Clean Slate Eligibility Explorer — Vermont

How eligibility is decided

The assumptions behind these rules

Each rule here is a policy choice we made to demonstrate feasibility, not a recommendation for what Vermont's policy should be. We label them as assumptions on purpose.

• Which grades are eligible. We treat non-convictions, petty misdemeanors, and misdemeanors as eligible by grade, and exclude all felonies. This is a deliberately narrow, misdemeanor-and-below line so the demonstration stays clear; where to actually draw it is a policy decision.
• How waiting periods are counted. There are several ways to define when a record's clock starts and how long it runs. We model two technically feasible approaches — counting from the disposition, and counting from sentence completion — so the choice can be compared side by side. We use four years for a misdemeanor and two for a petty misdemeanor as illustrative lengths; non-convictions are eligible immediately. We are not asserting any of these numbers is correct.
• Sentence completion is estimated. Neither the courts nor the repository reliably stores the date a sentence was actually completed, so the "sentence completion + time" model stands in the statutory maximum sentence for the real one. If an agency could read shorter actual sentences, people would become eligible sooner than this model shows.
• Disqualifying offenses. Because states sometimes carve violent offenses out of Clean Slate, we assume a conviction can be disqualified by its offense. We use an illustrative subset of Vermont's "listed crimes" (13 V.S.A. § 5301(7)) at misdemeanor grade — simple assault, reckless endangerment, threatening disorderly conduct, and domestic assault. Whether to disqualify these at all, and whether this is the right list, are policy questions for Vermont to confirm.

Understanding eligibility criteria vs. data points

Clean Slate works by reading records that already exist in electronic form. In Vermont, two different agencies could be the one to start the process, and they keep slightly different kinds of records. A court record in the Judiciary's Odyssey case-management system is the court's own account of a case; because the court is the source of dispositions, its disposition information is complete. A repository record at the Vermont Crime Information Center (VCIC) is the rap-sheet view, built from fingerprint-based arrests and the dispositions reported back to it. One important difference: the repository also holds arrests that never became a court case, while the court does not.

Each eligibility rule reads one of these fields. The offense-grade rule reads offense severity. Non-convictions are read straight from the disposition. The disqualified-offense check (convictions only) reads the offense together with the disposition. The waiting period reads the disposition date. The hanging-arrest rule reads the arrest date, which only the repository holds.

Two honest limits show up in this table. First, sentence completion is not reliably recorded in either system, so the sentence-based waiting-period model estimates it from the statutory maximum rather than reading a real end date. Second, hanging arrests exist only in the repository — the court never sees an arrest that was not charged — which is why the three hanging-arrest cases here appear only under VCIC. On the ten cases both agencies hold, they reach identical answers, which is the cross-check this POC verifies.

Showing that both a court system and a repository can run the same policy matters because either one could be the initiating agency. If the approach works whichever agency starts, the policy does not depend on a single system, and Vermont keeps flexibility in how it implements.

This is the core point: eligible records can only be identified by software when the eligibility criteria map to data points the agency actually holds in usable electronic form. A rule is only feasible if there is a field to read it from.

A look under the hood

The snippet below is the engine's actual decision logic, excerpted and lightly condensed from engine_template.py — the same engine that produced the cards above. Each charge passes through these checks in order, and the engine records a plain-language reason for whatever it decides.

• Non-conviction is eligible immediately: if the disposition is a non-conviction (dismissed, not guilty, and the like), the code returns eligible right away, with no waiting period.
• Offense grade must be on the eligible list: for a conviction, the code first checks that the grade is in the eligible set, so excluded grades stop here as not eligible.
• Optional disqualified-offense check for convictions: when a scenario turns it on, the code compares the conviction's offense against the disqualified list and marks it not eligible if it matches.
• Waiting-period check against the clear date: the code computes the charge's "clear date" from its disposition date and the policy's waiting period, then compares that date to the screening date, returning eligible only once it has arrived.
• A case is eligible only if all its charges are: after judging every charge, the case-level code marks the whole case not eligible if any single charge is not eligible.

Highlighted lines (1)–(5) are the key rules described above. Full source: engine_template.py.