Sunday, January 27, 2008

The CRA blog on NSF Funding

If you aren't regularly reading the Computing Research Policy blog at the CRA, I recommend stopping by once a month or so. This month: find out about the latest NSF funding disaster (or "how the NSF and science in general got shafted in the budget process yet again") , including whether CDI will actually get funded (after we suckers researchers sent in 1300 proposals for about 30 grants...)

17 comments:

Kevin McCurley said...

Every constituency tries to exert influence on the distribution of funds from the government. Examples include health insurance for children, renewable energy, the V-22 Osprey, education, SLAC, Head Start, etc., etc., etc. We all have our opinions about priorities, but the democratic process is what ends up setting priorities.

The real question is - in the face of tight budgets, what are the priorities for funding of computer science research? What serves the needs of the nation sufficiently to justify more funding? I don't mean to imply that NSF is not deserving, but the country faces a lot of crises, and the CS research community needs to articulate a compelling set of priorities in order to justify increased funding.

Michael Mitzenmacher said...

Kevin,

Let me forcefully disagree with your comment. First, while obviously in this blog the main interest is on NSF funding (and more specifically NSF funding for CS), the CRA policy blog makes clear it was all of science that got shafted this round. Please read their description, and in particular you might read the opinion piece by Craig Barrett they link to, which I quote below:

Two years ago, the National Academies published the seminal study on U.S. competitiveness entitled "Rising Above the Gathering Storm." The study identified major shortcomings in U.S. investments in basic scientific research as well as in math and science education for our youngsters. The suggestions contained in this study were immediately picked up by the Democratic House Leadership as their competitiveness strategy and later by President Bush in his State of the Union message under his American Competitiveness Initiative. Legislation in the form of the America Competes Act was passed in the House and Senate in 2007, and it appeared the United States was finally going to move forward after years of neglect to increase investment in math, science and basic research. All parties agreed that our competitiveness in the 21st century was at stake and we needed to act.

So much for political will.


I recommend the rest of the piece as well.

In the specific case of CS, I believe both the theory community (in the form of the "SIGACT funding committee") and the broader community (in the form of the CCC) has really moved to "articulate a compelling set of priorities in order to justify increased funding", and we've actually had the NSF listen. But when the NSF and other science agencies get beaten down like this by our legislators, after repeated promises that there would be increases in funding and the potential damage to US competitiveness was understood, I find your comment that the CS community needs to do more to better articulate a compelling set of priorities completely misguided. Perhaps, instead, we might entertain the possibility that something is not working quite right in our current version of the democratic process?

Adam Kirsch said...

Perhaps, instead, we might entertain the possibility that something is not working quite right in our current version of the democratic process?

Mike, as something of a politics junkie, I understand your frustration with the government, but even I think this comment is a little over the top. (Those who know me will attest that when I say something is over the top, it tends to be really over the top :-) )

The last couple of budget cycles have been particularly contentious, with all kinds of weird political maneuverings on all sides. But in the end, if science in general is being consistently shafted (and I agree that it is), then it is not really because of politicians. Ultimately, the responsibility for prioritizing issues rests with the electorate. That funding for science has consistently taken a back seat to all of the other issues-du-jour during the last several years is merely symptomatic of this: the general public does not really view broad funding for science as a crucial priority. (And when I say crucial, I mean post-Sputnik crucial. Like, "if we don't get more science the bad people are going to take over our country in the middle of the night" crucial.) I think that's what Kevin was pointing out.

I know that you've spent a lot of time on improving the NSF's view of CS theory research, but I think what's going on with these budgets is part of a much bigger picture. Unless there is a significant shift in the electorate's attitude towards science, we can expect scientific funding to take hits like this on a regular basis. It follows that any fault here is not with the process, but with the voters. In particular, the expectation that elected officials, particularly congressmen, should be "leaders" is sort of naive; our whole system has been set up from day 1 to bind their actions to their constituents' opinions. (This is one of my major problems with Barrett's piece.)

Kevin McCurley said...

It's not at all clear to me what we're disagreeing about. Science did get shafted in the recent budget, but I can only conclude that while we made the case to NSF for funding in CS, we didn't make a strong enough case for funding to Congress and the public. For whatever reasons, Congress decided in favor of the other priorities for funding. There's no disagreement that this is a problem.

This simply highlights the need for computer scientists (indeed ALL scientists) to articulate a vision to the public and their policy makers. Since we are approaching an election, it's also a time when we should pay attention to candidates who will understand importance of funding basic research. There's definitely something wrong with our national priorities.

Some further reading on the subject:
AAAS

Anonymous said...

One of the problems is that the US supposedly has a "privatized" higher education system but research (even at places like Harvard with 35 billion endowment) is supposed to be funded by voters (taxpayers) who never get to go Harvard, send their kids to Harvard or really benefit in any other direct way from funding Harvard. What I do not understand is that if theory funding is so meager (about 6 million per year from NSF) why don't Harvard professors lobby their own university? It must be much easier to get a million from Harvard than from NSF. And if we really are a privatized system of higher education that has the right to discriminate (not let all people in) why do we expect government funding at all? Why should taxpayers who have enough problems give another penny to one of the richest "non-profit" (i.e. tax-free") organizations in America?

Anonymous said...

Uhh ... maybe because those institutions produce inventions and discoveries that drive the national economy?

Anonymous said...

You mean like inventing weapon systems and bombs?

Michael Mitzenmacher said...

