How we research and test claims
Show the work. Record what we still do not know.
FJP should be judged by evidence, whether it can work under the law, whether it protects safety and fairness, and whether it creates new problems—not by how strongly anyone tells its origin story.
We separate three things: what an official source directly supports, what we infer from it, and what remains unknown. We keep limits next to claims, welcome evidence that challenges the proposal, and propose a small protected test before anyone considers broad adoption. Research terms explained →
Primary question
Can a structured protocol improve family-court decisions without creating a new layer of harm?
FJP proposes that courts could improve safety, fairness, accuracy, and lasting results by sorting cases more carefully, showing where evidence came from, requiring professionals to explain their reasoning, limiting assignments to the work actually needed, making cost visible, checking whether orders were carried out, and studying later outcomes. These are ideas to test—not established facts.
A successful pilot would improve several measures without making safety, legal fairness, unequal treatment, privacy, or access worse. A neutral or harmful result should change or end the part being tested.
Where the concept came from
The original working conversation is the idea source—not the evidence.
Version 0.1 reorganizes ideas developed in the referenced “Family Court Problem” conversation into a public model. That conversation is the master record of the concepts. It is not proof. Legal and research claims require independent sources.
The reorganization adds explicit safety and due-process gates, stronger privacy limits, anti-gaming rules, a distinction between ordinary tamper-evident records and blockchain, a correction protocol, and a pilot-ready measurement plan. These are presented as refinements for review, not settled policy.
Research process
Use the strongest available source and say what it cannot prove.
- Begin with original and official sources: laws, final court decisions, court reports, government data, and the original research study.
- Use news and advocacy to find leads: then trace important claims back to an official or original source when one exists.
- Label the kind of statement: verified fact, someone else’s attributed claim, our synthesis, our inference, a “what might have happened” test, or an unknown.
- Keep the boundary attached: identify the place, time period, people studied, measurement, and major limits.
- Look for evidence against the proposal: include neutral findings, other explanations, and risks that could make FJP fail.
The site uses that word for a conviction, guilty plea, formal disciplinary finding, official audit finding, or judicial determination—not for a disputed allegation, unusual result, anonymous account, or inference from a docket. Adjacent juvenile or child-protection matters are labeled so they cannot be mistaken for private domestic-relations cases.
Research for v0.1 was conducted through 14 September 2026. The source register links to the exact public item used. Citation availability and link persistence can change. Open the full source register →
Court records
Information can be legally public and still be harmful to republish.
Version 0.1 uses totals and trends from court data, published decisions, laws, official reports, and cases already discussed extensively in official public records. It does not collect or republish personal details from active family-case dockets. A short docket label is never enough to draw a conclusion about a family.
| Record type | Permitted use | Guardrail |
|---|---|---|
| Aggregate court statistics | Caseload, clearance, age, representation, hearing activity. | Keep the reporting definition, jurisdiction, and time period attached. |
| Published final judgments | Findings and procedural history stated by the deciding court. | Distinguish allegations, findings, holdings, dicta, and later posture. |
| Legislative accounts | Attributed statement of what a sponsor or body relied upon. | Do not convert into an adjudicated fact; identify missing case records. |
| Active family dockets | Only aggregate, de-identified research under lawful and ethical governance. | No targeting, republication, or inference about a family from sparse entries. |
Illinois Supreme Court Rule 8 distinguishes public, impounded, confidential, and sealed court records; access must follow controlling court rules and policy. S27
Case selection
Why these cases appear.
Cases were chosen to test different failure modes, not to produce a representative sample: order-implementation failure; delay and child voice; inability-to-pay procedure; appellate access; and widely cited child-safety legislative cases. The set intentionally includes both high-confidence judicial records and lower-completeness legislative accounts so the site can model different confidence levels.
Extreme cases are useful for hazard analysis but cannot estimate prevalence, typical performance, or comparative risk. Their emotional force must not substitute for base-rate evidence.
Privacy and data governance
Minimum necessary data, maximum accountability.
- Collect only information needed for a stated legal or research purpose.
- Keep names and identifying details separate from research data whenever possible.
- Limit who can see each item, protect stored and transmitted data, record access and changes, correct errors, and delete information when the lawful retention period ends.
