Alternative Structures for Research Institutions

Research Systems Shape What Gets Built

Academia and industry are both essential — and both incomplete

If science is fundamentally about experimentation, then we should also be willing to experiment with how research itself is conducted. Rethinking the structures, incentives, and funding models that underpin research institutions could unlock entirely new modes of discovery. This is at the heart of the growing field of metascience.

Universities produce most basic science: curiosity-driven, peer-reviewed, publicly funded. Industry translates discoveries into products. Together they have delivered extraordinary advances. But both are structurally limited in ways that create significant blind spots — and many of the most pressing challenges ahead require exactly the kind of work neither is designed to do.

But some of the most important breakthroughs in science and technology have emerged from environments that were designed differently. For example, Bell Labs produced the transistor and foundational work in information theory under a long-horizon industrial research model that was not constrained by immediate publication or product pressures. Similarly, DARPA enabled high-risk, programmatic research that led to breakthroughs such as the early internet, by funding ambitious bets without requiring traditional academic outputs.

These cases are not exceptions in outcome—they are exceptions in structure.

The Gap

Traditional structures can't do it all

Academic systems tend to reward publication volume and citation impact. Industrial systems tend to prioritize productization and revenue. Both can produce innovation, but both systematically filter out work that is uncertain, long-horizon, or non-incremental.

This duopoly of institutional models and their respective incentives don’t just shape how R&D is conducted; it constrains how we think about what’s possible. In doing so, it caps the ceiling of what research can achieve.

Together, these systems are powerful at producing incremental progress and deployable innovation—but leave a gap for long-horizon, high-risk, pre-competitive scientific work.

The Vision

We can design research institutions to do what others can't

The goal is not to replace academia or industry, or to simply make R&D more efficient. It is to build complementary structures that can pursue what neither can: ambitious, coordinated, long-horizon work oriented toward public goods. These institutions would be defined by stable long-term funding, explicit tolerance for failure, and success metrics beyond publications or profit.

Success would look like:

The Solution

Experimentation in research structures isn’t new—but we need more of it

Some institutions have already begun to redesign these constraints.

Example 1:

DARPA is an R&D agency funded by the US Department of Defense. Unlike most government R&D agencies, DARPA has engineered a distinctive incentive environment that is less bureaucratic and more entrepreneurial, systematically favoring risk-taking, speed, and breakthrough outcomes. DARPA hires Program Managers (PMs) as senior technical experts on short-term contracts (~3–5 years) to pursue specific, high-ambition goals, and gives them substantial control over funding allocations across research performers in industry, academia, and contracting organizations.

This structure reshapes incentives in four key ways:

1. Portfolio-level autonomy. Managing multiple parallel efforts rather than a single grant as a PI—enables active steering, rapid resource reallocation, and iterative course correction across competing technical approaches;
2. Time-bounded roles with no tenure create urgency to deliver meaningful impact within a fixed window rather than accumulate long-term institutional security;
3. Reputation becomes the primary currency, so career upside is driven by launching field-defining breakthroughs rather than publishing papers or generating near-term profits, reinforcing originality and high-risk bets; and
4. Explicit tolerance for failure means leadership expects a significant fraction of programs to fail, as long as they are well-designed, high-risk attempts.

Together, these incentives shift behavior toward proposing and pursuing ideas that would rarely survive traditional peer review, but that have disproportionate upside if they succeed.

DARPA has become the poster child of effective, non-traditional institutional structures for breakthrough R&D and has been extensively studied as such. Its model has increasingly served as a blueprint for new "ARPA-style" agencies in the US and abroad, including ARPA-H, SPRIND in Germany, and ARIA in the UK.

Example 2:

The Howard Hughes Medical Institute is a core-funded, philanthropy-backed research organization that shifts support away from short-term, project-based grants (~3 years) toward long-term investigator funding (typically ~7 years). This structure gives researchers greater freedom to pursue creative, higher-risk work that is less constrained by immediate publication pressures, compared to traditionally NIH-funded science. However, while this model meaningfully rethinks funding incentives, HHMI's underlying institutional structure remains largely academia-shaped.

Example 3:

Another example is the Human Genome Project, which succeeded through large-scale international coordination and sustained funding (planned for 15 years, but completed in 13) rather than competitive grant cycles or private market incentives. While not a research institution itself, the structure of this project enabled a level of standardization, scale, and shared infrastructure that individual labs operating under traditional incentive systems could not have achieved.

An ecosystem of structures, each suited to different needs

Research is not one uniform activity. It includes everything from low-hanging, incremental work to high-risk bets, and from single-disciplined to tightly coordinated, cross-disciplinary efforts. Different structures and incentives are needed for different parts of this ecosystem.

Positions are approximate and reflect typical operating mode. Industrial Labs are shown as a historical reference — the model is rare today.

The Landscape

Six alternative structures

The following structures represent distinct approaches to organising, funding, and governing research outside of conventional academia and industry. For a comprehensive breakdown and comparison, see Structures.