PIMS Collaborative Research Groups

(Edited below, 09/06/2011)

PIMS Collaborative Research Groups are groups of

[…] researchers with a common research interest and with a common desire to collaboratively develop some aspects of their research programs. The groups may already exist, organizing joint seminars and workshops, making joint PDF appointments, or developing joint graduate training programs, but will have the potential to do much more, given resources and organizational structure through PIMS.

A CRG must have a “critical mass” of participating researchers from PIMS member universities, that is, the major universities in western Canada plus the University of Washington. The smallest group that I can recall had 8 faculty, but usually there are many more, in the 20-30 range, and the numbers only seem to keep getting larger. (Once in a while a CRG includes a few participants from other institutions, but this is not common, and in any case it is a very small fraction of the group.) Each CRG has one or several “leaders” who develop the proposal and coordinate the project.

The PIMS 2010 Annual Report (link to PDF) tells us that PIMS spent $220,846 on CRGs in 2010. This figure is for the calendar year, which does not coincide with the fiscal year, and presumably does not include major items such as postdocs or summer schools, which are reported separately in their own categories. There were 4 CRGs ending in 2010, 2 ongoing, and 2 starting in 2010. Between all that and what I’ve gleaned in the past, the actual total CRG budget might be about 100-130K per year, per group. (If anyone here has better information, please correct me.) The money is parcelled out into amounts designated for specific purposes, i.e. this much for each of this many postdocs, this much for each distinguished visitor (one per year), etc.; the rest of the funding for each item comes from the Discovery Grants of the participants. Additionally, each CRG gets about one BIRS workshop per year, again from a separate budget (this time BIRS).

According to the program webpage, the CRGs “create new research opportunities,” “enhance training programs,” “generate new ways of having its [PIMS’s?] scientific programs driven by its member scientists,” integrate, facilitate, create a context, as well as foster a variety of things.

What could possibly go wrong?

I’ve said before that this is not my favourite funding program, basically because too much of it revolves around politics rather than scientific merit. Well, whatever. Not every research support program has to please me. But I’m concerned about the ongoing trend to support “big science,” as represented by large groups or major new initiatives and of which CRGs are a good example, at the expense of individual funding.

The Discovery Grants budget in mathematics took the first large cut in 2007 and has been decreasing ever since, culminating in this year’s mess. The appropriate response from NSERC would be to raise our DG budget by 70% or more, bringing it in line with other disciplines of science and resolving pretty much all of our financial problems. Instead, we get the long-range planning exercise where, I suspect, we’ll be asked to devise ways to do more with less. Between that and the upcoming renewal of institute funding, which will require a new framework again as the institutes are being removed from the MRS program, I’m worried that more money will get shifted away from individual grants and towards the institutes or other forms of collective funding.

There are plenty of political arguments for such a move: the PR value of large initiatives, the matching funds, the money doesn’t look spread so thin when it’s handed out in large chunks. But from the point of view of science, it would be catastrophic. I’ve written posts already on why thematic institute programs can’t replace stable individual funding. PIMS CRGs, on the other hand, support many of the same activities (postdocs, visitors) that are normally funded by Discovery Grants, but are awarded to groups rather than individuals. Let’s talk about how that works.

(Sorry about the length. I wanted to write this out in enough detail so that a non-mathematician could understand the problem, and didn’t want to run a whole series of CRG posts.)

Collaboration and geography. Science policymakers who promote and approve funding for “research groups” might well be thinking of experimental disciplines (such as physics or chemistry) where research is often done in large groups led by senior scientists, based at a single institution according to the physical location and institutional ownership of the equipment. Or that’s the stereotype, anyway.

This is not the case in mathematics. We’re in the business of trading ideas and expertise, not sharing equipment or conducting proprietary research for our employer. There is no reason for our collaborations to be confined to a single institution or geographic location. Indeed, the sharp upward climb in the rate of collaboration in mathematics, as measured for example by the percentage of multi-authored research articles, correlates strongly with the development of tools (especially the internet) that make long-distance collaboration more feasible. Our natural research networks are scattered around the globe: long-term collaborators, researchers with similar interests, people with whom we know we can work.

Collaborations and forced alliances. There are institutional mechanisms to support this type of activity. Many institutes and conference centers do so by inviting collaborators to visit at the same time and providing them an opportunity to work together in person. The NSF Focused Research Groups program funds research groups, often spanning several institutions. The group must make a convincing case that there will be actual scientific collaboration involved, supported by evidence such as joint publications by group members. According to NSF program officers, the evaluation panels take this very seriously and reject summarily any “groups” consisting of several unrelated individual proposals stapled together. Given the size and research strength of the American math community, it is not terribly difficult to find research groups in the U.S. that satisfy these requirements. It’s worth noting, though, that the FRGs actually tend to be smaller than most CRGs (three or more researchers are needed, according to the FRG program description), in complete disproportion to how the number of eligible researchers in the U.S. compares to that in western Canada.

The PIMS CRGs are based on a different model altogether, one that has been developed out of necessity. An average NSERC Discovery Grant in mathematics is about 20K per year as of this writing, maybe less, and most of us have no access to other sources of research funding. In an earlier post, I explained how we use that money; the bottom line is that this is not anywhere near enough for a single researcher to hire a postdoc. Instead, we are forced to form ad hoc alliances where 2-3 people, sometimes more, combine their funds to hire a candidate who’s acceptable enough to everyone. This can work well if the faculty are actually engaged in joint projects or at least close to each other research-wise, but if not, the postdoc could end up falling through the cracks, not quite able to work with any of the several nominal supervisors. Jim Colliander calls it a “failure factory”, and for good reasons.

PIMS evidently disagrees. The CRG program encourages us to expand this paradigm to all spheres of our activity, including conference organization, research visits, postdoctoral and graduate supervision.

