In November 2006, I received a call from a colleague suggesting I apply for a Program Director opening at the National Science Foundation. Prior to his call, I had determined to re-orient my research in machine learning towards environmental applications. It didn’t take long to decide that NSF would be much more a retooling for, rather than a distraction from, this new direction.
The following March I gave my first job talk in 20 years. The talk included a retrospective of my research, teaching and professional service, but I was instructed that my work be a minority focus; I was to reflect on the works of others, attempting to express themes in, and visions for, my research and education community. I learned early that synthesis was highly valued at NSF. My themes included evolutionary specialization and competition as metaphors for incremental and transformative research agenda; my visions included the application of artificial intelligence (AI) to mitigation of, and adaptation to, climate change, as well as the integration of ethics and contemporary issues into technical courses.
I received an offer. I took leave from Vanderbilt University and worked at NSF under the auspices of the Intergovernmental Personnel Act (“as an IPA”) —that is, NSF made an award to Vanderbilt for 100 percent of my time. This elegant mechanism allowed me to retain my salary and benefits through my university. In addition, I was given a living allowance by NSF, which just covered an unfurnished, one-bedroom apartment one block from NSF. I would joke that my vertical (elevator) commute was greater than my horizontal commute. I was on a one-year contract, renewable yearly as Vanderbilt and NSF allowed, for no more than 4 years.
‘Eyes on the Prize’
A NSF PD, a ‘rotator’ like me or a permanent Federal employee, is working for the advancement of science and for the United States—this ‘eyes on the prize’ is sustaining through very long hours. Part of the ‘prize’ can also start as an individual vision. My Division Director (DD) told me that, as a rotating PD, I could change the ways that things are done, though the changes might not be manifest until after I had left NSF! However, occasionally change occurs more quickly—one program grew from the compelling vision that a PD colleague expressed in her recruiting talk.
While my passions were for environmental applications of computing, I learned of, bought into and contributed to the visions of others in areas that included international collaboration, interdisciplinary research in biology and computing, creativity and computing, and social computational systems. Through negotiation and consensus building, both vertically and laterally, individual interests can grow into proposal solicitations and programs. Importantly, these interests are tempered, if not driven, by a strong ‘bottom-up’ research community influence — hypothetically, for example, PDs might notice an influx of research that intersects computer vision and human perception, which can lead to a new program a couple of years later. Such ‘serendipity’ is common at an agency like NSF, housed in a single building, with representatives of all scientific and engineering disciplines.
Day-to-Day Responsibilities of a Program Director
A core responsibility of a PD is the vetting of research proposals—clustering proposals into overlapping topic areas and panels, recruiting panelists and other reviewers, overseeing the review panels, and making funding recommendations. This latter task is typically more complex than simply passing on the recommendations of a panel, though these recommendations are the most important single factor a PD considers. A PD must make assessments across panels; in my case, these would typically be panels in areas such as planning, scheduling and search; machine learning; multi-agent systems; and knowledge representation. The PDs of a larger programmatic unit, in my case Robust Intelligence (RI), then come to joint recommendations across all the program’s areas—for RI these were robotics, natural language technologies, computer vision, computational neuroscience and other areas of AI that I covered. Attention to factors such as research area balance, both for the current year and recent years, play a role in addition to panel recommendations. What sympathy I have for current Federal budget woes stems from my experience at NSF trying to reach consensus with other PDs over funding recommendations under tight budget constraints. After reaching agreement, PDs write up the recommendations as ‘review analyses,’ which justify the expenditure (or not) of public funds for proposals.
PDs spend a good deal of time talking with researchers. There is no greater treat than emailing investigators with news that I (RI) was considering recommending their proposals for awards—the follow-up talks on budgets and award specifics were very pleasant. Investigators also contact PDs to describe research ideas, wondering where they fit into NSF programs. This usually happens by email and phone, but NSF welcomes investigators to come to its offices with an appointment. Initially, talks with investigators of declined proposals were difficult, but it’s safe to say that these discussions changed me most and I quickly grew to appreciate them. I would often reflect with investigators on the main messages of their proposals’ reviews to better inform possible revisions and resubmissions.
