Moving the Needle on STEM Workforce Development through Fellowships and Mentorship Support in the CHIPS and Science Act

The CHIPS and Science Act ushered in unprecedented opportunities for American manufacturing, science, and innovation – and yet, current underfunding leaves the outcomes at risk.

The legislation directs the federal government to invest $280 billion to bolster U.S. semiconductor capacity, catalyze R&D, create regional high-tech hubs, and develop a larger, more inclusive STEM workforce. The federal investment of $50 billion in semiconductor manufacturing is estimated to add $24.6 billion annually to the American economy and create 185,000 jobs from 2021 to 2026. However, at the current rate of STEM degree completion, the U.S. may not be able to produce enough qualified workers to fill these jobs. Left unaddressed, this labor market gap will have cascading effects on the U.S. economy and compromise the nation’s global competitiveness.

Supporting STEM Workforce Development by Expanding Fellowship and Mentorship Programs

Despite the progress that has been made in recent years to grow the STEM pipeline, STEM graduates continue to lack the opportunity to contribute to the research enterprise and are not equipped to translate their scientific knowledge into actionable policy solutions. The CHIPS and Science Act attempts to address the shortfall in the U.S. STEM workforce and create more career pathways for graduates by authorizing federal agencies to expand their fellowship programs. 

For example, the legislation directs the National Science Foundation (NSF) to expand the number of new graduate research fellows supported annually over the next 5 years to no fewer than 3,000 fellows. This provision echoes the recommendations from a 2021 Federation of American Scientists (FAS) policy memo calling for the expansion of the Graduate Research Fellowship Program in order to catalyze and train a new workforce that would maintain America’s leading edge in the industries of the future. Another important provision has led to the launch of NSF’s Entrepreneurial Fellowships in September 2022, with the goal of supporting STEM entrepreneurs from diverse backgrounds in turning breakthroughs from the laboratory into products and services that benefit society. 

In addition to fellowships, the legislation also includes federal funding for graduate student and postdoctoral research mentorship and professional development, which are critical elements to developing our nation’s research enterprise. Supportive mentors and advisors can guide career planning for future scientists and help them develop the necessary critical thinking and problem solving skills. This is also the case for students from underrepresented minority backgrounds (URMs), where positive research and mentorship experiences contribute to persistence in intention to pursue a STEM career following graduation.

While these provisions are promising, more can be done to ensure better oversight and support of mentorship programs within federal funded research programs. The GRAD Coalition, which was established to support the Congressional Graduate Research and Development Caucus, has called on Congress to expand mentorship oversight and support, specifically to: 

The National Institutes of Health (NIH) has long recognized the need for mentorship at the post-doctoral level. In 2023,the NIH Advisory Committee to the Director (ACD) Working Group on Re-envisioning NIH-Supported Postdoctoral Training held listening sessions resulting in a report detailing many aspects of the postdoctoral experience in biomedical fields: lack of adequate compensation, concerns about postdoctoral quality of life and challenges with diversity, equity, inclusion, and accessibility. Many of these postdoctoral issues have been known for some time but continue to be insufficiently addressed. The report calls for increasing oversight and accountability of faculty for mentoring, specifically for NIH to: 

While boosts for science and education provisions in the legislation have been authorized, funding for the “and science” portion of the act has fallen short in several areas. FAS analysis shows that the FY 2024 appropriations for NSF are approximately $6 billion-short or 39% below the CHIPS and Science authorization levels, which has the potential to set the U.S. back in several areas of science and technology.

Maintaining the U.S. scientific and research enterprise requires a whole-of-government approach. Expanding fellowship programs and better incorporating mentorship in federal-funded programs can have far-reaching consequences for the STEM pipeline and maintaining our nation’s edge in scientific research and innovation. The CHIPS and Science Act provides specific opportunities for federal agencies, Congress, and the executive branch to grow the U.S. STEM workforce pipeline by expanding fellowships and mentorship support for graduate students and postdoctoral researchers. Our nation’s global leadership in science and technology is dependent upon the research and innovation driven by graduate students and postdoctoral researchers, and fully funding the authorized programs and new initiatives in the CHIPS and Science Act will help ensure that this trend continues. 

Expanding Pathways for Career Research Scientists in Academia

Summary

The U.S. university research enterprise is plagued by an odd bug: it encourages experts in science, technology, engineering, and math (STEM) to leave it at the very moment they become recognized as experts. People who pursue advanced degrees in STEM are often compelled by deep interest in research. But upon graduation from master’s, Ph.D., or postdoctoral programs, these research-oriented individuals face a difficult choice: largely cede hands-on involvement in research to pursue faculty positions (which increasingly demand that a majority of time be spent on managerial responsibilities, such as applying for grants), give up the higher pay and prestige of the tenure track in order to continue “doing the science” via lower-status staff positions (e.g., lab manager, research software engineer), or leave the academic sector altogether. 

Many choose the latter. And when that happens at scale, it harms the broader U.S. scientific enterprise by (i) decreasing federal returns on investment in training STEM researchers, and (ii) slowing scientific progress by creating a dearth of experienced personnel conducting basic research in university labs and mentoring the next generation of researchers. The solution is to strengthen and elevate the role of the career research scientist1 in academia—the highly trained senior research-group member who is hands-on and in the lab every day—in the university ecosystem. This is, fundamentally, a fairly straightforward workforce-pipeline issue that federal STEM-funding agencies have the power to address. The National Institutes of Health (NIH) and the National Science Foundation (NSF) — two of the largest sources of academic research funding — could begin by hosting high-level discussions around the problem: specifically, through an NSF-led workshop and an NIH-led task force. In parallel, the two agencies can launch immediately tractable efforts to begin making headway in addressing the problem. NSF, for instance, could increase visibility and funding for research software engineers, while NSF and/or NIH could consider providing grants to support “co-founded” research labs jointly led by an established professor or principal investigator (PI) working alongside an experienced career research scientist.

The collective goal of these activities is to infuse technical expertise into the day-to-day ideation and execution of science (especially basic research), thereby accelerating scientific progress and helping the United States retain world scientific leadership.

Challenge and Opportunity

The scientific status quo in the United States is increasingly diverting STEM experts away from direct research opportunities at universities. STEM graduate students interested in hands-on research have few attractive career opportunities in academia: those working as staff scientists, lab managers, research software engineers, and similar forego the higher pay and status of the tenure track, while those working as faculty members find themselves encumbered by tasks that are largely unrelated to research. 

Making it difficult for STEM experts to pursue hands-on research in university settings harms the broader U.S. scientific enterprise in two ways. First, the federal government disburses huge amounts of money every year—via fellowship funding, research grants, tuition support, and other avenues—to help train early-career STEM researchers. This expenditure is warranted because, as the Association of American Universities explains, “There is broad consensus that university research is a long-term national investment in the future.” This investment hinges on university contributions to basic research; universities and colleges account for just 13% of overall U.S. research and development (R&D) activity, but nearly half (48%) of basic research. Limited career opportunities for talented STEM researchers to continue “doing the science” in academic settings therefore limits our national returns on investment in these researchers.

Box 1. Productivity benefits of senior researchers in software-driven fields.
Cutting-edge research in nearly all STEM fields increasingly depends on software. Indeed, NSF observes that software is “directly responsible for increased scientific productivity and significant enhancement of researchers’ capabilities.” Problematically, there is minimal support within academia for development and ongoing maintenance of software. It is all too common for a promising research project at a university lab to wither when the graduate student who wrote the code upon which the project depends finishes their degree and leaves.

The field of deep learning (a branch of artificial intelligence (AI) and machine learning) underscores the value of research software. Progress in deep learning was slow and stuttering until development of user-friendly software tools in the mid-2010s: a development spurred mostly by private-sector investment. The result has been an explosion of productivity in deep learning. Even now, top AI research teams cite software-engineering talent as a critical input upon which their scientific output depends. But while research software engineers are some of the most in-demand and valuable team members in the private sector, career positions for research software engineers are uncommon at academic institutions. How much potential scientific discovery are U.S. university labs failing to recognize as a result of this underinvestment?

