03 · What You Need to Know
Understanding Dual-Use Research and Its Ethical Implications
What Does Dual-Use Research Mean?
Dual-use research concerns scientific work whose knowledge, methods, materials, or resulting capabilities can serve more than one purpose, including beneficial and harmful applications.
The concept is particularly prominent in discussions of biological research, where scientific findings intended to improve public health may also have implications for biological security.
However, dual-use concerns extend beyond biological sciences in broader research ethics discussions.
Advances in artificial intelligence may support accessibility and personalized learning while also enabling intrusive surveillance or discriminatory classification. Cybersecurity research may improve system defenses while producing knowledge relevant to exploitation.
These examples do not imply that all research in these fields is equally risky.
The significance of a dual-use concern depends on the capability created, its accessibility, the context of potential use, and the consequences of misuse.
Is Dual-Use Research the Same as Dual-Use Research of Concern?
No. The terms should be distinguished.
Dual-use research is a broad concept describing research with potentially beneficial and harmful applications.
Dual-use research of concern, often abbreviated DURC, is a more specific term used in particular research oversight frameworks.
In the United States, the government's policy on dual-use research of concern and pathogens with enhanced pandemic potential establishes defined oversight requirements for specified life-sciences research.
That policy should not be treated as a universal classification system for every AI, educational technology, or social science study.
An educational data analytics project may raise genuine dual-use ethical questions without falling within the scope of a biological DURC regulation.
How Does Dual-Use Research Differ From Ordinary Research Risk?
Research risk is a broader category. It may include physical, psychological, social, economic, privacy, or other harms arising during a study.
Dual-use concerns focus particularly on the possibility that research outputs or capabilities may subsequently be applied for harmful purposes.
Direct research risk
Potential harm arising from conducting the study, such as exposing participants' confidential information during data collection.
Dual-use risk
Potential harm arising when the knowledge, methods, or capabilities produced by the research are applied for a harmful purpose.
Research integrity concern
Potential problems involving honesty, accuracy, accountability, or compliance in conducting and reporting research.
These concerns can overlap. A study may be conducted with valid participant consent and accurate methods while producing a technology that later enables harmful surveillance.
Ethical review of the research process and assessment of downstream misuse therefore address related but different questions.
What Are Examples of Dual-Use Research?
| Research area |
Potential beneficial application |
Potential harmful application |
| Educational learning analytics |
Identify students who may benefit from additional academic support. |
Enable discriminatory profiling or punitive monitoring. |
| Generative artificial intelligence |
Support accessible explanations, translation, and educational content creation. |
Enable deceptive content or large-scale manipulation. |
| Emotion recognition research |
Investigate adaptive interfaces or accessibility applications. |
Support intrusive monitoring or unsupported judgments about individuals. |
| Cybersecurity research |
Identify vulnerabilities and strengthen defensive systems. |
Provide knowledge that may assist malicious exploitation. |
| Biological research |
Improve understanding of pathogens and disease prevention. |
Create information or capabilities relevant to biological threats. |
| Behavioral prediction |
Improve access to timely support and relevant services. |
Facilitate manipulative targeting or discriminatory decision-making. |
These are illustrative possibilities, not claims that each research area necessarily produces the listed harms. Whether a specific project creates a meaningful dual-use risk requires contextual assessment.
Why Can Research With Good Intentions Be Harmful?
Research intentions do not determine every subsequent application of scientific knowledge.
A researcher may develop a method for a particular beneficial purpose, while other actors later adapt the method to different institutional, commercial, or political objectives.
The same technical capability may have different consequences depending on governance, access, incentives, and the populations affected.
For example, predicting academic disengagement could help identify students needing support. The same predictions could be used to deny opportunities to students classified as unlikely to succeed.
The harmful application is not inevitable, but it is sufficiently connected to the capability that responsible researchers may need to consider it.
Scientific benefit should therefore be evaluated alongside foreseeable pathways to harm rather than treated as proof that misuse is impossible.
Does Dual-Use Mean Every Technology Has Exactly Two Uses?
No. The expression dual-use should not be interpreted as a claim that research has precisely one beneficial and one harmful application.
A single technology may support numerous legitimate applications and several forms of misuse.
For example, a language model may assist academic writing, accessibility, translation, administrative communication, and instructional design.
Its misuse possibilities may also differ across contexts.
