Article Introduction

Author:Michael F. Joseph, Assistant Professor, Department of Political Science, University of California San Diego; Michael Poznansky, Associate Professor, Strategy and Operations Research Department, U.S. Naval War College

Source text: Joseph, Michael F., and Michael Poznansky. “Secret Innovation.” *International Organization* 78, no. 4 (2024): 766–799.

CompileZhu Kaize, PhD candidate in the Department of Public and International Affairs at City University of Hong Kong

IntroductionThe conventional wisdom is that open, collaborative, and transparent organizations are more innovative, as they help break down departmental silos and foster knowledge flow and information sharing. However, many groundbreaking technological innovations of history—such as satellites, lithium batteries, and the internet—often originated from small, secretive teams in national security domains. Does the innovativeness of these organizations stem from their secrecy, or is it unrelated to their organizational form? This paper constructs a principal-agent model of public sector innovation to explore how secrecy can play a role in the early stages of high-risk, high-reward projects.

Research has found that secrecy helps drive project success through initial stages via a “cost shifting” mechanism. In transparent organizations, managers often avoid approving high-risk projects due to fear of political costs. However, in a secretive environment, researchers take on more personal costs to tentatively advance projects, and then use the evidence collected to persuade managers to approve them.This mechanism was validated in two cases from the early Cold War: the CIA's MKULTRA mind control program and the CORONA reconnaissance satellite program. This study offers new perspectives for understanding the technological catalysis of secrecy systems and innovation strategies in great power competition.

01 Theoretical Framework: The Paradox of Innovation

National prosperity and security depend on continuous innovation. It is widely believed in academia that transparent organizations that encourage internal information sharing foster innovation, as such organizations can break down departmental barriers, enhance competition and cooperation, promote knowledge flow, and drive the free dissemination of information. However, some institutions defy this principle. For instance, national security agencies like the U.S. Central Intelligence Agency and the British Secret Intelligence Service are known for their information secrecy, yet they continue to produce significant technological breakthroughs, a phenomenon that challenges the consensus that “openness promotes innovation.”

To explain this phenomenon, this paper constructs a two-stage innovation model: Researchers (R) can choose an open or secret path to advance their projects in the initial stage, while managers (D) decide whether to approve based on expected benefits and political costs. The model introduces the following key parameters: research cost ($k_i$) and deployment cost ($c_i$) characterize the sensitivity of behavioral actors to controversial activities; the cost-sharing coefficient ($x \in (0,1)$) quantifies the extent to which managers bear only a portion of the responsibility costs due to “plausible deniability.” The core logic of the model is that when the expected net benefit to managers ($e_0$) under the open path falls below the research cost ($k_D$), the secret path becomes the only option. Researchers advance trials by bearing initial costs ($k_R$). If the results are positive, managers approve the subsequent stage with $x$ proportion of the costs. The dynamic balance between parameters makes it easier for secret organizations to initiate high-risk projects.

02 Cost pass-through mechanism

Regarding how confidentiality plays a role in innovation, this paper proposes a core mechanism: the cost-passing mechanism. In transparent organizations, managers are responsible for the full lifecycle risks of a project (such as public scrutiny and accusations of fund misuse). Therefore, due to political cost considerations, they tend to avoid high-risk exploration in the early stages. In contrast, in confidential organizations, researchers secretly advance trials by bearing the initial costs themselves (such as professional reputation and legal accountability). They then seek management approval after obtaining evidence of feasibility, shifting the deployment costs (such as large-scale capital investment and international public opinion pressure) to the institution as a whole. Thus, the core role of confidentiality lies in reallocating political costs within the innovation process. This mechanism also enables confidential organizations to overcome the efficiency advantages of open organizations and focus on high-risk, high-reward technological exploration.

03 “Don't ask, don't tell” strategy

The research findings also present new challenges to the principal-agent theory in the field of international relations. First, even if researchers and managers have perfectly aligned goals, researchers may still choose to leverage secrecy to advance their research. Second, even if monitoring costs are extremely low and perfect information is available, managers may still refrain from monitoring. If managers actively seek to understand project details, they will be held fully responsible for potential ethical or political risks. Conversely, by maintaining “reasonable ignorance,” they can share in the benefits of success while bearing less responsibility if a scandal erupts by claiming “ignorance.” This “Don't Ask, Don't Tell” interaction model allows managers to greenlight projects at opportune moments without explicitly knowing the research details.

Taking as an example Project MKUltra, a secret research program initiated by the U.S. Central Intelligence Agency (CIA) in the 1950s to explore mind control techniques, its core objective was to manipulate human behavior through drugs and psychological means. The program's secrecy mechanisms were exemplary: funds were allocated through “unaccounted-for accounts,” eliminating the need to submit expenditure details to Congress; only the core team knew the experimental details, and before the scandal broke in 1973, director Sidney Gottlieb was ordered to destroy 90%study records; after experimental subject Frank Olson died from LSD hallucinations, CIA Director Allen Dulles only issued an informal warning to Gottlieb, avoiding written repercussions. The 1973 Church Committee investigation found that these experiments “lasted for many years and revealed a fundamental disregard for ethical boundaries,” especially the scale of experiments testing hallucinogens on unsuspecting citizens and the disregard for the value of human life, sparking severe ethical controversies.

The success of the CORONA reconnaissance satellite program is another classic case demonstrating how secrecy mechanisms can function in high-risk, high-reward projects. In 1955, the U.S. Air Force rejected the WS-117L satellite program due to technical risks and budget pressures. The research team employed a “Second Story” strategy, publicly announcing the cancellation of the WS-117L program and repackaging it as a seemingly harmless “Earth Science Experiment” as a cover. This circumvented the political controversy of “militarizing space.” Meanwhile, the program was secretly restarted with the support of the CIA. Funding bypassed congressional review through a handwritten directive from President Eisenhower, which simply stated, “Proceed at once without hearings.” This dual strategy not only avoided political controversy but also secured valuable research and development time for the project. Ultimately, the program launched a cumulative total of 144 satellites, generating over $1 trillion in economic benefits and establishing U.S. dominance in space reconnaissance.

04 Conclusion

Research indicates that secretive organizations tend to favor innovative projects with highly uncertain prospects, while transparent organizations lean towards incremental innovation. This difference is not only evident in the technological realm but also reflected in organizational culture and decision-making logic. By controlling information flow and reducing political risks, secretive organizations drive the development of high-risk, high-reward projects. Conversely, transparent organizations rely on broad information sharing and collaboration to enhance innovation efficiency. Simultaneously, secrecy systems provide political shelter for high-risk technology exploration through cost transfer mechanisms, but their success rate is less than 5%and accompanied by ethical control risks. Future technology governance needs to balance these two types of systems: secretive organizations focus on the initial validation of breakthrough technologies, while transparent organizations promote the incremental optimization of mature technologies. Institutional diversity is key to maintaining the innovation ecosystem—it requires tolerating the ‘controlled black box’ of secretive organizations while also constraining their risks through ethical review mechanisms at the stage of outcome.