Second, attrition of STEM talent from academia slows the pace of U.S. scientific progress because most hands-on research activities are conducted by graduate students rather than more experienced personnel. Yet, senior researchers are far more scientifically productive. With years of experience under their belt, senior researchers possess tacit knowledge of how to effectively get research done in a field, can help a team avoid repeating mistakes, and can provide the technical mentorship needed for graduate students to acquire research skills quickly and well. And with graduate students and postdocs typically remaining with a research group for only a few years, career research scientists also provide important continuity across projects. The productivity boosts that senior researchers can deliver are especially well established for software-driven fields (see box).

The absence of attractive job opportunities for career research scientists at most academic institutions is an anomaly. Such opportunities are prevalent in the private sector, at national labs (e.g., those run by the NIH and the Department of Energy) and other government institutions, and in select well-endowed university labs that enjoy more discretionary spending ability. As the dominant funder of university research in the United States, the federal government has massive leverage over the structure of research labs. With some small changes in grant-funding incentives, federal agencies can address this anomaly and bring more senior research scientists into the academic research system. 

Plan of Action

Federal STEM-funding agencies — led by NSF and NIH, as the two largest sources of federal funding for academic research — should explore and pursue strategies for changing grant-funding incentives in ways that strengthen and elevate the role of the career research scientist in academia. We split our recommendations into two parts. 

The first part focuses on encouraging discussion. The problem of limited career options for trained STEM professionals who want to engage in hands-on research in the academic sector currently flies beneath the radar of many extremely knowledgeable stakeholders inside and outside of the federal government. Bringing these stakeholders together might result in excellent actionable suggestions on how to retain talented research scientists in academia. Second, we suggest two specific projects to make headway on the problem: (i) further support for research software engineers and (ii) a pilot program supporting co-founded research labs. While the recommendations below are targeted to NSF and NIH, other federal STEM-funding agencies (e.g., the Departments of Energy and Defense) can and should consider similar actions. 

Part 1. Identify needs, opportunities, and options for federal actions to support and incentivize career research scientists.2

Shifting academic employment towards a model more welcoming to career research scientists will require a mix of specific new programs and small and large changes to existing funding structures. However, it is not yet clear which reforms should be prioritized. Our first set of suggestions is designed to start the necessary discussion.

Specifically, NSF should start by convening key community members at a workshop (modeled on previous NSF-sponsored workshops, such as the workshop on a National Network of Research Institutes [NNRI]) focused on how the agency can encourage creation of additional career research scientist positions at universities. The workshop should also (i) discuss strategies for publicizing and encouraging outstanding STEM talent to pursue such positions, (ii) identify barriers that discourage universities from posting for career research scientists, and (iii) brainstorm solutions to these barriers. Workshop participants should include representatives from federal agencies that sponsor national labs as well as industry sectors (software, biotech, etc.) that conduct extensive R&D, as these entities are more experienced employers of career research scientists. The workshop should address the following questions:

The primary audience for the workshop will be NSF leadership and policymakers. The output of the workshop should be a report suggesting a clear, actionable path forward for those stakeholders to pursue.

NIH should pursue an analogous fact-finding effort, possibly structured as a working group of the Advisory Committee to the Directorate. This working group would identify strategies for incentivizing labs to hire professional staff members, including expert lab technicians, professional biostatisticians, and RSEs. This working group will ultimately recommend to the NIH Director actions that the agency can take to expand the roles of career research scientists in the academic sector. The working group would address questions similar to those explored in the NSF workshop.

Part 2. Launch two pilot projects to begin expanding opportunities for career research scientists.

Pilot 1. Create a new NSF initiative to solicit and fund requests for research software engineer (RSE) support. 

Research software engineers (RSEs) build and maintain research software, and train scientists to use that software. Incentivizing the creation of long-term RSE positions at universities will increase scientific productivity and build the infrastructure for sustained scientific progress in the academic sector. Though a wide range of STEM disciplines could benefit from RSE involvement, NSF’s Computer and Information Science and Engineering (CISE) Directorate is a good place to start expanding support for RSEs in academic projects. 

CISE has previously invested in nascent support structures for professional staff in software and computing fields. The CISE Research Initiation Initiative (CRII) was created to build research independence among early-career researchers working in CISE-related fields by funding graduate-student appointments. Much CRII-funded work involves producing — and in turn, depends on — shared community software. Similarly, the Campus Research Computing Consortium (CaRCC) and RCD Nexus are NSF-supported programs focused on creating guidelines and resources for campus research computing operations and infrastructure. Through these two programs, NSF is helping universities build a foundation of (i) software production and (ii) computing hardware and infrastructure needed to support that software. 

However, effective RSEs are crucial for progress in scientific fields outside of CISE’s domain. For example, one of this memo’s authors has personal experience with NSF-funded geosciences research. PIs working in this field are desperate for funding to hire RSEs, but do not have access to funding for that purpose. Instead, they depend almost entirely on graduate students.

As a component of the workshop recommended above, NSF should highlight other research areas hamstrung by an acute need for RSEs. In addition, CISE should create a follow-on CISE Software Infrastructure Initiative (CSII) that solicits and funds requests from pre-tenure academic researchers in a variety of fields for RSE support. Requests should explain how the requested RSE would (i) catalyze cutting-edge research, and (ii) maintain critical community open-source scientific software. Moreover, academia severely lacks strong mentorship in software engineering. A specific goal of CSII funding should be to support at least a 1:3 ratio of RSEs to graduate students in funded labs. Creative evaluation mechanisms will be needed to assess the success of CSII. The goal of this initiative will be a community of university researchers productively using software created and supported by RSEs hired through CSII funding. 

Pilot 2. Provide grants to support “co-founded” research labs jointly led by an established professor or principal investigator (PI) working alongside an experienced career research scientist.

Academic PIs (typically faculty) normally lead their labs and research groups alone. This state of affairs leads to high rates of burnout, possibly leading to poor research success. In some cases, starting an ambitious new project or company with a co-founder makes the endeavor more likely to succeed while being less stressful and isolating. A co-founder can provide a complementary set of skills. For example, the startup incubator Y Combinator is well known for wanting teams to include a CEO visionary and manager working alongside a CTO builder and designer. By contrast, academic PIs are expected to be talented at all aspects of running a modern scientific lab. Developing mechanisms to help scientists come together and benefit from complementary skill sets should be a high priority for science-funding agencies.

We recommend that NSF and/or NIH create a pilot grant program to fund co-founded research labs at universities. Formally co-founded research groups have been successful across scientific domains (e.g., the AbuGoot Lab at MIT and the Carpenter-Singh Lab at the Broad Institute), but remain quite rare. Federal grants for co-founded research labs would build on this proof of concept by competitively awarding 5–7 years of salary and equipment funding to support a lab jointly run by an early-career PI and a career research scientist. A key anticipated benefit of this grant program is increased retention of outstanding researchers in positions that enable them to keep “doing the science.” Currently, the most talented STEM researchers become faculty members or leave academia altogether. Career research scientist positions simply cannot offer competitive levels of compensation and prestige. Creating a new, high-profile, grant-funded opportunity for STEM talent to remain in hands-on university lab positions could help shift the status quo. Creating a pathway for co-founded and co-led research labs would also help PIs avoid isolation and burnout while building more robust, healthy, and successful research teams.

Conclusion

Many breakthroughs in scientific progress have required massive funding and national coordination. This is not one of them. All that needs to be done is allow expert research scientists to do the hands-on work that they’ve been trained to do. The scientific status quo prevents our nation’s basic research enterprise from achieving its full potential, and from harnessing that potential for the common good. Strengthening and elevating the role of career research scientists in the academic sector will empower existing STEM talent to drive scientific progress forward.

Frequently Asked Questions
Are there places where research scientists are common?

