Over the past decade, criminal and juvenile courts around the world have begun delegating a task once entrusted to human judgment—deciding who is credible—to risk assessment tools. These operate according to a fixed formula to score a young person's likelihood of reoffending. Conventionally, caseworkers tallied the factors by hand, but a growing number of these are automated. Troublingly, the score, however it is produced, increasingly stands in for a judgment a person used to make. Compounding the issues created by risk assessment tools is the turn toward video, as opposed to in-person, hearings. Proponents describe both developments as ways to increase efficiency and neutrality. Both, in practice, influence whether a child is believed, detained, or released, and they do so in ways that a growing body of research suggests harm these young people.
The Convention on the Rights of the Child, the most widely ratified human rights treaty in the world, guarantees children capable of forming views the right to express them in judicial proceedings. The United Nations Committee on the Rights of the Child has affirmed that this is a binding obligation on states, not a courtesy, and that it applies squarely to juvenile justice. When the design of a proceeding adopted by the courts undercuts a child’s credibility before they speak, that obligation is eroded.
What risk scores actually measure
In the United States, and increasingly elsewhere, risk assessment tools assign young people a risk score based on factors that can include prior arrests, family circumstances, and neighborhood. However, it is not always clear what those inputs capture. As one widely used guide to juvenile probation practice acknowledges, scoring a child higher based on prior arrests or court referrals perpetuates the racial and economic disparities responsible for those contacts in the first place. A federal literature review is more direct: a high risk score often reflects a young person’s likelihood to be detected and rearrested, not their likelihood to offend.
The tool, in other words, does not measure danger. It measures the footprint the system has already left on a child's life and recasts this as evidence of future threat. This is the deeper problem with framing credibility as something a formula can compute. These systems do not assess credibility so much as they redefine it according to whatever the available data can capture, ignoring important context, such as unstable housing, well-founded distrust of authority, or the difference between a bad month and a bad character.
How screens distort court hearings
Remote court proceedings compound the problem. The Brennan Center for Justice has synthesized studies showing that judges routinely perceive people who appear in court by video as less credible than those who appear in person. In one study of more than 645,000 felony bond decisions in Cook County, Illinois, average bail amounts increased by about 51% after hearings shifted to closed-circuit video; cases that remained in person saw a much smaller and statistically insignificant change. Perhaps most tellingly, immigration judges are significantly more likely to order deportation when hearings are held by video rather than in person. Researchers attribute this disparity in part to how screens distort facial expressions, body language, and other cues judges use to decide whether someone is credible.
For a young person, the stakes of this distortion are particularly high. Adolescents are more likely to display exactly the behaviors—averted eyes, hesitation, visible nervousness—that observers read as evasion, and a screen amplifies those cues while stripping away the context that would explain them. The video medium thus shapes the judgment of their credibility before the substance of the case is even broached.
A local practice becomes a global default
Risk assessment tools and remote hearings are spreading through criminal justice systems worldwide. Often private vendors promote these to under-resourced courts and police forces as a means of modernization. The Netherlands, for instance, developed a risk assessment tool that lets police automatically calculate a young person's risk of offending from data already in the system. Meanwhile, the international framework meant to protect children's right to expression has not adapted to technologies capable of filtering, scoring, and flattening their voices before anyone weighs them. The existing human rights architecture assumes a child who can speak will be heard directly. It does not account for proceedings in which an algorithm and an interface have prestructured what “being heard” means.
This gap is bridgeable, but closing it means listening to the growing coalition of criminologists and legal scholars who insist that judicial systems must assess algorithmic legal tools through the lens of rights rather than procurement. Courts can decline to let a risk score stand in for a hearing and can preserve in-person proceedings for children whenever a credibility judgment is central to the case. Any algorithmic tool bearing on a young person's liberty should be independently audited and open to challenge by the child's representatives. And, as a minimal baseline, a system should have to demonstrate that it does not discount a child's testimony prior to use.
The window to act is narrowing. These tools are still being adopted, still being incorporated into policies and budgets, and still open to revision. Once they insinuate themselves into the default architecture of juvenile justice, the question of whether a child is entitled to belief will already have been answered. It is a question the human rights community should insist on addressing openly and deliberately while the answer is still ours to shape.