The term highlights the coexistence of beneficial and harmful potential rather than a strict numerical classification.
Does Dual-Use Research Require Malicious Intent?
No. Dual-use risk can exist even when researchers, participants, and immediate collaborators have legitimate intentions.
Intent matters when evaluating individual responsibility for particular conduct, but it does not eliminate the possibility of harmful applications.
A researcher who publishes an analytical technique for educational support may have no intention of enabling discrimination.
Nevertheless, the technique may be adapted by another organization for decisions that disadvantage particular groups.
The ethical assessment should distinguish foreseeable misuse from accusations that the original researcher intended the harmful outcome.
Are Dual-Use Risks Limited to Natural Sciences?
No. Although formal dual-use oversight has a particularly developed history in the life sciences, broader dual-use questions can arise in computational and social research.
Studies of behavioral influence, information diffusion, predictive classification, and automated decision-making may produce capabilities with competing applications.
In educational technology, researchers may examine engagement prediction, identity verification, remote proctoring, and automated assessment.
Such systems may support legitimate educational functions while raising questions about surveillance, exclusion, or institutional power.
However, applying the broad dual-use concept to these fields should not imply that their regulatory treatment is identical to biological DURC oversight.
How Does Artificial Intelligence Complicate Dual-Use Research?
AI systems may be adaptable across multiple tasks and settings, sometimes beyond those originally anticipated by their developers.
Capabilities developed for benign applications may be incorporated into systems with different purposes.
For example, a model trained to classify patterns of student interaction may later be adapted for workforce monitoring or behavioral profiling.
The extent of risk depends on the model's actual capabilities, accessibility, data requirements, and deployment environment.
Researchers should avoid assuming that every AI model creates severe misuse risks merely because it can be repurposed.
Risk assessment should examine plausible applications and their consequences rather than rely on the general claim that AI is powerful.
Can Educational Technology Research Be Dual-Use?
Yes, in the broader ethical sense.
Educational technology research frequently produces methods for observing, classifying, predicting, or influencing learning behavior.
These capabilities may support personalized instruction, accessibility, timely intervention, or more informed educational decisions.
They may also be used to intensify surveillance, restrict opportunities, or justify decisions based on incomplete or biased indicators.
For example, a model predicting student dropout risk could inform supportive advising or become a mechanism for excluding applicants perceived as academically vulnerable.
Researchers should therefore examine how their outputs could be interpreted and used within actual educational institutions.
The ethical issue is not that prediction is inherently harmful, but that predictive information can be used under different institutional incentives and safeguards.
Is Dual-Use the Same as Unintended Consequences?
Not exactly. Unintended consequences include a wide range of outcomes that researchers or developers did not anticipate or desire.
Dual-use concerns focus more specifically on the possibility that knowledge or capabilities may support harmful applications alongside legitimate ones.
An educational platform that unexpectedly increases teachers' administrative workload may create an unintended consequence without necessarily presenting a classic dual-use problem.
By contrast, a behavioral prediction technique that can support academic assistance or punitive classification illustrates competing uses of the same capability.
The distinction is useful, although the categories can overlap in practice.
How Can Researchers Assess Whether Misuse Is Foreseeable?
Foreseeability involves examining whether a harmful application is reasonably connected to the research output and the circumstances in which it may be used.
Researchers should consider what capability the study creates, who could access it, what additional resources misuse would require, and what harms might follow.
A speculative scenario requiring numerous improbable assumptions may deserve less weight than an application already supported by existing institutional practices.
However, uncertainty about probability does not automatically make a potentially severe harm irrelevant.
Risk assessment should make assumptions explicit and distinguish evidence from conjecture.
Where specialized expertise is required, independent ethical, technical, or security advice may be appropriate.
What Makes a Dual-Use Risk Serious?
The seriousness of a dual-use concern depends on several interacting factors.
These may include the severity of potential harm, the plausibility of misuse, the accessibility of the capability, the scale of possible consequences, and the availability of safeguards.
For example, a research output enabling minor inconvenience under unusual conditions differs from one that could facilitate systematic discrimination affecting large populations.
However, risk cannot always be reduced to a single numerical score.
Likelihood estimates may be uncertain, and the consequences may depend on social and institutional conditions that are difficult to model.
Researchers should document their reasoning rather than present subjective risk ratings as objective measurements.