Yes. The tech sector is a good example. Multiple tech companies have developed senior individual contributor (IC) career paths. These IC career paths allow people to grow their influence while remaining mostly in a hands-on technical role. The most common role of a senior software engineering IC is that of the “tech lead”, guiding the technical decision making and execution of a team. Other paths might involve prototyping and architecting a critical new system or diving in and solving an emergency problem. For more details on this kind of career, look at the Staff Engineer book and accompanying discussion.

Why is now the time for federal STEM-funding agencies to increase support for career research scientists?

The United States has long been the international leader in scientific progress, but that position is being threatened as more countries develop the human capital and infrastructure to compete in a knowledge-oriented economy. In an era where humankind faces mounting existential risks requiring scientific innovation, maintaining U.S. scientific leadership is more important than ever. This requires retaining high-level scientific talent in hands-on, basic research activities. But that goal is undermined by the current structure of employment in American academic science.

Which other federal agencies fund scientific research, and could consider actions similar to those proposed in this memo for NSF and NIH?

Key federal STEM-funding agencies that could also consider ways to support and elevate career research scientist positions include the Departments of Agriculture, Defense, and Energy, as well as the National Aeronautics and Space Administration (NASA).

Addressing the Mental Health Crisis Among Predoctoral and Postdoctoral Researchers in STEM

Summary

The growing mentalhealth crisis among science, technology, engineering, and math (STEM) doctoral and postdoctoral researchers threatens the future and competitiveness of science and technology in the United States. The federal government should tackle this crisis through a four-part approach to (i) improve data collection on the underlying drivers of mental-health struggles in STEM, (ii) discourage behaviors and cultures that perpetuate stress, (iii) require Principal Investigators (PIs) to submit a statement of their mentoring philosophy as part of applications for federally supported research grants, and (iv) increase access to mental-health care for predoctoral and postdoctoral researchers.

Challenge and Opportunity

The prevalence of mental-health problems is higher among Ph.D. students than in the highly educated general population: fully half of Ph.D. students experience psychological distress. In a survey of postdoctoral researchers conducted by Nature, 51% of respondents reported considering leaving science due to work-related mental-health concerns. 65% of respondents reported experiencing power imbalances or bullying during their postdoctoral appointments, and 74% reported observing the same. Stress accumulation not only leads to the development of neuropsychiatric disorders among the developing STEM workforce — it also contributes to burnout. At a time when advancing U.S. competitiveness in science and technology is of utmost importance, the mental-health crisis is depleting our nation’s STEM pipeline when we should be expanding and diversifying it. This is a crisis that the federal government is well-positioned to and must solve. 

Plan of Action

The federal government should counter the mental-health crisis for U.S. doctoral and postdoctoral researchers through a four-part approach to (i) improve data collection on the underlying drivers of mental-health struggles in STEM, (ii) discourage behaviors that perpetuate stress, (iii) require PIs to submit a statement of their mentoring philosophy as part of applications for federally supported research grants, and (iv) increase access to mental-health care for doctoral and postdoctoral researchers. Detailed recommendations associated with each of these steps are provided below.

Part 1. Improve data collection

Data drives public policy. Various organizations conduct surveys evaluating the mental health of doctoral and postdoctoral researchers in STEM, but survey designs, target audiences, and subsequent follow-up and monitoring are inconsistent. This fragmented information ecosystem makes it difficult to integrate and act on existing data on mental health in STEM. To provide a more comprehensive picture of the STEM mental-health landscape in the United States, the National Institutes of Health (NIH) and the National Science Foundation (NSF) should work together to conduct and publish biennial evaluations of the state of mental health of the STEM workforce. The survey format could be modeled on the NSF’s Survey of Doctorate Recipients or the Survey of Earned Doctorates — and, like those surveys, resultant data could be maintained at NSF under the National Center for Science and Engineering Statistics. Once established, the data from the survey can be used to track effectiveness of programs that are implemented and direct the federal government to change or start new initiatives to modify the needs of doctoral and postdoctoral researchers. Additionally, the NSF and NIH could partner with physicians within HHS to define and establish what “healthy” means in terms of mental-health guidelines in order to establish new program guidelines and goals. 

Part 2. Discourage problematic behaviors

The future of a doctoral or postdoctoral researcher depends considerably on the researcher’s professional relationship with their PI(s). Problems in the relationship — including bullying, harassment, and discrimination — can put a trainee in a difficult situation, as the trainee may worry that confronting the PI could compromise their career opportunities. The federal government can take three steps to discourage these problematic behaviors by requiring PIs to submit and implement training and mentorship plans for all grant-supported trainees. 

First, the White House Office of Science and Technology Policy (OSTP) should assemble a committee of professionals in psychology, social sciences, and human resources to define what behaviors constitute bullying and harassment in academic work environments. The committee’s findings should be publicized via a web portal (similar to NSF’s website on Sexual Harassment), and included in all requests for grant applications issued by federal STEM-funding agencies (in order to raise awareness among PIs).

Second, federal STEM-funding agencies should require universities to submit annual reports of bullying to federal, grant-issuing agencies. NSF already requires institutions to report findings of sexual harassment and other forms of harassment and can revoke grants if a grantee is found culpable. NSF and other STEM-funding agencies should add clarity to this definition and broaden this reporting to include bullying and retaliation to include bullying and retaliation attempts by PIs, with similar consequences for repeated offenses. Reinstatement of privileges (e.g., reinstatement of eligibility for federal grant funding) would be considered on a case-by-case basis by the grant-issuing institution and could be made contingent on implementation of an adequate “re-entry” plan by the PI’s home institution. The NIH Office of Behavioral and Social Science Research should be consulted to help formulate such “re-entry” plans to benefit both researchers and PIs.

Third, STEM-funding agencies could work together to establish a mechanism whereby trainees can anonymously report problematic PI behaviors. NSF has a complaint form for those who wish to report incidents for incidents of sexual harassment or harassment. Thus, NSF could expand their system to accept broader incidents such as bullying and retaliation attempts and NIH could use this complaint form as a template for reporting as well. In conjunction with reporting misconduct, a “two-strike” accountability system should be imposed if a PI is found guilty of harassment, bullying, or other behaviors that could contribute to the development of a neuropsychiatric disorder. After receiving a first strike (report of problematic behavior and a guilty verdict), the PI would be given a warning and be required to participate in relevant training workshops and counseling using a plan outlined by social science professionals at NIH. If a second strike is received, the PI would lose privileges to apply for federal grant funding and opportunities to serve on committees that are often favored for tenure and promotion, such as grant review committees. Again, reinstatement of privileges would be considered on a case-by-case basis by the grant-issuing institution and could be made contingent on implementation of an adequate “re-entry” plan.

Part 3. Require submission of mentoring philosophies

NIH F31 predoctoral and F32 postdoctoral award applications already require PIs to submit mentoring plans for their trainees to receive professional-development training. Federal STEM-funding agencies should build on this precedent by requiring PIs applying for federal grants to submit not just mentoring plans, but brief summaries of their mentoring philosophies. As the University of Colorado Boulder explains, a mentoring philosophy

“…defines [a mentor’s] approach to engaging with students as [they] guide their personal growth and professional development, often explaining [the mentor’s] motivation to mentor with personal narratives while highlighting their goals for successful relationships and broader social impact. These statements may also be considered ‘living documents’ that are updated as [the mentor] refine[s[ [their] approach and the context and goals of [their] work changes.”

Mentoring philosophies help guide development of and updates to individualized mentoring plans. Mentoring philosophies also promote equity and inclusion among mentees by providing a common starting point for communication and expectations. Requiring PIs to create mentoring philosophies will elevate mental health among doctoral and postdoctoral researchers in STEM by promoting effective top-down mentorship and discouraging unintended marginalization. And since a growing number of university faculty are already creating mentoring philosophies, this new requirement shouldn’t be seen as just another administrative burden; rather, it would serve as a means to quickly perpetuate a best practice that is already spreading. The federal government can support PIs in adhering to this new requirement by working with external partners to collect and broadly share resources related to preparing mentoring philosophies. The Center for the Improvement of Mentored Experiences in Research, for instance, has already assembled a suite of such resources on its web platform. 

