists, since better public support of science presumably means more funding.

I have a couple of contacts on the Hill (not very good ones, unfortunately) will see if I can track down some answers. If you have any pointers to good sources, please contact me!

Stress of Science, Science of Stress

2007-02-23T19:00:00Z

I am scheduled to give a seminar in 2 weeks on a topic that I have less familiarity with than I'd like. The people from whom I am supposed to get a crucial data set for the talk aren't returning my calls. My backup plan has been scooped by a seminar in the same series on Monday. So I'm a little panicked. Not too panicked, mind you - I have learned from long experience that I will eventually figure something out, and fear has proven quite an effective stimulus for creative thinking in the past for me. But it's enough to cause all the physiological effects of stress to kick in.

Back when I was an undergraduate, I had some vague notion that work as a researcher would be motivated primarily by a quest for a state of flow. That's definitely there, but it is frequently punctuated by moments of terror. Research careers can be quite stressful, even in the best of times. Experiments can go awry, conference deadlines can sneak up, and there is always last minute lecture preparation.

I recently finished an excellent book by Robert Sapolsky on the physiology of stress, Why Zebras Don't Get Ulcers. I highly recommend it.

Why Don't Zebras Get Ulcers?

First, it's a great piece of science writing. Sapolsky manages to keep things lively and engaging without dumbing things down. Part of what keeps things moving along is his use of an informal tone, something that unfortunately is drummed out of a lot of people by writing endless journal articles. The book is a lot closer to sitting in an undergraduate seminar with a friendly but brilliant prof than reading a bunch of journal articles. He's also generous with credits - I discovered that a friend of mine did an undergraduate internship with Sapolsky because he spent a half page describing the work she had done. He mixes established knowledge with informed speculation, being careful to distinguish between the two.

Second, the subject matter hits pretty close to home. Having spent the past 20 years or so in fairly high-stress environments, I've tended to brush aside suggestions that I try to adjust (sorry, Mom!). Vague notions that stress is "bad" somehow just don't do it for me. Sapolksy, in contrast, spells out in grim detail exactly what chronic stress does, and believe me, I'm listening now.

Two effects of stress are particularly relevant to researchers:

Chronic stress wreaks havoc with the hippocampus, which plays a key role in memory and learning. So frantic cramming can, in the long term, reduce your ability to retain information.
Chronic stress is linked to depression, which in turn can reduce your ability to produce by sapping your motivation.

There's an interesting policy implication: it suggests that universities might do well to invest in stress-reduction / mental health services for researchers. Given that a sense that one is not in control of one's work is a major risk factor for depression (google "learned helplessness") and other stress-related morbidity, graduate students and postdocs could be some of the biggest beneficiaries.

A recent survey at Berkeley suggests stress-related mental health issues are startlingly common among graduate students:

"In the last 12 months, 45.3% of respondents had experienced an emotional or stress-related problem that significantly affected their well being and/or academic performance."
"9.9 % of respondents seriously considered suicide in the past 12 months."

Fortunately, remedies are fairly straightforward and inexpensive. Exercise, for example, has big benefits, so simple things like intramural sports for grad students and gym access for postdocs can help a lot. I learned at a recent conference that Vanderbilt has hired a full-time counselor for its postdocs, and apparently it's been a very successful experiment.

Another interesting thing I discovered in Sapolsky's book: there is interesting evidence that men and women's responses to stress are fairly different. The well-known "fight or flight" response is apparently a better description of the male response to stress than the female. The female response is characterized as "tend and befriend":

In particular, they propose that females respond to stressful situations by protecting themselves and their young through nurturing behaviors--the "tend" part of the model--and forming alliances with a larger social group, particularly among women--the "befriend" part of the model. Males, in contrast, show less of a tendency toward tending and befriending, sticking more to the fight-or-flight response, they suggest.

This work dovetails nicely with some research by Anne Preston on why people leave science. Preston found that female graduate students who did not have a mentor were much more likely to drop out of graduate school than those with a mentor. Intriguingly, there was no such effect for males. Tend-and-befriend provides a potential explanation: a mentor may play a more important role in women's mechanisms for coping with the stress of graduate study than in those of men.

This is potentially good news, if true: the Sigma Xi Postdoc Survey found that women who worked for female PIs were substantially more likely to consider their PI to be a mentor than women working with male PIs. As the ranks of women increase in the professoriate, we may well see a virtuous cycle: as female students and postdocs have greater opportunities to find a female mentor, retention rates may increase, which in turn could lead to further increases in women in the faculty.

Getting Research Ideas

2007-01-24T15:53:00Z

How does one come up with good research ideas? Grad school teaches you a lot about how to pursue them once you have them. You also learn a lot of techniques for evaluating your ideas so you can sift through and find good ones. But are there things you can do to help ensure that your pool of ideas actually includes some good ones?

