By Eric Krebs
This week, on “Proof over Precedent,” Shannon Sewards, current Administrative Director of the Dartmouth Health IRB and former IRB administrator at Harvard University, joined host Jim Greiner for the third in a series of discussions demystifying Institutional Review Boards.
This installment explores one deceptively simple but highly consequential question: who counts as a “human subject” under the federal Common Rule, and why does that matter for legal field experiments?
Though it might seem like a purely academic regulatory puzzle, Greiner and Sewards explain how the way researchers, IRBs, and courts answer that question will shape the future of evidence-based reform in access to justice and beyond.
The Common Rule—codified at 45 C.F.R. Part 46—sets out the baseline protections for research involving human beings. 45 C.F.R § 46.102(e) defines “human subject” as such:
Human subject means a living individual about whom an investigator (whether professional or student) conducting research:
(i) Obtains information or biospecimens through intervention or interaction with the individual, and uses, studies, or analyzes the
information or biospecimens; or
(ii) Obtains, uses, studies, analyzes, or generates identifiable private information or identifiable biospecimens.
That legalistic phrasing masks a huge practical consequence: if someone counts as a human subject, researchers must obtain their consent or secure a waiver from an IRB before proceeding. If not, those protections don’t apply. For biomedical studies, the boundaries can be relatively clear. For legal field experiments—say, randomized trials in eviction cases or bail hearings—the lines blur quickly.
An Illustration: When the Tenant Is a Subject but the Landlord May Not Be
Greiner begins with a concrete scenario. Suppose researchers randomize the provision of legal assistance to tenants in eviction cases. Clearly, the tenant receiving the intervention is a human subject: the study interacts with them, collects data, and measures outcomes.
But what about the landlord on the other side of the case? Here, Sewards emphasizes the importance of boundaries. Unless the landlord is directly interacted with, or private data about them is collected, they are not a human subject under the regulations. Even if the landlord’s experience is indirectly shaped by the study—for instance, tenants might lose fewer eviction cases—Sewards explains that IRBs should avoid “mission creep,” the tendency to expand oversight so far that everyone possibly affected in any way by a study is a subject. As Sewards explains, IRBs must apply the definitions set forth by the Common Rule, not simply what “feels right.” Else, “the entirety of the world might be subject to higher oversight.”
As Greiner explains, this distinction matters. If landlords were deemed human subjects, they could effectively veto randomized trials in access-to-justice settings, an outcome that could cripple empirical research in the legal system.
A second scenario concerns risk assessment instruments in criminal bail decisions. Researchers may randomize whether judges receive these predictive tools. Here, the judge is clearly a subject, since their environment is directly manipulated.
But what about the defendants appearing before those judges? Greiner and Sewards probe the tension. On one hand, the tool changes the context in which defendants’ cases are decided—sometimes resulting in pretrial release, sometimes in bail being imposed. On the other hand, there is no direct contact with the defendants, and their behavior is not altered by the intervention. Sewards stresses that IRBs must look back to the text of the regulation: is there intervention, interaction, or identifiable private data? If the answer is no, then defendants may not qualify as human subjects, even though the outcomes of their cases are affected.
The conversation here highlights a broader irony. Governments can adopt or reject risk assessment instruments wholesale without triggering federal protections. Yet randomized trials designed to test whether these tools actually help or harm trigger scrutiny precisely because they carry the label “research.” As Sewards puts it, that tension is one of the great challenges of extending evidence-based methods into the legal arena.
One recurring theme is the ripple effect problem. If interventions affect not just direct participants but also adversaries, judges, or bystanders—should they all count as subjects? Sewards cautions against such an expansion. If downstream or speculative effects were enough, “everyone would be a human subject,” and research would grind to a halt. The key is to draw lines based on the regulations: interaction, intervention, or identifiable private information.
Greiner then poses a hypothetical question from legal scholar Christopher Robertson. Imagine a law school offering a preparatory program only to first-generation students, randomized to test whether it improves performance. Those students are clearly subjects. But are the non-first-generation students also subjects, since grades are curved and any benefit for one group could disadvantage the other?
Sewards’s answer: no. The mere fact that outcomes are competitive doesn’t transform everyone into a subject. Moreover, Greiner and Sewards identify the issue of selecting subjects based on the ex-post outcome of the study—whether the program actually helps some students therefore harms others—to identify who is a subject ex-ante, “you may be going down the wrong path.”
This approach mirrors debates in law about statutory interpretation. Should judges apply the law as written, or expand it to serve perceived justice and broader-stroke goals? Greiner notes that IRBs in many ways function like quasi-courts, wrestling with similar dilemmas. Sewards also acknowledges the parallel: both endeavors require balancing text with purpose and consequence. For IRBs, striking the right balance is critical to ensuring that productive, humane field research can continue—and, hopefully, expand—in the field.
Listen to the full Proof Over Precedent podcast episode on this topic.