Part 4. Increase access to mental health care

Concurrent with reducing causes of mental health burdens, the federal government should work to expand doctoral and postdoctoral researchers’ access to adequate mental-health care. Current access may vary considerably depending on the level of insurance coverage offered by a researcher’s home institution. Inspired by legislation (S. 3048 – Stopping the Mental Health Pandemic Act, where funds can be used to support and enhance mental health services) introduced in the 117th Congress, the Department of Health and Human Services (HHS) should partner with federal STEM-funding agencies to design and implement new pathways, programs, and opportunities to strengthen mental-health care among early-career STEM professionals. In particular, the federal government could create a library of model policies that federally funded public and private institutions could adopt to strengthen mental-health care for employed early-career researchers. Examples include allowing trainees to take time off during the workday to receive mental-health treatment without expectations to make up hours outside of business hours, providing a supplemental stipend for trainees to pay for therapy costs that are not covered by insurance, and addressing other sources of stress that can exacerbate stressful situations, such as increasing stipends to decrease financial stress. 

Conclusion

The U.S. science and technology enterprise is only as strong as the workforce behind it. Failing to address the mental-health crisis that plagues early-career researchers will lead the United States to fall behind in global research and development due to talent attrition. President Biden’s 2022 State of the Union address cited mental health as a priority area of concern. There is an especially clear need for a culture change around mental health in academia. The four actions detailed in this memo align with the President’s policy agenda. By improving data collection on the mental-health status of STEM doctoral and postdoctoral researchers, discouraging behaviors and cultures that produce stress among this population, improving training and mentorship at universities, and expanding access to mental-health care among STEM doctoral and postdoctoral researchers, the federal government can ensure that success for early-career STEM professionals does not demand mental-health sacrifice.

Frequently Asked Questions
Why does this proposal focus on early-career professionals in STEM and not on other fields?

STEM fields are closely tied to the U.S. economy, supporting two-thirds of U.S. jobs and 69% of the U.S. Gross Domestic Product (GDP). Attrition of U.S. researchers from STEM fields due to mental-health challenges has disproportionately adverse effects on American society and undermines U.S. competitiveness. Policymakers should prioritize actions designed to combat the mental-health crisis in STEM.

Bullying and harassment are subjective behaviors. How can the federal government prevent false allegations from being submitted by doctoral and postdoctoral researchers?

NSF already requires that universities who receive federal research funding conduct internal investigations to validate claims of harassment and sexual harassment. Similar policies could be implemented regarding reported bullying and/or workplace harassment. If an allegation is found to be false, it should be handled by university-specific policies.

If bullying and harassment are causing serious issues in STEM training, why should a PI be allowed “re-entry” to apply for federal funding to mentor students and postdocs after workshops and therapy are completed?

The goal of requiring PIs to attend workshops on mentorship and therapy sessions is to help them better themselves and improve their ability to mentor the next generation of STEM professionals. Re-entry to mentoring trainees will be closely monitored by leadership faculty who should conduct surveys of both mentors and mentees to determine if the PI understands (a) their previous misconduct and (b) the lasting mental health effects that their previous actions inflicted on their trainees.

NIH and NSF aren’t the only federal agencies that provide funding for training early career researchers. What about the others?

NIH and NSF are arguably the two leading federal agencies when it comes to providing federal funding for graduate students. That said, recommendations presented in this memo could easily be extended to other STEM-funding agencies. For instance, there is a timely opportunity to extend these recommendations to the Department of Energy (DOE). DOE is currently working to manage the President’s major FY23 investment in clean energy and sustainability, including through significant research-grant funding. Coupling these new grants with policies designed to mitigate mental-health burdens among early-career researchers could help foster a more resilient and productive clean-energy workforce and serve as a pilot group for the NIH and NSF to follow.

Requiring the reporting of bullying or harassment by a PI is an administrative burden. Why should universities take on increased responsibilities in this area?

The administrative responsibilities for reporting are minimal. NSF’s Organizational Notification of Harassment Form can — at a minimum — be used as a template for NSF, NIH, and other agencies to notify the federal government of guilty verdicts from universities. Alternatively, doctoral and postdoctoral researchers can submit incidents for reporting by federal agencies similar to NSF’s existing complaint form, which would reduce the initial administrative burden of university employees but may create additional hours of work once federal agencies conduct their investigations.

Some universities are offering free yoga and meditation classes for predoctoral and postdoctoral researchers. Others are offering training courses on developing resilience to stress. Aren’t these opportunities sufficient for alleviating mental health concerns?

While the strategies above teach researchers how to cope with stress, a long-term, more supportive approach would be to reduce stress by going straight to the source. Actions such as addressing harassment and bullying will benefit not only the researcher themselves, but others in the work environment by fostering a responsible, low-stress culture.

7. How are mentoring philosophies different from mentoring plans?

The submission of mentoring plans by PIs are currently required for NIH pre- and post-doctoral fellowship applications. They are meant to supplement the training of a researcher by focusing on the logistics of skill building. However, mentorship of a researcher transcends knowledge and skill-building — it also encompasses the holistic development of a researcher, supporting and respecting their interests, values, and considerations of their individual situations. Thus, submission of a mentoring philosophy is meant to stimulate thoughts and conversations about how a PI wants to communicate openly and honestly with their trainee and how they can adapt to support the mentoring style that best fits their trainee.

Creating a Public System of National Laboratory Schools

Summary

The computational revolution enables and requires an ambitious reimagining of public high-school and community-college designs, curricula, and educator-training programs. In light of a much-changed — and much-changing — society, we as a nation must revisit basic assumptions about what constitutes a “good” education. That means re-considering whether traditional school schedules still make sense, updating outdated curricula to emphasize in-demand skills (like computer programming), bringing current perspectives to old subjects (like computational biology); and piloting new pedagogies (like project-based approaches) better aligned to modern workplaces. To do this, the Federal Government should establish a system of National Laboratory Schools in parallel to its existing system of Federally Funded Research & Development Centers (FFRDCs).

The National Science Foundation (NSF) should lead this work, partnering with the Department of Education (ED) to create a Division for School Invention (DSI) within its Technology, Innovation, and Partnerships (TIP) Directorate. The DSI would act as a platform analogous to the Small Business Innovation Research (SBIR) program, catalyzing Laboratory Schools by providing funding and technical guidance to federal, state, and local entities pursuing educational or cluster-based workforce-development initiatives.

The new Laboratory Schools would take inspiration from successful, vertically-integrated research and design institutes like Xerox PARC and the Mayo Clinic in how they organized research, as well as from educational systems like Governor’s Schools and Early College High Schools in how they organized their governance. Each Laboratory School would work with a small, demographically and academically representative cohort financially sustainable on local per-capita education budgets.
Collectively, National Laboratory Schools would offer much-needed “public sandboxes” to develop and demonstrate novel school designs, curricula, and educator-training programs rethinking both what and how people learn in a computational future.

Challenge and Opportunity

Education is fundamental to individual liberty and national competitiveness. But the United States’ investment in advancing the state of the art is falling behind. 

Innovation in educational practice has been incremental. Neither the standards-based nor charter-school movements departed significantly from traditional models. Accountability and outcomes-based incentives like No Child Left Behind suffer from the same issue.

The situation in research is not much better: NSF and ED’s combined spending on education research is barely twice the research and development budget of Nintendo. And most of that research focuses on refining traditional school models (e.g. presuming 50-minute classes and traditional course sequences).

Despite all these efforts, we are still seeing unprecedented declines in students’ math and reading scores.