Regarding Anonymous comments 5-7 (particularly 5): The NSF (and related institutions) do not fund Harvard. They fund research in the sciences and engineering, particularly basic research. The argument for why the government (that is, taxpayers) should fund basic research was laid out decades ago; as a starting point, I'd suggest reading the Wikipedia articles on the NSF and Vannevar Bush, and specifically for CS, I found the recent testimony at

http://www7.nationalacademies.org/ocga/testimony/Computer_Science_Research_in_the_United_States.asp

pretty interesting. (And yes, building better bombs and weapons, or promoting the national defense, is part of the argument for basic science, as is promoting the national defense by promoting the economy through science.)

Harvard is, generally, quite generous with supporting the research of its faculty. But Harvard is an educational institution, not a private research facility or corporation, so I don't agree with Anonymous #5's reasoning.

Anonymous said...

I have to agree with Kevin here. It's hard for me (as a computer scientist who counts on NSF dollars!) to make a case for why 99.9% of *any* CS research (not just theory) should be funded at the expense of, say, feeding the homeless (or, if you are Republican, military spending). How many important innovations in computer science *with practical impact to the average citizen* were driven by academic research? I am actually hoping someone will suggest some good examples.

Anonymous said...

Right. Nobody but dweebs use the Internet, after all.

Anonymous said...

Right. Nobody but dweebs use the Internet, after all.

I'm not an expert on the history of the Internet, but my understanding is that this is a bad example: development of the Internet was greatly aided by substantial government and industry support, and I would say that academics played an important but less significant role.

Besides, the Internet is now almost 40 years old. How about a more recent example?

The best example I can think of is public-key cryptography, more specifically RSA and Diffie-Hellman, which came solely from academia (though governments also came up with the idea independently) and have had measurable impact on the average person. This is a 30 year old example. Any more recent ones?

Michael Mitzenmacher said...

Anonymous forgets that quite a bit of the funding the government provides goes to training graduate students. A lot of the stuff you're labeling as advances by industry might not have been accomplished without the research training in graduate school.

But leaving that aside, because I'm too busy to look more examples up, I'll simply egotistically throw out my research as having an impact on the real-world that far exceeds the value spent on it by the government. (Mind you, I'm not saying that this is due to me personally -- I am, of course, just one link in the scientific chain.) My work on LDPC codes has been generalized and commercialized and is used in military projects (there's that one motivation, the national defense thing again) and in multiple civilian projects. (I'm not sure the average citizen uses it every day, but now or in the near future they might be without even knowing it.) My work on Bloom filters and multiple choice hashing has found application in routers (at least, that's what I hear from Cisco people, although I know of no details), which the average citizen uses every day, again possibly without knowing it. Even my more theoretical work on deletion channels, I hope, might someday lead to improvements in actual systems.

Of course, I think of myself as a "practical theorist". But I wouldn't deny the impressive value of my more complexity-oriented friends. Notions of expander graphs have numerous applications, such as in P2P networks, which I hear the average citizen uses. Primality testing algorithms, quite theoretical when they first were being thought of, have practical cryptography applications. Never mind just how our computational frameworks are now affecting sciences across the board.

I think there's a strong case that Google, Yahoo, Inktomi and many other Internet companies started in universities, and they've had a pretty big effect. (I think the person who gave the Internet as an example was speaking in the large, including efforts such as these.)

I think you're not getting a response because people think your statement is pretty ludicrous. At least, I do. I can point to lots of direct and indirect accomplishments of CS research funded by the NSF, and overall I think the return on investment has been tremendous.

Anonymous said...

Michael, you may be an outlier. =)

I've served on NSF panels and see what gets funded. I see the research people do in my department. Most (not all) of it is stuff that will generate papers but has little-to-no practical relevance. Of course this includes fantastic theory stuff that I like to see funded but couldn't justify to an average citizen. It also includes terrible stuff that pretends to have relevance but is really just a way of generating LPUs. (Are you seriously claiming that most academic research on P2P networks has any bearing on what gets used in practice?!)

I'm not saying that nothing is worth funding, and I'm not saying that academics shouldn't have the freedom to work on whatever they like. But scan through 100 random NSF grants (and don't limit yourself to people from the top 10 schools) and justify why the US government should fund more than 25 of them.

Anonymous said...

Of course, you are suggesting that if a quarter of the grants turn out to be useful, any fool could have known which quarter, in advance. We suffer, if anything, from too much short-term thinking already, too much emphasis on things whose practical utility is obvious. I suppose those who don't think TCS funding can usually be justified would defund mathematics research almost entirely.

Anonymous said...

Anonymous 9 here.

Let me first concede the point: there is plenty of useful CS research that should, in an ideal world, be funded.

However, the main points I am trying to make are:
First, Kevin is correct that "every constituency tries to exert influence on the distribution of funds". Let us not think that we are acting "for the benefits of society" when we are really just acting in our own self interests.

Second, the question is not "should CS research be funded" but rather "should CS research be funded at the expense of XXX" where XXX might be "biology research" or XXX might be "feeding the homeless". You can make the case that CS research is more important than military spending or paying down the national debt, but I never hear the argument phrased this way.

This will be my final post on the matter; I don't want to continue wasting everyone's time. =)

Anonymous said...

FYI, this talk is of utmost relevance to the discussion going on here:

Anonymous said...

The value of theory as a whole can be defended in the strongest of terms: RSA, Elliptic curve crypto, linear programming, internet algorithms, the coding and hashing work of MM, etc. Yet at the same time this list (if it were to be completely filled in) would give us a good indication of certain subfields within TCS that have not been particularly successful in producing valuable results either economically or in terms of leading to further developments outside the given subfield.

In those cases it behooves us as proponents and evaluators of research proposals to question the fruitlesness of a given area and move away from it if there are no reasonable returns on investment.