Working together, or not. For most of us, our natural research networks are too diverse geographically to be relevant for the purpose of forming a CRG, the latter being confined to PIMS member institutions. Even if we’re lucky to have a collaborator or three in western Canada, that’s not enough, because as I said earlier, it generally takes 15-30 faculty to form a CRG. Instead, we have to negotiate with local colleagues who work in a subarea of an area that has some other subarea that is tangentially related to some subarea of the area that we work in. We can only hope that they will listen to us when we try to explain why our postdoc candidate should be supported, or at least that they will agree to have our postdoc funded if they get the visitor money.

There are sensible ways of doing this. For instance, a CRG might start with a core group of several researchers who do work together. (This is where it helps to have a few local collaborators.) They pick the postdocs and organize the summer schools. Others are tacked on for the numbers and the geographical diversity. They’re more peripheral to the group and don’t benefit as much from it, maybe a conference invitation with all expenses covered, but then all that’s required from them is to allow their names to be included in a proposal, so it can be a fair trade. Or else a few senior researchers, mostly isolated otherwise, might decide to start a group. The purported collaboration would not necessarily pass muster at the NSF, but they each deserve to be funded and they can agree well enough on how to spend the money, so why not.

The problem with the second option is that it isn’t really a collaboration. The researchers involved would benefit more from having their individual grants increased instead by 5-10K per year, an amount comparable to their share of CRG budget minus postdocs and BIRS workshops, and from having the CRG share of BIRS workshops and especially PIMS postdocs returned to the general competition.

The problem with the first option is that the juicy part of it – membership in a core group – is not available to everyone, and it’s not scientific merit that makes the difference. It’s having several colleagues working in the same area. It helps if you’ve had some influence in the last few rounds of hiring in your department, but your CRG plans may just as well depend on the outcome of the hiring process at other PIMS universities where you don’t even work. (The expectation, I think, is that CRG leaders should be renowned researchers. However, there has been a case where a CRG leader was denied tenure a few years later.)

Between the size requirements and geographical restrictions, a CRG will often include faculty with quite different research agendas. This leads to more politics as they have to negotiate the splitting of the proceeds with no arms-length panel to adjudicate the relative scientific merit of each request. Sometimes, compromises are made. A conference combining several different topics might be acceptable to everyone, even if the different subsets of participants don’t interact much. Other times, it boils down to everyone’s relative influence in the group. It helps again to have a tightly knit group of departmental colleagues in your area.

Why is this a problem? The CRG program benefits, first and foremost, large research groups within departments. The more isolated researchers are pushed to the sidelines, formally not excluded, but in practice unable to get the same benefits from the program. A senior researcher who cuts a major figure in Canadian mathematics might be able to lead a CRG anyway, but a mid-career researcher may have to choose between being excluded from the program altogether and accepting the leadership of someone who works in a very different area and may well be a lesser mathematician, but was more aggressive in the last round of departmental hiring.

The general principle in science funding is that the wider you open the competition, the higher the level of the winning proposals. The CRG program is restricted to PIMS institutions to begin with, and then it is further restricted to groups that can muster sufficient “critical mass” while maintaining their locality. It’s not surprising that PIMS has already started to fund the same CRG areas more than once. PIMS in fact seems to be shifting further in that direction, supporting selected groups pretty much on a permanent basis (IGTC in math biology, summer schools in probability almost every year).

Associated with this is the loss of visibility. Outside of the CRG-eligible institutions, PIMS has been slowly dropping off the radar. I was surprised to learn over the years how many of my colleagues from Europe and the U.S. either weren’t aware at all that PIMS existed, or else thought that PIMS and BIRS were the same institution. This includes generally well informed scientists who have held administrative positions as department chairs, NSF officers or institute board members. By contrast, pretty much every senior mathematician I know has heard of Fields, CRM or BIRS.

The CRG program has eclipsed, and drawn funds away from, other PIMS programs that might well have been more valuable. I was glad to learn, way back when, that PIMS had a Postdoctoral Fellowships program. The PIMS support amount of 20K per year for each selected fellow can be a blessing for individual faculty, enabling them to make postdoc offers that their Discovery Grant alone would not allow. This is especially important for those of us who are isolated in our departments and can’t pool the money together with other faculty. Unfortunately, in recent years almost all PIMS postdoctoral fellowships have been bookmarked for CRGs. This defeats the main benefit of the program as I see it. According to the PIMS web page, the CRG postdocs “go through the same nomination and rigorous review processes.” Based on my knowledge of several non-CRG candidates who got rejected, I don’t see how I could have much faith in that.

This is not just a PIMS problem. Any similar program that funds “collaborative groups” that don’t actually collaborate is bound to have similar issues. NSERC had a short-lived “leadership grants” program in mathematics some years ago. I was approached once about joining a group under this program, submitted my paperwork, then didn’t hear back about it again for a long time. Eventually I asked, and was told that the group had removed my name from the application so that the number of participating faculty from each institution would be the same. I don’t know whether this was the real reason, and anyway in this particular case it was probably just as well; still, just the fact that it was a credible excuse says a lot about how a research group can turn into a political group where the need for equal representation (or proportional, or whatever) trumps scientific merit, shared (or not) research interests, or common sense.

Edited to add: I have been asked specifically about the CRG in Geometric and Harmonic Analysis, 2006-2008. Since the archived webpage does not list the names of participating faculty, I should clarify that I was not associated with that group in any way and did not participate in any of its activities. The group was mostly based in Alberta and had a research focus very different from mine.

1 Comment

Filed under research funding

One response to “PIMS Collaborative Research Groups

  1. Pingback: Weekly Picks « Mathblogging.org — the Blog