Meanwhile, the computational revolution is widening the gap between what school teaches and the skills needed in a world where work is increasingly creative, collaborative, and computational. Computation’s role in culture, commerce, and national security is rapidly expanding; computational approaches are transforming disciplines from math and physics to history and art. School can’t keep up.

For years, research has told us individualized, competency- and project-based approaches can reverse academic declines while aligning with the demands of industry and academia for critical thinking, collaboration, and creative problem-solving skills. But schools lack the capacity to follow suit.

Clearly, we need a different approach to research and development in education: We need prototypes, not publications. While studies evaluating and improving existing schools and approaches have their place, there is a real need now for “living laboratories” that develop and demonstrate wholly transformative educational approaches.

Schools cannot do this on their own. Constitutionally and financially, education is federated to states and districts. No single public actor has the incentives, expertise, and resources to tackle ambitious research and design — much less to translate into research to practice on a meaningful scale. Private actors like curriculum developers or educational technologists sell to public actors, meaning private sector innovation is constrained by public school models. Graduate schools of education won’t take the brand risk of running their own schools, and researchers won’t pursue unfunded or unpublishable questions. We commend the Biden-Harris administration’s Multi-Agency Research and Development Priorities for centering inclusive innovation and science, technology, education, and math (STEM) education in the nation’s policy agenda. But reinventing school requires a new kind of research institution, one which actually operates a school, developing educators and new approaches firsthand.Luckily, the United States largely invented the modern research institution. It is time we do so again. Much as our nation’s leadership in science and technology was propelled by the establishment ofland-grant universities in the late 19th century, we can trigger a new era of U.S. leadership in education by establishing a system of National Laboratory Schools. The Laboratory Schools will serve as vertically integrated “sandboxes” built atop fully functioning high schools and community colleges, reinventing how students learn and how we develop in a computational future.

Plan of Action

To catalyze a system of National Laboratory Schools, the NSF should establish a Division for School Invention (DSI) within its Technology, Innovation, and Partnerships (TIP) directorate. With an annually escalating investment over five years (starting at $25 million in FY22 and increasing to $400 million by FY26), the DSI could support development of 100 Laboratory Schools nationwide.

The DSI would support federal, state, and local entities — and their partners — in pursuing education or cluster-based workforce-development initiatives that (i) center computational capacities, (ii) emphasize economic inclusion or racial diversity, and (iii) could benefit from a high-school or community-college component.

DSI support would entail:

  1. Competitive matching grants modeled on SBIR grants. These grants would go towards launching Laboratory Schools and sustaining those that demonstrate success.
  2. Technical guidance to help Laboratory Schools (i) innovate while maintaining regulatory compliance, and (ii) develop financial models workable on local education budgets.
  3. Accreditation support, working with partner executives (e.g., Chairs of Boards of Higher Education) where appropriate, to help Laboratory Schools establish relationships with accreditors, explain their educational models, and document teacher and student work for evaluation purposes.
  4. Responsible-research support, including providing Laboratory Schools assistance with obtainingFederalwide Assurance (FWA) and access to partners’ Institutional Review Boards (IRBs).
  5. Convening and storytelling, raising awareness of and interest in Laboratory Schools’ mission and operations.

Launching at least ten National Laboratory Schools by FY23 would involve three primary steps. First, the White House Office of Science and Technology Policy (OSTP) should convene an expert group comprised of (i) funders with a track record of attempting radical change in education and (ii) computational domain experts to design an evaluation process for the DSI’s competitive grants, secure industry and academic partners to help generate interest in the National Laboratory School System, and recruit the DSI’s first Director.

In parallel, Congress should issue one appropriations report asking NSF to establish a $25 million per year pilot Laboratory School program aligned with the NSF Directorate for Technology, Innovation, and Partnerships (TIP)’s Regional Innovation Accelerators (RIA)’s Areas of Investment. Congress should issue a second appropriations report asking the Office of Elementary and Secondary Education (OESE) to release a Dear Colleague letter encouraging states that have spent less than 75% of their Elementary and Secondary School Emergency Relief (ESSER) or American Recovery Plan funding to propose a Laboratory School.

Finally, the White House should work closely with the DSI’s first Director to convene the Department of Defense Education Activity (DDoEA) and National Governors Association (NGA) to recruit partners for the National Laboratory Schools program. These partners would later be responsible for operational details like:

Focus will be key for this initiative. The DSI should exclusively support efforts that center:

  1. New public schools, not programs within (or reinventions of) existing schools.
  2. Radically different designs, not incremental evolutions.
  3. Computationally rich models that integrate computation and other modern skills into all subjects.
  4. Inclusive innovation focused on transforming outcomes for the poor and historically marginalized.

Conclusion

Imagine the pencil has just been invented, and we treated it the way we’ve treated computers in education. “Pencil class” and “pencil labs” would prepare people for a written future. We would debate the cost and benefit of one pencil per child. We would study how oral test performance changed when introducing one pencil per classroom, or after an after-school creative-writing program.

This all sounds stupid because the pencil and writing are integrated throughout our educational systems rather than being considered individually. The pencil transforms both what and how we learn, but only when embraced as a foundational piece of the educational experience.

Yet this siloed approach is precisely the approach our educational system takes to computers and the computational revolution. In some ways, this is no great surprise. The federated U.S. school system isn’t designed to support invention, and research incentives favor studying and suggesting incremental improvements to existing school systems rather than reimagining education from the ground up. If we as a nation want to lead on education in the same way that we lead on science and technology, we must create laboratories to support school experimentation in the same way that we establish laboratories to support experimentation across STEM fields. Certainly, the federal government shouldn’t run our schools. But just as the National Institutes of Health (NIH) support cutting-edge research that informs evolving healthcare practices, so too should the federal government support cutting-edge research that informs evolving educational practices. By establishing a National Laboratory School system, the federal government will take the risk and make the investments our communities can’t on their own to realize a vision of an equitable, computationally rich future for our schools and students.

Frequently Asked Questions

Who

1. Why is the federal government the right entity to lead on a National Laboratory School system?

Transformative education research is slow (human development takes a long time, as does assessing how a given intervention changes outcomes), laborious (securing permissions to test an intervention in a real-world setting is often difficult), and resource-intensive (many ambitious ideas require running a redesigned school to explore properly). When other fields confront such obstacles, the public and philanthropic sectors step in to subsidize research (e.g., by funding large research facilities). But tangible education-research infrastructure does not exist in the United States.

Without R&D demonstrating new models (and solving the myriad problems of actual implementation), other public- and private-sector actors will continue to invest solely in supporting existing school models. No private sector actor will create a product for schools that don’t exist, no district has the bandwidth and resources to do it themselves, no state is incentivized to tackle the problem, and no philanthropic actor will fund an effort with a long, unclear path to adoption and prominence.

National Laboratory Schools are intended primarily as research, development, and demonstration efforts, meaning that they will be staffed largely by researchers and will pursue research agendas that go beyond the traditional responsibilities and expertise of local school districts. State and local actors are the right entities to design and operate these schools so that they reflect the particular priorities and strengths of local communities, and so that each school is well positioned to influence local practice. But funding and overseeing the National Laboratory School system as a whole is an appropriate role for the federal government.

2. Why is NSF the right agency to lead this work?

For many years, NSF has developed substantial expertise funding innovation through the SBIR/STTR programs, which award staged grants to support innovation and technology transfer. NSF also has experience researching education through its Directorate for Education and Human Resources (HER). Finally, NSF’s new Directorate for Technology, Innovation, and Partnerships (TIP) has a mandate to “[create] education pathways for every American to pursue new, high-wage, good-quality jobs, supporting a diverse workforce of researchers, practitioners, and entrepreneurs.” NSF is the right agency to lead the National Laboratory Schools program because of its unique combination of experience, in-house expertise, mission relevance, and relationships with agencies, industry, and academia.

3. What role will OSTP play in establishing the National Laboratory School program? Why should they help lead the program instead of ED?