Early in my time as a PD I realized that I could not recommend funding for nearly all proposals that were worthy of funding. This was a source of genuine dismay; it’s an irony to me that I felt more empathy for investigators because I stood on the other side of the fence from them in these conversations, and of course because I had stood where they were, too (and would again). Tough decisions often resulted in small but significant points of pride as well—I recall one such instance when four PDs came together, pooling their very limited discretionary resources, to ‘save’ an excellent proposal that would not have otherwise been funded. One remembers these small victories. My experiences are also translating to civic action now that I have left NSF. In particular, I am writing my congressional representatives on the under-funding of science and other issues—something I hadn’t done before (and I don’t recall it having occurred to me to do before).
Within a few months of getting to NSF I was running panels. I felt some responsibility that panelists’ lives were significantly disrupted because of travel, particularly those coming from a great distance and/or with family and professional responsibilities—empathy again. I also found it hard to recruit panelists from among some groups, such as parents of young children, who may well constitute segments of the population weighting the broader impacts of computing more heavily than the rest! Generally, I was encouraged to innovate for the good of the organization, and I chose to pursue remote (‘virtual’) panel participation on a large scale as one such innovation. With few exceptions I gave panelists a choice of whether to come to NSF for a panel or to participate remotely. Some panels were exclusively composed of remote panelists, and I ran two such panels from my Vanderbilt office. That said, one of the few NSF activities that I would characterize as ‘fun’ was dinner out with panelists who physically came to NSF; informal time with panelists after hours is a real social and intellectual perk. There is much I have said elsewhere on panelist choice to travel or ‘phone in’ relative to ecological footprints, broadening participation, panelist wear and tear, and other factors. Suffice it to say here that I am satisfied that I helped sprout a seed through this experience—aware of my DD’s cautious encouragement that changes might not manifest while I was there.
PDs are responsible for post-award oversight as well, through annual and final reports, and site visits—typically the PD recruits the site visit team. In addition to vetting and oversight duties in a core area like RI, most PDs also participate in like roles for cross-cutting programs, which span two or more of NSF’s programmatic units. For example, I participated in various roles in Science, Engineering and Education for Sustainability (SEES), Social Computational Systems (SoCS), and Cyber-enabled Discovery and Innovation (CDI), and I served as the cognizant program officer for a large Science of Learning Center (SLC). I gained a greater appreciation for the place of AI (and computing broadly) in a larger scientific and societal landscape, in large part because of my participation in cross-cutting programs. I know that other IPAs similarly walk away with a broader appreciation of their chosen fields.
To stay up on the latest research developments and to gauge community interests, PDs are expected to attend conferences and other meetings in their respective fields—and they have a travel budget to do so. As a machine learning person who oversaw other areas of AI such as planning and knowledge representation, I would attend a broad range of talks at a conference to learn more about my areas of coverage. I would be invited to some meetings to give talks, typically on NSF programs. Ideally, after each professional meeting PDs prepare and circulate, by email or an internal wiki/blog, a conference report for the benefit of NSF colleagues.
In sum, being a NSF PD was richly rewarding, changing my approach to research, education and citizenship. There is no single stereotype of a NSF IPA, but I would guess that to be happy, or at least rewarded as an IPA, one would have a significant interest in broader impacts of science, skills at and motivation for synthesis, and a desire to learn from tightly coupled social interactions—in which faculty are often not practiced—with NSF colleagues, researchers, as well as others in government. Learning the “inside story” on funding is a benefit, but for purposes of improved grantsmanship, I think it’s very over-rated and probably not a motivation that would sustain a person for long through the challenges of being a NSF PD. Rather, the chief sustenance comes from a dedicated team of colleagues and service to country, science, planet and the people of those constituencies.
Dr. Douglas Fisher is Associate Professor of Computer Science and Computer Engineering at Vanderbilt University. From July 2007 to August 2010, he served as a Program Director in the Division of Information and Intelligent Systems (IIS) within the National Science Foundation’s (NSF) Directorate for Computer and Information Science and Engineering (CISE). He has agreed to describe something of his life as a NSF PD so that others can judge whether it might be something they would like to try.