Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.
Research Strategies Geoffrey Fox November 2 2007
Research Strategy • Always aim to do world-class work and structure research strategy and proposals to reflect this • It is very unlikely that your nifty idea is new unless you are really familiar with field • Don’t start serious work unless you know current state of the art • It usually takes a long time to develop a fundable proposal • Need to nurture future areas while “living off” current areas • Currently I am developing multi-core research and I suspect it will dominate over distributed systems (Grids) 3-5 years from now • It took me 4 years from co-organizing a computational earthquake science meeting in 1997 to obtaining NASA funding • Several NASA projects funded since 2001 but all NSF proposals in this area turned down. • NSF proposals were technically as strong as NASA ones but different reviewer base • Note communities are surprisingly distinct; NSF does not acknowledge our work even though NASA gave it accolades • Sometimes it doesn’t work out: in 1997 I developed pretty innovative web-based Crisis Management/Collaboration system; all follow up NSF/NASA/DoD proposals turned down; finally got a DoD grant in 2005 in a different area but based on contacts I made from previous work • The rejected proposals were very good like accepted ones but not in areas reviewers related to as most work in this area from industry; • Some areas are hard to fund!
Collaborate • US Funding agencies love collaborative multi-institution proposals • Easier to get 20% of a $1M proposal with 5 institutions than one $200K proposal • 20 years ago I got a Caltech only proposal for $1.8M/year (mainly people) – such days have past for me • Note very unlikely (statistically) that best work in any area done locally so expect to need (inter)national collaboration • Only prepare proposals with only people from your institution if all components are world class and perceived to be world class • Build long term partners; today’s research colleague is tomorrow’s program manager at Darpa • Pursue an agenda identified with you in a collaboration • Once I developed an innovative technology; nobody took it seriously as I was collaborating with a really great researcher in that technology area; key people assumed he did work and there was nothing he could do to change this perception • In Grid area, do not improve Globus; only Argonne/Chicago/USC can get credit for this and there is nothing they can do to give you credit • Obvious lessons for students of well known faculty `
Know your Funding Agencies and their review strategy/peer review community • US agencies below: Need a Chinese version! • NSF: The research and education community • NIH: Not an expert but in between NSF and NASA in style of successful proposals • NASA: Work with laboratories (Goddard, JPL etc.) • DoE: Work with laboratories (Argonne, Oak Ridge etc.) • DoD/Darpa: Must know the real intent of solicitation and program manager who often has strong technical impact on program • Industry: Very erratic in USA; stronger in Europe • Local (your own university): Obviously pursue but won’t clearly add to national reputation
Know your Reviewers • Reviewing involves “peer” review by mail from funding agency, panels and agency program managers • Different agencies have different balances here • Varies from NSF peer reviewers and panels • to DoD program managers • Other USA agencies are in between • Many reviews are incorrect as the reviewers do not understand your proposal and if it is too innovative, cannot understand it • Example: My “best” proposals in 1995 for web-based computing and web-based education were soundly rejected even though in retrospect “right-on” • Make your proposals exciting but not too far out • NSF OCI and CISE are computer/computational scientists; NSF EHR are Cognitive Science/School of Education • Neither unit will easily fund researchers from the other • All my EHR proposals turned down except for a small $50,000 grant • Implies education technology very weak in USA
Interdisciplinary work • Many people extol the value of interdisciplinary work and much of my research is of this type • However some dangers as hard to get respect from people of a different field • Make certain your work has components (e.g. papers) where respect clear • Only do interdisciplinary work where each involved field is world class
Glittering Diamonds • Often reviewers judge proposals on people involved and not the content (which they don’t in fact understand because it is “too far out” or an area outside their expertise) • Thus good to put “glittering diamonds” on proposals; researchers who are and are perceived to be world class • However reviewers note “fake collaborations”; only put those really involved on proposal and best to have pre-existing collaborations documented with joint papers etc.
The Institutional Advantage • Often artifacts – hardware, software, power – are very important and can be leveraged for success • Argonne/Chicago/USC leverage Globus software • Open Science Grid and TeraGrid “own” certain Grid areas • MIDAS owns epidemiology for NIH? • DoE labs own Homeland Security research ….. • SDSC NCSA leverage supercomputer infrastructure • I leverage “power” of “Alliance for Equity in Higher Education” which represents 335 Minority serving institutions • NSF funding and indeed project successes partnering with AIHEC, HACU and NAFEO • In outreach look for systemic not point solutions • e.g. in current NSF CPATH, do not write an IU proposal; write an (inter)national Informatics proposal • Create and nurture your artifacts and glittering diamonds
Further Principles and Issues • Don’t waste time on hopeless idealistic proposals • Safe strategy is to get started as a partner with one or more “Glittering Diamonds” • Best to be funded as a servant in heaven rather than be rejected as a ruler in hell ….. • The Glittering Diamond is a perfect tenure reference • Do not tabulate a lot of wishful thinking i.e. possible but not real activities • Publications and papers benefit from results with good graphics • Have clearly stated ideas and activities in your proposal; make it clear you know competing work • Focus – do not be too broad; quality better than quantity • Involve PhD not Masters students! • Not important to be PI; co-PI role in many ways best • If you put together a joint proposal, the PI must expect to do 95% of work; organize brainstorming sessions as they create links between collaborators and this shows in proposal quality • Organizing specialized workshops is a good way to become known • Letters of support of dubious value; all escalated so not useful for reviewer; letter writers are restricted from being NSF reviewers for your proposal • Now you should be exploiting your current knowledge but thinking of the new thrusts that you will exploit 5 years from now • I have made many mistakes here; early on I dismissed Grids as obviously wrong but it was me that was wrong as it evolved to tackle different problems where it is a good idea. • Budgets take a lot of effort but remember even if proposal approved, agency will change budget – so budget should “compile correctly” and clearly match proposal but details not important; do justify what you put in!