ED focuses on the concerns and priorities of existing schools. Ensuring that National Laboratory Schools emphasize invention and reimagining of educational models requires fresh strategic thinking and partnerships grounded in computational domain expertise.

OSTP has access to bodies like the President’s Council of Advisors on Science and Technology (PCAST)and the National Science and Technology Council (NSTC). Working with these bodies, OSTP can easily convene high-profile leaders in computation from industry and academia to publicize and support the National Laboratory Schools program. OSTP can also enlist domain experts who can act as advisors evaluating and critiquing the depth of computational work developed in the Laboratory Schools. And annually, in the spirit of the White House Science Fair, OSTP could host a festival showcasing the design, practices, and outputs of various Laboratory Schools.

Though OSTP and NSF will have primary leadership responsibilities for the National Laboratory Schools program, we expect that ED will still be involved as a key partner on topics aligned with ED’s core competencies (e.g., regulatory compliance, traditional best practices, responsible research practices, etc.).

4. What makes the Department of Defense Education Activity (DoDEA) an especially good partner for this work?

The DoDEA is an especially good partner because it is the only federal agency that already operates schools; reaches a student base that is large (more than 70,000 students, of whom more than 12,000 are high-school aged) as well as academically, socioeconomically, and demographically diverse; more nimble than a traditional district; in a position to appreciate and understand the full ramifications of the computational revolution; and very motivated to improve school quality and reduce turnover

5. Why should the Division for School Invention (DSI) be situated within NSF’s TIP Directorate rather than EHR Directorate?

EHR has historically focused on the important work of researching (and to some extent, improving) existing schools. The DSI’s focus on invention, secondary/postsecondary education, and opportunities for alignment between cluster-based workforce-development strategies and Laboratory Schools’ computational emphasis make the DSI a much better fit for the TIP, which is not only focused on innovation and invention overall, but is also explicitly tasked with “[creating] education pathways for every American to pursue new, high-wage, good-quality jobs, supporting a diverse workforce of researchers, practitioners, and entrepreneurs.” Situating the DSI within TIP will not preclude DSI from drawing on EHR’s considerable expertise when needed, especially for evaluating, contextualizing, and supporting the research agendas of Laboratory Schools.

6. Why shouldn’t existing public schools be eligible to serve as Laboratory Schools?

Most attempts at organizational change fail. Invention requires starting fresh. Allowing existing public schools or districts to launch Laboratory Schools will distract from the ongoing educational missions of those schools and is unlikely to lead to effective invention. 

7. Who are some appropriate partners for the National Laboratory School program?

Possible partners include:

8. What should the profile of a team or organization starting a Laboratory School look like? Where and how will partners find these people?

At a minimum, the team should have experience working with youth, possess domain expertise in computation, be comfortable supporting both technical and expressive applications of computation, and have a clear vision for the practical operation of their proposed educational model across both the humanities and technical fields.

Ideally, the team should also have piloted versions of their proposed educational model approach in some form, such as through after-school programs or at a summer camp. Piloting novel educational models can be hard, so the DSI and/or its partners may want to consider providing tiered grants to support this kind of prototyping and develop a pipeline of candidates for running a Laboratory School.

To identify candidates to launch and operate a Laboratory School, the DSI and/or its partners can:

What

1. What is computational thinking, and how is it different from programming or computer science?

A good way to answer this question is to consider writing as an analogy. Writing is a tool for thought that can be used to think critically, persuade, illustrate, and so on. Becoming a skilled writer starts with learning the alphabet and basic grammar, and can include craft elements like penmanship. But the practice of writing is distinct from the thinking one does with those skills. Similarly, programming is analogous to mechanical writing skills, while computer science is analogous to the broader field of linguistics. These are valuable skills, but are a very particular slice of what the computational revolution entails.

Both programming and computer science are distinct from computational thinking. Computational thinking refers to thinking with computers, rather than thinking about how to communicate problems and questions and models to computers. Examples in other fields include:

These transitions each involve programming, but are no more “about” computer science than a philosophy class is “about” writing. Programming is the tool, not the topic.

2. What are some examples of the research questions that National Laboratory Schools would investigate?

There are countless research agendas that could be pursued through this new infrastructure. Select examples include:

  1. Seymour Papert’s work on LOGO (captured in books like Mindstorms) presented a radically different vision for the potential and role for technology in learning. In Mindstorms, Papert sketches out that vision vis a vis geometry as an existence proof. Papert’s work demonstrates that research into making things more learnable differs from researching how to teach more effectively. Abelson and diSessa’s Turtle Geometry takes Papert’s work further, conceiving of ways that computational tools can be used to introduce differential geometry and topology to middle- and high-schoolers. The National Laboratory Schools could investigate how we might design integrated curricula combining geometry, physics, and mathematics by leveraging the fact that the vast majority of mathematical ideas tackled in secondary contexts appear in computational treatments of shape and motion.
  2. The Picturing to Learn program demonstrated remarkable results in helping staff to identify and students to articulate conceptions and misconceptions. The National Laboratory Schools could investigate how to take advantage of the explosion of interactive and dynamic media now available for visually thinking and animating mental models across disciplines.
  3. Bond graphs as a representation of physical dynamic systems were developed in the 1960s. These graphs enabled identification of “effort” and “flow” variables as new ways of defining power. This in turn allowed us to formalize analogies across electricity and magnetism, mechanics, fluid dynamics, and so on. Decades later, category theory has brought additional mathematical tools to bear on further formalizing these analogies. Given the role of analogy in learning, how could we reconceive people’s introduction to natural sciences in cross-disciplinary language emphasizing these formal parallels.
  4. Understanding what it means for one thing to cause (or not cause) another, and how we attempt to establish whether this is empirically true is an urgent and omnipresent need. Computational approaches have transformed economics and the social sciences: Whether COVID vaccine reliability, claims of election fraud, or the replication crisis in medicine and social science, our world is full of increasingly opaque systems and phenomena which our media environment is decreasingly equipped to tackle for and with us. An important tool in this work is the ability to reason about and evaluate empirical research effectively, which in turn depends on fundamental ideas about causality and how to evaluate the strength and likelihood of various claims. Graphical methods in statistics offer a new tool complementing traditional, easily misused ideas like p-values which dominate current introductions to statistics without leaving youth in a better position to meaningfully evaluate and understand statistical inference.

The specifics of these are less important than the fact that there are many, many such agendas that go largely unexplored because we lack the tangible infrastructure to set ambitious, computationally sophisticated educational research agendas.

3. How will the National Laboratory Schools differ from magnet schools for those interested in computer science?

The premise of the National Laboratory Schools is that computation, like writing, can transform many subjects. These schools won’t place disproportionate emphasis on the field of computer science, but rather will emphasize integration of computational thinking into all disciplines—and educational practice as a whole. Moreover, magnet schools often use selective enrollment in their admissions. National Laboratory Schools are public schools interested in the core issues of the median public school, and therefore it is important they tackle the full range of challenges and opportunities that public schools face. This involves enrolling a socioeconomically, demographically, and academically diverse group of youth.

4. How will the National Laboratory Schools differ from the Institute for Education Science’s Regional Education Laboratories?

The Institute for Education’s (IES’s) Regional Education Laboratories (RELs) do not operate schools. Instead, they convene and partner with local policymakers to lead applied research and development, often focused on actionable best practices for today’s schools (as exemplified by the What Works Clearinghouse). This is a valuable service for educators and policymakers. However, this service is by definition limited to existing school models and assumptions about education. It does not attempt to pioneer new school models or curricula.

5. How will the National Laboratory Schools program differ from tech-focused workforce-development initiatives, coding bootcamps, and similar programs?

These types of programs focus on the training and placement of software engineers, data scientists, user-experience designers, and similar tech professionals. But just as computational thinking is broader than just programming, the National Laboratory Schools program is broader than vocational training (important as that may be). The National Laboratory Schools program is about rethinking school in light of the computational revolution’s effect on all subjects, as well as its effects on how school could or should operate. An increased sensitivity to vocational opportunities in software is only a small piece of that.