Seth Roberts, a psychology professor at UC Berkeley, has a fascinating essay about how he comes up with good ideas. His system is not for everyone - he obsessively records dozens of measurements of himself and periodically sifts through the data. Because he studies appetite and weight control, he has a lot of opportunities for self-experimentation, and he seems to take them all up: an all-sushi diet, drinking gallons and gallons of water per day, shots of olive oil, and more.

This past weekend I attended the NC Science Blogging Conference where Jean-Claude Bradley from Drexel suggested another interesting approach. He searches Google Scholar for phrases like, "what is needed now", "what is missing is", "there is a pressing need", "what is now needed", "needs to be synthesized", "pressing problem", etc. For example, "there is a pressing need" 2005 chemistry turns up phrases for things that people have identified in (mostly) chemistry papers in 2005 as being important.

Postdoc Leadership Mentoring Project

2007-01-11T19:14:00Z

On the subject of cultivating leadership in scientists, the following showed up in my mailbox recently:

The NPA is pleased to announce the Postdoc Leadership Mentoring Project, sponsored by the NPA and the Alfred P. Sloan Foundation. This innovative program seeks to connect experienced leaders in the postdoctoral community with individuals and institutions interested in establishing postdoc offices and/or associations at their home institutions. Experienced leaders will serve as mentors to those forming new organizations or working to revive past organizations.

...

At the Annual Meeting [March 30-April 1], recipients of the awards will participate in a special session on establishing postdoc offices and/or associations and will connect with their mentors or mentees face-to-face. Following the meeting, mentors and mentees will continue to strengthen their relationship through participation in monthly teleconferences. Finally, some award recipients may be eligible for site visits to their research institutions by a team of experts to further the growth of their new postdoc organizations.

I think this kind of thing can potentially have a big impact on the quality of the training scientists receive.

Notice who initiated it: postdocs, not senior scientists. That's good and bad: good in that postdocs are taking the initiative (showing leadership in the process), but bad in that they have had to do so.

Kudos to the National Postdoc Association.

You can find out more here

Tip: Find Your Citations

2007-01-02T19:37:00Z

Quick tip: Want to see who is citing your work? Use Google Scholar!

Find your papers by searching for your name (if it's distinctive) or by searching for the titles of your papers. Helpful hints:

Use the exact form of your name that you use in papers and put it in quotes. In my case, "Geoffrey M Davis" (without the quotes Google can return papers for which one author has the first name Geoffrey and a different author has last name Davis).
Eliminate erroneous matches by restricting the search to the subject area of your research (look under Advanced Scholar Search).

For each search result, you'll see a link entitled "Cited by N" (where N is the number of citations) underneath. Click the link to see a list of papers citing your work.

Voila

This could be a big help if you're looking for a postdoctoral position: people who have cited your work clearly have similar interests and are already familiar with your work.

Leadership in Science

2006-12-06T15:30:00Z

Like Peter, I was a problem grad student, spending much of my time engaged in non-research projects. I did student government. I helped to start a community-service focused undergraduate orientation program. And I took a semester off to write virtual reality video games (long story).

The student government and orientation program were unusual activities for me, since I had been a pretty hardcore nerd in college and high school. The inspiration for these activities came from two sources. A public policy class taught by Bruce Payne changed my perspective on what it meant to be a leader. The muddy idea of leadership in my 19-year-old brain involved people on stages in front of large crowds, not researchers in labs or college kids in class. Leaders were people very different from me. Payne's class changed that idea in a big way. He made me realize that leadership wasn't just an activity of distant, famous people working in a national or global arena. Rather, it's about taking the initiative to make change in your own world. And, just as importantly, the necessary skills aren't innate -- they can be learned. My work with an innovative outdoor education program gave me the opportunity to try out some of the ideas from Payne's class for myself. These two experiences were among the most important things I gained from my time as an undergraduate.

The reason I bring this up is that last night I attended a little reunion of people who had been a part of Payne's program at Duke. The people in attendance described their careers -- all had gone on to do interesting things. One thing struck me, though: I was the only scientist there.

There is a huge need for leadership in the sciences. Stem cells, global climate change, and other areas in which science ends up on the national agenda are obvious places. But I think the places where better leadership skills would make the biggest difference are less visible and more mundane: Running departments. Administering programs at funding agencies. Running labs. Teaching.

We as a community don't do a lot to prepare younger scientists to take on leadership roles. There are few rewards and little recognition for leadership activities, except possibly for very senior people (quick: name an award for leadership in science that would ever go to someone young). Leadership positions are often viewed as not particularly desirable or even as a step down (are people lining up to be the your next department chair? your dean?). And there is relatively little training in leadership, especially for junior people.