Does Dual-Use Risk Mean the Research Should Not Be Conducted?
No. The presence of potential misuse does not automatically make a study ethically unacceptable.
Many valuable research areas produce knowledge that can be used in competing ways.
Prohibiting every project with a conceivable harmful application could prevent socially important discoveries.
However, some risks may be sufficiently serious that additional safeguards, methodological changes, restricted access, or reconsideration of the research question are warranted.
The appropriate decision depends on the expected benefits, plausible harms, alternatives, and relevant governance requirements.
The question of when misuse risks should affect the choice of research question requires particular attention when safeguards cannot adequately reduce foreseeable harm.
Can Publication Increase Dual-Use Risk?
Potentially. Publishing research can make methods, data, software, and technical capabilities available to a wider audience.
Open scientific communication supports scrutiny, reproducibility, and cumulative knowledge.
However, some information may also make harmful applications easier to develop or implement.
Researchers should assess whether particular details create meaningful additional risks and whether those risks can be reduced without unnecessarily undermining scientific transparency.
For example, sharing an educational prediction model may require attention to whether the accompanying dataset exposes sensitive student information or enables inappropriate profiling.
The broader tension between scientific openness and responsible research should be evaluated according to the specific information involved.
Should Researchers Withhold Potentially Sensitive Methods?
Sometimes, but withholding should be justified rather than treated as the automatic response to dual-use concerns.
Possible measures include controlled access to sensitive datasets, limiting distribution of particular materials, or consulting appropriate oversight authorities before dissemination.
However, withholding methodological information may reduce reproducibility and prevent independent scrutiny.
Researchers should examine whether a less restrictive safeguard can address the relevant risk.
Any restriction should be consistent with applicable law, institutional requirements, funder conditions, and publication policies.
The detailed ethical considerations are discussed in withholding methodological details to prevent harm.
Who Is Responsible for Evaluating Dual-Use Risks?
Responsibility may be shared among researchers, institutions, ethics committees, funders, publishers, and specialized oversight bodies.
Researchers are often best positioned to explain the capabilities and limitations of their work.
However, they may not possess all the expertise needed to assess security, social, legal, or institutional consequences.
Interdisciplinary consultation can therefore be valuable, particularly when a project raises unfamiliar risks.
Formal oversight requirements depend on the research field and jurisdiction.
A study that falls within a specific dual-use regulation may require procedures that do not apply to an ordinary educational technology project.
Can Ethical Review Approval Eliminate Dual-Use Concerns?
No. Ethics approval may establish that a study satisfied particular review requirements at a specific point in time.
It does not guarantee that every future application of the research will be beneficial or that all possible misuse has been anticipated.
Some ethics reviews focus primarily on risks to participants during the study, while broader downstream consequences may require additional expertise.
Researchers should therefore remain attentive to material changes in capabilities, intended applications, and deployment contexts.
However, continuing responsibility should be proportionate to the researcher's actual role and reasonable ability to influence outcomes.
What Is the Difference Between Misuse Risk and Misconduct?
Dual-use risk concerns potential harmful applications of research. Research misconduct concerns violations of applicable standards governing research conduct.
A study may be conducted honestly and accurately while producing knowledge that could later be misused.
Conversely, fabricated research may constitute misconduct even if its findings have no meaningful dual-use implications.
Researchers should not accuse someone of misconduct merely because their research has potential harmful applications.
Ethical evaluation should distinguish the researcher's actions from the possible conduct of future users.
How Can Researchers Reduce Dual-Use Risks?
Risk reduction should address the specific pathways through which harmful use could occur.
Possible measures include restricting unnecessary access to sensitive information, designing systems with appropriate safeguards, testing for foreseeable misuse, and communicating limitations clearly.
Researchers may also consult affected communities or relevant specialists to identify risks overlooked by the original research team.
However, safeguards should be evaluated for feasibility and effectiveness rather than assumed to work because they are described in a protocol.
The aim is to support beneficial scientific work while reducing reasonably foreseeable opportunities for harm.
Watch Out
Do not assume that a beneficial research objective guarantees beneficial applications. Equally, do not classify research as unacceptable solely because someone can imagine a harmful use. Dual-use assessment requires a defensible account of the capability, plausible misuse pathways, potential consequences, and available safeguards.
