Abstract: Is more novel research always desirable? We develop a model in which knowledge shapes society's policies and guides the search for discoveries. Researchers select a question and how intensely to study it. The novelty of a question determines both the value and difficulty of discovering its answer. We show that the benefits of discoveries are nonmonotone in novelty. Knowledge expands endogenously step‐by‐step over time. Through a dynamic externality, moonshots—research on questions more novel than what is myopically optimal—can improve the evolution of knowledge. Moonshots induce research cycles in which subsequent researchers connect the moonshot to previous knowledge.
The Tension in Scientific Discovery
Scientific progress often oscillates between two poles: incremental refinement and radical novelty. While "scientific freedom" allows researchers to choose their own paths, they frequently gravitate toward "safe" problems—those Nobel laureate Peter Medawar called "the art of the soluble." This paper develops a formal economic model to investigate whether more novelty is always desirable and how the current state of knowledge shapes future discoveries.
Modeling the Evolution of Knowledge
The authors represent the "universe of questions" on a real line, where the underlying truth follows a realized path of a Brownian motion. This captures the fundamental intuition that similar questions tend to have correlated answers. "Knowledge" consists of specific known points on this path. For any unasked question, society forms a conjecture—a normal distribution—based on the nearest known points.
Society derives value from this knowledge by using these conjectures to set policies. The precision of a conjecture (its variance) determines the payoff: more precise knowledge leads to better decision-making. Researchers, in turn, choose which questions to study (their "novelty" or distance from existing knowledge) and how much effort to exert. Crucially, the cost of research is endogenous; it is cheaper to find an answer when existing conjectures are already precise.
The "Sweet Spot" of Novelty
A key insight of the model is that the benefits of discovery are non-monotone in novelty. If a question is too close to existing knowledge, it adds little new information. Conversely, if it is too far, it remains too disconnected to improve society's general understanding. This creates an intermediate "optimal" distance for immediate value.
However, individual researchers are often "myopic." They focus on their own immediate success and ignore the "dynamic externality" of their work—how their discovery might guide and lower the costs for future generations. Left to their own devices, researchers tend to produce a "ladder" of knowledge: small, sequential steps that expand the frontier slowly.
Moonshots and Research Cycles
The paper highlights the transformative power of "moonshots"—research on questions far beyond the current frontier. While a moonshot may have low immediate value due to its isolation, it acts as vital infrastructure. By establishing a distant anchor point, it creates a "research area" between itself and the old frontier.
This gap "inspires" subsequent researchers to "deepen" knowledge by filling in the space. This process, termed a "research cycle," is often more productive than simple step-by-step expansion. As shown in the right panel of Figure 1, having two anchor points significantly reduces uncertainty in the region between them, making subsequent research less risky and more valuable.
Key Results and Policy Implications
The model demonstrates that moonshots can reduce the "hazard rate" of science—the probability that the chain of discovery breaks because a researcher fails to find an answer. By providing multiple anchor points, the model shows that subsequent research becomes more robust.
From a policy perspective, a social planner (such as a grant agency) should incentivize moonshots specifically when research costs are intermediate and society is patient. The model also shows that research output peaks at intermediate levels of novelty; very incremental research has high success but low value, while extreme novelty faces prohibitive costs.
Limitations and Future Directions
While the 1D model provides clear insights, real-world knowledge is multidimensional. The authors suggest that future work could explore "trees" or "oceans" of questions. Additionally, the model focuses on sequential researchers, leaving room to explore the effects of competition and "races" for discovery. Finally, incorporating more complex institutional frictions, such as publication bias or "winner-take-all" reward structures, would further refine our understanding of the quest for knowledge.
Summary updated