The only program I'm familiar with for cultivating leadership in young scientists is the congressional fellows program. I don't know of anyone who has done that program who has returned to an academic career (though I am sure that such people do exist), so I think its impact on academic science is fairly limited. (Project Kaleidoscope also appears to do some leadership training, but I don't know anything about the program or anyone who has ever participated - perhaps it is small? or focused at a narrow audience?)

The attitude I sensed while I was an academic was that leadership activities were not for younger people. Younger people should keep focused on science. Only when you get tenure, when your work slows down, when you can't do science 24x7 should you engage in such things. To see the problem with this approach, turn it around. What kind of science would we have if there was no training and people were discouraged from engaging in any real research activities until they were well into their 50's?

At the reunion we had a brainstorming session about ways we could engage current undergraduates. I suggested trying to work with graduate students as well, but was met with blank stares.

There were a lot of MBAs present, and some suggested working with undergraduates to put together business plans for interesting projects or nontraditional companies. They described a recent group of students who had worked with the city government and had raised money to transform an abandoned lot in a bad neighborhood in Durham into a sports field lighted for nighttime play; other recent graduates had just started an import business designed to help unemployed South Africans in Capetown. The ideas that were bandied about were intriguing, but not really right for grad students or postdocs. The discussion got me thinking, though, about some variants that might work.

While it would be nice if more scientists had skills for changing the world, I think a more attainable goal is to try to increase the ranks of those capable of making change within their institutions. What if, instead of pairing MBAs with undergrads to write business plans, one paired, say, an interested professor and/or someone with a background more like mine or Peter's with some grad students / postdocs to put together a plan for some sort of institutional change? A postdoc office, say, or a new institution-level training program. Instead of going to a random dean with some half-baked idea and being surprised when s/he says no, one could go to the right person with a detailed plan in hand, complete with operating costs, sources of personnel and funding, and ideas of impacts on other programs. I think it would make a new idea a much easier sell, plus it would give people some valuable experiences negotiating the labyrinth that is the modern research university.

What do people think? Refinements? Other ideas?

Writing Effective Grant Proposals

2006-11-17T14:58:00Z

I have been reviewing a few grant proposals recently. One in particular struck me: a poorly written proposal that described what I thought was a pretty good idea. In the end, I couldn't recommend that it be funded -- there were just too many unanswered questions. I imagine it will be revised and resubmitted and that the reviewer feedback will result in improvements.

I learn a lot from reviewer feedback, and having the occasional proposal get dinged really drives some lessons home. But on the whole having your stuff rejected is a really slow and unpleasant way to learn.

One way to reduce your odds of rejection is to make sure your proposal addresses Heilmeier's Catechism, a set of questions that George Heilmeier, the director of ARPA in the 1970's, used to ask of all new research programs:

What is the problem and why is it hard?
How is it solved today?
What is the new technical idea, and why can we succeed now?
What is the impact if successful?
How will the program be organized?
How will intermediate results be generated?
How will you measure progress?
What will it cost?

Questions 1-4 are the kinds of things you need for any good paper or talk, and 8 is obvious. The larger your project, the more important questions 5-7 become. They provide the funder with some assurance that you will be taking steps to ensure that you do not stray too far off course.

Tim Finin, on whose blog I discovered the Catechism, learned the questions in a graduate course. In contrast, grad school taught me pretty much zero about how to write a good grant proposal -- I don't think the subject was ever mentioned. Apparently I'm not alone: 37% of the postdocs surveyed in the Sigma Xi Postdoc Survey reported that they received no training in grant writing. Most of those who did receive training indicated that it had involved informal, on-the-job training rather than a formal course or seminar.

(An aside: I suspect that things are both better and worse than the Sigma Xi numbers suggest. The “no training” number is probably a bit high because some of the respondents are fairly new postdocs who just haven't received experience yet. On the other hand, I would bet that part of the “informal, on-the-job training” corresponds to people doing legwork for a PI's proposal without gaining much experience with bigger picture grant-writing issues, and part corresponds to people writing fellowship applications, again, without necessarily receiving much feedback.)

The relative dearth of grantsmanship training is is pretty surprising. Unless you are independently wealthy, some fraction of your time as an independent researcher is going to be spent convincing other people to invest resources in your ideas. This is true whether you are in academia, a government lab, or industry. If you don't know how to write a good proposal, it's going to be tough to move your own ideas forward.

I have collected a set of grant writing resources on phds.org. The resources on effective paper writing are also useful.

Have you learned about grantsmanship as a grad student / postdoc? What resources have you found helpful?