- Never place sealed evidence, children’s accounts, addresses, therapy records, or intimate images on a public record system.
- Before any live test, identify what could go wrong, test for bias and disability access, and prepare a response to any security breach.
NIST’s Privacy Framework offers a general risk-management structure; NCSC’s National Open Court Data Standards provide a court-data vocabulary and governance reference. Neither endorses FJP. S23 S22
An ordinary secure database can record who changed what and when. That may be enough, with less privacy and management risk. A shared blockchain-style system should be considered only if several independent institutions genuinely need to verify the same history without relying on one record keeper.
Proposed pilot
Start narrow, compare fairly, stop if gates fail.
Design
Selected parts could be introduced in steps across similar court teams, allowing researchers to compare results while every participating site eventually receives the tested change. Researchers call these approaches a stepped-wedge or matched-site design. Independent evaluators should record the analysis plan in advance and watch for safety problems. The study must respect judicial authority, emergency procedures, and participants’ rights.
Begin with lower-risk parts
First test a visible evidence history, clear assignments for neutral professionals, written-reason templates, cost and time records, and checks that ordered safeguards actually happened. Software that predicts a person’s risk is outside v0.1.
Balanced measures
| Domain | Illustrative measures | Anti-gaming companion |
|---|---|---|
| Safety | Verified safety events; time to protective response; restrictions. | False-positive and false-negative review; independent case audit. |
| Process | Time to key events; pending-case age; continuances. | Reopening, postjudgment activity, completeness of evidence gates. |
| Fairness | Notice, opportunity to respond, comprehension, procedural-fairness survey. | Representation and language/disability stratification. |
| Cost | Party legal/professional spend, lost wages, travel, child care. | Missing necessary evidence, uncompensated work, unequal access. |
| Child/family | School/home stability; child-appropriate wellbeing; parenting capacity. | Baseline risk, nonresponse, privacy burden, reporter source. |
| Durability | Compliance, modification, relitigation, user-reported workability. | Necessary modification, unsafe silence, settlement under coercion. |
Reasons to stop
Pause a part of the pilot if safety appears to worsen, people cannot participate meaningfully, private information is seriously exposed, one group is harmed more than another, or the added work becomes unmanageable. Publish harmful events and results showing no improvement—not only success stories.
Independent review
Participation is not endorsement.
Before journal submission or policy advocacy, v0.1 should be reviewed by people with intentionally different responsibilities: family-law attorneys; judges with appropriate independence; child representatives; evaluators; domestic-violence and coercive-control experts; civil-liberties and due-process counsel; court administrators; economists; statisticians; child-development researchers; privacy/security technologists; disability and language-access experts; and people with diverse lived experience.
Reviewers receive specialty-specific questions and an objection ledger. Each material objection receives an ID (for example, FJP-023), severity, source, response, revision status, and unresolved-risk statement. Reviewer names are disclosed only with consent. Participation never implies endorsement.
Limitations
Known weaknesses in version 0.1.
- The framework was synthesized from a working dialogue and has not yet undergone formal interdisciplinary peer review.
- The source review is targeted rather than a systematic review with dual screening and risk-of-bias scoring.
- U.S. family-court data are fragmented; cross-jurisdiction comparison is limited by different definitions and privacy rules.
- Extreme cases are selected for failure-mode analysis and are not representative.
- Some notorious-case facts are available only through legislative accounts; underlying family records were not independently reviewed.
- The protocol may add burden, formalize bias, produce false reassurance, or shift power toward administrators and vendors.
- Outcome measurement can intrude on family privacy and can be gamed.
- Legal feasibility depends on state law, court rule, funding, labor, local technology, and constitutional constraints.
Versioning and corrections
Make disagreement traceable.
| Version | Date | Status | Material changes |
|---|---|---|---|
| 0.1 | 14 Sept. 2026 | Public working draft | Initial integrated protocol, source register, nine case stress tests, pilot outline, privacy limits, and review commons prototype. |
Corrections should identify the page, exact claim, proposed change, source, and whether the issue is factual, interpretive, legal, methodological, or ethical. Accepted corrections remain visible in a public change log rather than silently replacing the record.