by Claude Opus 5.5
What kind of judgement is a decision not to charge: an evidential prediction, a normative assessment, or a resource allocation? Which should it be?
The district attorney’s original decision not to charge in the case behind these seminars was explained in evidential terms: the complainant’s signed statement, as the DA read it, described the conduct as voluntary and consensual, and that made a prosecution untenable. But decisions not to charge are rarely only one kind of judgement. They combine predictions about what a court would do, assessments of what ought to happen, and choices about where limited resources should go. Which of these the decision is, and which it should be, matters for how we evaluate it and for what accountability it should face.
Three models
On the predictive model, the prosecutor asks: if this case went to trial, how likely is it that a jury would convict? The Crown Prosecution Service’s “realistic prospect of conviction” test is explicitly framed this way, as an objective assessment of whether a court properly directed would be more likely than not to convict. American practice varies, but many prosecutors describe their charging decisions similarly.
On the normative model, the prosecutor asks: do I believe this person committed the crime, and is prosecution the right response? The American Bar Association’s standards for prosecutors, for instance, suggest that a prosecutor should not file charges without a belief that the evidence is sufficient to support conviction, and should consider whether prosecution serves the interests of justice. The CPS’s second stage, the public interest test, also invites normative judgement about culpability, harm and proportionality.
On the allocative model, the prosecutor asks: given finite resources, is this case a priority? This is rarely stated openly in individual cases, but it shapes practice. Offices with heavy caseloads decline cases that would require extensive investigation or are unlikely to succeed.
The problem with pure prediction
Prediction seems the most neutral and defensible model. Prosecutors should not put people through trials they will almost certainly win. But in sexual offence cases, pure prediction has a serious flaw. If juries are influenced by stereotypes about how victims behave, for instance that a woman who took drugs voluntarily or did not resist was consenting, then predicting jury behaviour means incorporating those stereotypes into charging decisions. The prosecutor becomes a filter that reproduces jury prejudice before any jury sits.
This problem has been debated extensively in England and Wales. CPS guidance at one time endorsed a “merits-based approach”, instructing prosecutors to predict the verdict of an idealised, unprejudiced jury rather than an actual one. Its later removal from training materials was challenged in litigation and criticised as contributing to the collapse in rape prosecutions, and subsequent reform programmes have tried to restore a merits-based focus. The debate shows that “prediction” is not neutral: one must decide whose verdict is being predicted.
The problem with pure assessment
If the prosecutor instead relies on her own view of guilt, the risk runs the other way. A prosecutor’s private conviction is not tested by any adversarial process, and it may itself be distorted by bias, pressure or incomplete evidence. A normative model also threatens to blur the roles of prosecutor and judge. The prosecutor’s task is not to decide guilt but to decide whether the question of guilt should be put to a court.
The problem with allocation
Resource constraints are real, but they are a poor justification for declining serious cases. If allocation drives decisions about sexual offences, cases that require painstaking investigation, such as those involving intoxication, multiple suspects or digital evidence, will systematically be dropped. That disadvantages precisely the complainants whose cases are hardest, and it does so invisibly, because allocative reasons are seldom stated.
What the decision in this case reveals
The available information suggests that the original decision was presented as predictive and evidential, resting on the signed statement. But several features indicate that other considerations may have been in play. If the accused were not interviewed, and if the prosecutor did not review all available material, the decision may have reflected investigative limits, an allocative constraint, rather than a full evidential assessment. And the reading of “voluntary” and “consensual” in the statement involved an interpretive judgement about what those words meant in context, which is partly normative.
What it should be
The best account is a structured combination. The core of a charging decision should be evidential, but predictive in the merits-based sense: whether a fair-minded tribunal, properly directed and free of stereotypes, could be satisfied of guilt on the available evidence. This keeps the decision anchored in evidence and respects the prosecutor’s role, while refusing to incorporate prejudice.
A normative element belongs in a second stage, as a public interest assessment of whether prosecution is warranted given the seriousness of the alleged offence, its effects and the circumstances of the parties. For serious sexual offences, this stage will usually favour prosecution where the evidential test is met.
Resource allocation should operate only at the level of investigation and policy, not as an unstated reason for declining serious cases. Where resource constraints affect a decision, they should be disclosed, so that they can be challenged and remedied.
Finally, the decision should be accompanied by reasons that make clear which kind of judgement was made. A complainant told that there is “insufficient evidence” deserves to know whether that means the evidence does not support guilt, that a jury would not believe her, or that the case was not investigated fully.
Conclusion
A decision not to charge is, in practice, a mixture of prediction, assessment and allocation. It should be primarily a merits-based evidential judgement, with a separate public interest stage, and with resource constraints confined to investigation and openly disclosed. That structure would not guarantee correct decisions, but it would make them intelligible and contestable, which is what this case suggests was missing.