6. Can computation really change classes other than math and science?

Yes. The easiest way to prove this is to consider how professional practice of non-STEM fields has been transformed by computation. In economics, the role of data has become increasingly prominent in both research and decision making. Data-driven approaches have similarly transformed social science, while also expanding the field’s remit to include specifically online, computational phenomena (like social networks). Politics is increasingly dominated by technological questions, such as hacking and election interference. 3D modeling, animation, computational art, and electronic music are just a few examples of the computational revolution in the arts. In English and language arts, multimedia forms of narrative and commentary (e.g., podcasts, audiobooks, YouTube channels, social media, etc.) are augmenting traditional books, essays, and poems. 

7. Why and how should National Laboratory Schools commit to financial and legal parity with public schools?

The challenges facing public schools are not purely pedagogical. Public schools face challenges in serving diverse populations in resource-constrained and highly regulated environments. Solutions and innovation in education need to be prototyped in realistic model systems. Hence the National Laboratory Schools must commit to financial and legal parity with public schools. At a minimum, this should include a commitment to (i) a per-capita student cost that is no more than twice the average of the relevant catchment area for a given National Laboratory School (the 2x buffer is provided to accommodate the inevitably higher cost of prototyping educational practices at a small scale), and (ii) enrollment that is demographically and academically representative (including special-education and English Language Learner participation) of a similarly aged population within thirty minutes’ commute, and that is enrolled through a weighted lottery or similarly non-selective admissions process.

8. Why are Xerox PARC and the Mayo Clinic good models for this initiative?

Both Xerox PARC and the Mayo Clinic are prototypical examples of hyper-creative, highly-functioning research and development laboratories. Key to their success inventing the future was living it themselves.

PARC researchers insisted on not only building but using their creations as their main computing systems. In doing so, they were able to invent everything from ethernet and the laser printer to the whole paradigm of personal computing (including peripherals like the modern mouse and features like windowed applications that we take for granted today).

The Mayo Clinic runs an actual hospital. This allows the clinic to innovate freely in everything from management to medicine. As a result, the clinic created the first multi-specialty group practice and integrated medical record system, invented the oxygen mask and G-suit, discovered cortisone, and performed the first hip replacement.

One characteristic these two institutions share is that they are focused on applied design research rather than basic science. PARC combined basic innovations in microelectronics and user interface to realize a vision of personal computing. Mayo rethinks how to organize and capitalize on medical expertise to invent new workflows, devices, and more.

These kinds of living laboratories are informed by what happens outside their walls but are focused on inventing new things within. National Laboratory Schools should similarly strive to demonstrate the future in real-world operation.

Why?

1. Don’t laboratory schools already exist? Like at the University of Chicago?

Yes. But there are very few of them, and almost all of those that do exist suffer from one or more issues relative to the vision proposed herein for National Laboratory Schools. First, most existing laboratory schools are not public. In fact, most university-affiliated laboratory schools have, over time, evolved to mainly serve faculty’s children. This means that their enrollment is not socioeconomically, demographically, or academically representative. It also means that families’ risk aversion may constrain those schools’ capacity to truly innovate. Most laboratory schools not affiliated with a university use their “laboratory” status as a brand differentiator in the progressive independent-school sector.

Second, the research functions of many laboratory schools have been hollowed out given the absence of robust funding. These schools may engage in shallow renditions of participatory action research by faculty in lieu of meaningful, ambitious research efforts. 

Third, most educational-design questions investigated by laboratory schools are investigated at the classroom or curriculum (rather than school design) level. This creates tension between those seeking to test innovative practices (e.g., a lesson plan that involves an extended project) and the constraints of traditional classrooms.

Finally, insofar as bona fide research does happen, it is constrained by what is funded, publishable, and tenurable within traditional graduate schools of education. Hence most research reflects the concerns of existing schools instead of seeking to reimagine school design and educational practice.

2. Why will National Laboratory Schools succeed where past efforts at educational reform (e.g., charter schools) have failed?

Most past educational-reform initiatives have focused on either supporting and improving existing schools (e.g., through improved curricula for standard classes), or on subsidizing and supporting new schools (e.g., charter schools) that represent only minor departures from traditional models.

The National Laboratory Schools program will provide a new research, design, and development infrastructure for inventing new school models, curricula, and educator training. These schools will have resources, in-house expertise, and research priorities that traditional public schools—whether district or charter or pilot—do not and should not. If the National Laboratory Schools are successful, their output will help inform educational practice across the U.S. school ecosystem. 

3. Don’t charter schools and pilot schools already support experimentation? Wasn’t that the original idea for charter and pilot schools—that they’d be a laboratory to funnel innovation back into public schools?

Yes, but this transfer hasn’t happened for at least two reasons. First, the vast majority of charter and pilot schools are not pursuing fundamentally new models because doing so is too costly and risky. Charter schools can often perform more effectively than traditional public schools, but this is just as often because of problematic selection bias in enrollment as it is because the autonomy they’re given allows for more effective leadership and organizational management. Second, the politics around charter and pilots has become increasingly toxic in many places, which prevents new ideas from being considered by public schools or advocated for effectively by public leaders.

4. Why do we need invention at the school rather than at the classroom level? Wouldn’t it be better to figure out how to improve schools that exist rather than end up with some unworkable model that most districts can’t adopt?

The solutions we need might not exist at the classroom level. We invest a great deal of time, money, and effort into improving existing schools. But we underinvest in inventing fundamentally different schools. There are many design choices which we need to explore which cannot be adequately developed through marginal improvements to existing models. One example is project-based learning, wherein students undertake significant, often multidisciplinary projects to develop their skills. Project-based learning at any serious level requires significant blocks of time that don’t fit in traditional school schedules and calendars. A second example is the role of computational thinking, as centered in this proposal. Meaningfully incorporating computational approaches into a school design requires new pedagogies, developing novel tools and curricula, and re-training staff. Vanishingly few organizations do this kind of work as a result.

If and when National Laboratory Schools develop substantially innovative models that demonstrate significant value, there will surely need to be a translation process to enable districts to adopt these innovations, much as translational medicine brings biomedical innovations from the lab to the hospital. That process will likely need to involve helping districts start and grow new schools gradually, rather then district-wide overhauls.

5. What kinds of “traditional assumptions” need to be revisited at the school level?

The basic model of school assumes subject-based classes with traditionally licensed teachers lecturing in each class for 40–90 minutes a day. Students do homework, take quizzes and tests, and occasionally do labs or projects. The courses taught are largely fixed, with some flexibility around the edges (e.g., through electives and during students’ junior and senior high-school years).

Traditional school represents a compromise among curriculum developers, standardized-testing outfits, teacher-licensure programs, regulations, local stakeholder politics, and teachers’ unions. Attempts to change traditional schools almost always fail because of pressures from one or more of these groups. The only way to achieve meaningful educational reform is to demonstrate success in a school environment rethought from the ground up. Consider a typical course sequence of Algebra I, Geometry, Algebra II, and Calculus. There are both pedagogical and vocational reasons to rethink this sequence and instead center types of mathematics that are more useful in computational contexts (like discrete mathematics and linear algebra). But a typical school will not be able to simultaneously develop the new tools, materials, and teachers needed to do so.

6. Has anything like the National Laboratory School program been tried before?

No. There have been various attempts to promote research in education without starting new schools. There have been interesting attempts by states to start new schools (like Governor’s Schools),there have been some ambitious charter schools, and there have been attempts to create STEM-focused and computationally focused magnet schools. But there has never been a concerted attempt in the United States to establish a new kind of research infrastructure built atop the foundation of functioning schools as educational “sandboxes”.

How?

1. How will we pay for all this? What existing funding streams will support this work? Where will the rest of the money for this program come from?

