ARPA-ED: A Qualified Thumbs-Up
Senator Michael Bennet (D-Colorado) has proposed an "Education-ARPA," modeled on the famed Defense Advanced Research Projects Agency (DARPA). The Obama administration has included a similar proposal, carving the dollars out of i3. Projected funding seems to hover in the $30 to $70 million range. (The proposals are cost neutral, meaning they'd be paid for by off-setting cuts.)
The idea intrigues me, but I've been as confused as most others about what ARPA-ED would look like or actually do. To try to get a clearer picture of what Sen. Bennet and the Obama administration have in mind, I invited the Senator and some key authorities over on Wednesday morning to explain it all at more length. The conversation included a public event, featuring Bennet; Jim Shelton, chief of ED's Office of Innovation and Improvement; John Easton, Director of the Institute for Education Sciences; and Ken Gabriel, the Deputy Director of DARPA (you can watch it here).
I came away more comfortable with the proposal. It didn't hurt that Gabriel did a terrific job of clearly and bluntly discussing DARPA and its role. Created in 1958 in the aftermath of the Soviet launch of Sputnik, DARPA operates as a nimble operation able to pursue strategically critical R&D.
Gabriel argued that the key word in the agency's title is not "advanced" or "research," but "project." This gives the agency a tightly defined mission--to pursue and create applied, transformative capabilities. This tack obviously requires a willingness to invest in programs with uncertain prospects and which may take a number of years to deliver, if they ever do. One current project is seeking to design a plane than can fly Mach 20 (twenty times the speed of sound). Previous projects have eventually yielded success like unmanned aircraft, stealth technology, and the internet. This patient, applied approach is very different from how R&D typically unfolds in education, where research is far less likely to focus on developing applications and where there is a demand for big answers that can be used right now. DARPA's focus on specific breakthrough capabilities strikes me as analogous to what I've previously described in Education Unbound as creating new "one percent" solutions--where you develop the ability to get profoundly better at solving one small but significant problem. One percent solutions don't "fix" schools; but if you develop one, and then another, and then another, stubborn problems start to become more tractable.
Any reason to think this kind of approach can usefully translate to the world of schooling, higher education, and workforce training, where it's about complex human interactions and not stealth technology? I think so. First, it's not as if the Department of Defense doesn't deal with teams, mentoring, training, and such. More to the point, ARPA-ED would make more sense for some kinds of capabilities than others. It would need to be directed accordingly. I find it much easier to envision a promising program to help students master conversational Mandarin or the principles of calculus in six weeks than one which "turns around" schools or figures out "optimal" pay systems. I could imagine projects that help an ELL student make up three years of English acquisition in three months, an iPad app that can help identify and remediate problems in early reading, or an assessment that can capture cognitive development in more robust, compelling ways.
Gabriel alluded to an intriguing DARPA project that enjoyed enormous success at workforce training. The Institute for Defense Analysis has reported, "In a recent study, DARPA compared students who trained to be Navy Information System Technicians, and found that those who had been trained by a new digital tutor outperformed traditionally-trained students by two standard deviations. In other words, the average student trained by the new digital tutor outperforms around 98 percent of students trained using traditional instruction."
Gabriel was very clear about what enables DARPA to be successful despite its modest size (DARPA spends just $2.8 billion of the Department of Defense's sprawling, $60 billion annual research and development budget). An independent agency of about 220 employees, DARPA's director reports directly to the Secretary of Defense. It's a remarkably flat operation, with just a half-dozen office directors, a deputy director, and a director providing all the management. The agency has freedom from civil service and hiring restrictions, and purposefully recruits outsiders from the worlds of academe, research, science, and technology to generate fresh thinking and ideas. Employees typically stay three to five years, and the work mostly consists at any given moment of different staff pursuing about 100 projects, each costing perhaps $15 or $20 million. The actual work is not done by DARPA, but by research universities, private sector firms, and other contracted parties. Gabriel was excruciatingly clear that DARPA wouldn't work if it was run like a typical federal agency. He remarked that "people are the lifeblood of the agency" and that "ideas are fragile," and therefore that the agency must be free to seek technical expertise; to recruit "with a sense of urgency;" to be small, autonomous, and agile; to have staff serve and then leave; and to be buffered from personal and political agendas on Capitol Hill or in DoD.
For any ARPA-ED proposal to make sense, it has to incorporate the same capabilities as DARPA. And that's an immense challenge. One savvy DC veteran enumerated on Wednesday the things that could go wrong: Congressional micro-management and an insistence on "immediately useful" results; an inability to attract or hire real talent due to hiring rules; an inability to move quickly due to cumbersome agency procedures and procurement rules; the desire of established research outfits, advocacy groups, and regional interests for earmarks and set-asides; and ED's aversion to dealing with for-profits.
There are a bunch of interesting details I won't wade into right now. One is that DARPA can count on DoD to be a giant customer for its best work. An ARPA-ED would have to instead depend on districts, states, and colleges to buy its product. Shelton spoke to this, arguing that the logic of ARPA-ED does depend on the faith that demand will emerge for capabilities which clearly represent a profound improvement. Another issue is that DARPA seeds innovation by being clear that intellectual property is owned not by the government but by those doing the work; that ensures it seeds the marketplace and doesn't stifle it. It's also unclear how ARPA-ED would proceed with funding which would amount to just one or two percent of what DARPA enjoys.
Here's where I come out. I like the idea. I believe there's a vital role for this kind of project-driven federal R&D. This can, as Shelton notes, help equip U.S. firms to thrive in the multi-trillion dollar international education marketplace. But this is only worth doing if it's done right, and that will be enormously difficult. There will be huge temptations to leave the design vague, get the thing launched, and then hope for the best. I think that would be a catastrophic mistake. And, we don't have new money to spend. So, I'm in...if the proposal is cost-neutral and, more importantly, if the legislation carefully and explicitly incorporates and safeguards the features that have fueled DARPA's success.