For budgeting purposes, assume that each Laboratory School enrolls a small group of forty high school or community college students full-time at an average per capita rate of $40,000 per person per year. Half of that budget will support the functioning of schools themselves. The remaining half will support a small research and development team responsible for curating and developing the computational tools, materials, and curricula needed to support the School’s educators. This would put the direct service budget of the school solidly at the 80th percentile of current per capita spending on K–12 education in the United States.With these assumptions, running 100 National Laboratory Schools would cost ~$160 million. Investing $25 million per year would be sufficient to establish an initial 15 sites. This initial federal funding should be awarded through a 1:1 matching competitive-grant program funded by (i) the 10% of American Competitiveness and Workforce Improvement Act (ACWIA) Fees associated with H1-B visas (which the NSF is statutorily required to devote to public-private partnerships advancing STEM education), and (ii) the NSF TIP Directorate’s budget, alongside budgets from partner agency programs (for instance, the Department of Education’s Education Innovation and Research and Investing in Innovation programs). For many states, these funds should also be layered atop their existing Elementary and Secondary School Emergency Relief (ESSER) and American Rescue Plan (ARP) awards.

2. Why is vertical integration important? Do we really need to run schools to figure things out?

Vertical integration (of research, design, and operation of a school) is essential because schools and teacher education programs cannot be redesigned incrementally. Even when compelling curricular alternatives have been developed under the auspices of an organization like the NSF, practical challenges in bringing those innovations to practice have proven insurmountable. In healthcare, the entire field of translational medicine exists to help translate research into practice. Education has no equivalent.

The vertically integrated National Laboratory School system will address this gap by allowing experimenters to control all relevant aspects of the learning environment, curricula, staffing, schedules, evaluation mechanisms, and so on. This means the Laboratory Schools can demonstrate a fundamentally different approach, learning from great research labs like Xerox PARC and the Mayo Clinic, much of whose success depended on tightly-knit, cross-disciplinary teams working closely together in an integrated environment.

3. What would the responsibilities of a participating agency look like in a typical National Laboratory School partnership?

A participating agency will have some sort of educational or workforce-development initiative that would benefit from the addition of a National Laboratory School as a component. This agency would minimally be responsible for:

4. How should success for individual Laboratory Schools be defined?

Working with the Institute of Education Sciences (IES)’ National Center for Education Research(NCER), the DSI should develop frameworks for collecting necessary qualitative and quantitative data to document, understand, and evaluate the design of any given Laboratory School. Evaluation would include evaluation of compliance with financial and legal parity requirements as well as evaluation of student growth and work products.

Evaluation processes should include:

Success should be judged by a panel of experts that includes domain experts, youthworkers and/or school leaders, and DSI leadership. Dimensions of performance these panels should address should minimally include depth and quality of students’ work, degree of traditional academic coverage, ambition and coherence of the research agenda (and progress on that research agenda), retention of an equitably composed student cohort, and growth (not absolute performance) on the diagnostic/formative assessments.In designing evaluation mechanisms, it will be essential to learn from failed accountability systems in public schools. Specifically:, it will be essential to avoid pushing National Laboratory Schools to optimize for the particular metrics and measurements used in the evaluation process. This means that the evaluation process should be largely based on holistic evaluations made by expert panels rather than fixed rubrics or similar inflexible mechanisms. Evaluation timescales should also be selected appropriately: e.g., performance on diagnostic/formative assessments should be measured by examining trends over several years rather than year-to-year changes.

5. What makes the Small Business Innovation Research (SBIR) program a good model for the National Laboratory School program?

The SBIR program is a competitive grant competition wherein small businesses submit proposals to a multiphase grant program. SBIR awards smaller grants (~$150,000) to businesses at early stages of development, and makes larger grants (~$1 million) available to awardees who achieve certain progress milestones. SBIR and similar federal tiered-grant programs (e.g., the Small Business Technology Transfer, or STTR, program) have proven remarkably productive and cost-effective, with many studies highlighting that they are as or more efficient on a per-dollar basis when compared to the private sector via common measures of innovation like number of patents, papers, and so on.

The SBIR program is a good model for the National Laboratory School program; it is an example of the federal government promoting innovation by patching a hole in the funding landscape. Traditional financing options for businesses are often limited to debt or equity, and most providers of debt (like retail banks) for small businesses are rarely able or incentivized to subsidize research and development. Venture capitalists typically only subsidize research and development for businesses and technologies with reasonable expectations of delivering 10x or greater returns. SBIR provides funding for the innumerable businesses that need research and development support in order to become viable, but aren’t likely to deliver venture-scale returns.

In education, the funding landscape for research and development is even worse. There are virtually no sources of capital that support people to start schools, in part because the political climate around new schools can be so fraught. The funding that does exist for this purpose tends to demand school launch within 12–18 months: a timescale upon which it is not feasible to design, evaluate, refine an entirely new school model. Education is a slow, expensive public good: one that the federal government shouldn’t provision, but should certainly subsidize. That includes subsidizing the research and development needed to make education better.

States and local school districts lack the resources and incentives to fund such deep educational research. That is why the federal government should step in. By running a tiered educational research-grant program, the federal government will establish a clear pathway for prototyping and launching ambitious and innovative schools.

6. What protections will be in place for students enrolled in Laboratory Schools?

The state organizations established or selected to oversee Laboratory Schools will be responsible for approving proposed educational practices. That said, unlike in STEM fields, there is no “lab bench” for educational research: the only way we can advance the field as a whole is by carefully prototyping informed innovations with real students in real classrooms.

7. Considering the challenges and relatively low uptake of educational practices documented in the What Works Clearinghouse, how do we know that practices proven in National Laboratory Schools will become widely adopted?

National Laboratory Schools will yield at least three kinds of outputs, each of which is associated with different opportunities and challenges with respect to widespread adoption.

The first output is people. Faculty trained at National Laboratory Schools (and at possible educator-development programs run within the Schools) will be well positioned to take the practices and perspectives of National Laboratory Schools elsewhere (e.g., as school founders or department heads). The DSI should consider establishing programs to incentivize and support alumni personnel of National Laboratory Schools in disseminating their knowledge broadly, especially by founding schools.

The second output is tools and materials. New educational models that are responsive to the computational revolution will inevitably require new tools and materials—including subject-specific curricula, cross-disciplinary software tools for analysis and visualization, and organizational and administrative tools—to implement in practice. Many of these tools and materials will likely be adaptations and extensions of existing tools and materials to the needs of education.

The final output is new educational practices and models. This will be the hardest, but probably most important, output to disseminate broadly. The history of education reform is littered with failed attempts to scale or replicate new educational models. An educational model is best understood as the operating habits of a highly functioning school. Institutionalizing those habits is largely about developing the skills and culture of a school’s staff (especially its leadership). This is best tackled not as a problem of organizational transformation (e.g., attempting to retrofit existing schools), but rather one of organizational creation—that is, it is better to use models as inspirations to emulate as new schools (and new programs within schools) are planned. Over time, such new and inspired schools and programs will supplant older models.

8. How could the National Laboratory School program fail?

Examples of potential pitfalls that the DSI must strive to avoid include:

Creating a National Fellowship for Entrepreneurial Scientists and Engineers

Summary

The next administration should establish a national fellowship for scientists and engineers to accelerate the transformation of research discoveries into scalable, market-ready technologies. Entrepreneurship is driving innovation across the U.S. economy—with the troubling exception of early-stage science. Transitioning scientific discoveries from the laboratory into prototypes remains too speculative and costly to garner significant support from industry or venture-capital firms. This makes it difficult for many of our nation’s science innovators to translate their research into new products and puts the United States at risk of falling behind in the quickly evolving global economy.

Entrepreneurial fellowships for scientists and engineers have emerged as an effective strategy for translating research into new products and businesses, showing tremendous early impact and a readiness to scale. The next administration should advance this proven strategy at the federal level by creating a national entrepreneurial fellowship. This new entrepreneurial fellowship would leverage our nation’s investments in science to drive national prosperity, security, and global competitiveness.