How we score you — and what it means.

SA CoParents is an educational and preparation tool. Whenever we show you a score, this page tells you exactly where that number came from, what it measures, and where it stops.

1. Mediation Readiness (0–100)

What it measures. Your preparation across four dimensions:

• Child-focused focus — the specificity + child-directed framing of the goals you named.
• Priorities clarity — whether you've ranked and rationalised your top mediation topics.
• Communication readiness — your self-assessed patterns around requests, listening, and de-escalation.
• Emotional regulation — the presence of concrete plans for reacting to triggers.

How it's calculated. Deterministic rule-based scoring of the fields you fill in: word-count thresholds, presence of required components, penalty for one-word answers. The exact rules live in backend/summary_scoring.py. There is no ML model in this score.

What it does NOT measure. The score cannot see emotional labor, real-world power dynamics, or coercion. A high score means you've written a lot of child-focused content — it does not mean mediation will be safe or successful for your specific family.

2. Family Violence and Communication Safety Screen

Source instruments. The Section-1 immediate-safety screen is adapted from the widely-used Danger Assessment (Campbell, 1986) and the HITS screening tool (Sherin et al., 1998), with wording tailored for co-parenting contexts. Section 2 (communication readiness) is original but reviewed against the WEB Scale and the ODARA framework.

Output. A four-level communication pathway — No-direct contact · Structured written · Assisted · Cooperative — with an explanation of why. Any single "yes" in the Section-1 immediate-safety questions triggers the No-direct-contact pathway.

Important limits. This is a screening tool, not a validated clinical assessment. It is not the Danger Assessment, MASIC-S, or HITS. The Danger Assessment estimates lethal / near-lethal risk in abusive intimate relationships and its scoring is tied to specialized training; HITS was validated in family-practice settings, not in divorce, custody, or mediation populations. This screen has not been independently validated in custody or mediation populations, and its scoring is heuristic. It should not be used in court proceedings without accompanying professional evaluation. Texas crisis hotlines are surfaced in every result regardless of level, and the "pathway" and "Texas resources" sections of the PDF cannot be hidden by the user when sharing.

What this screen decides — and what it does not. The screen recommends a communication structure for the parent to consider. It does not decide mediation suitability, visitation, custody, protective orders, or any other legal or clinical outcome. Decisions of that weight require a qualified human assessment (family-law attorney, licensed clinician or forensic evaluator, DV advocate). For Level 1 & 2 results the app suppresses AI communication coaching, because coaching a survivor to communicate "better" with a coercive co-parent shifts responsibility onto the person experiencing harm.

3. AI Coach responses

The interactive coach uses Anthropic Claude 4.5 via the Emergent LLM key. Prompts are constrained by a system instruction that forbids diagnosis, dosage advice, legal advice, or crisis triage. When triggered by risk keywords, the coach immediately surfaces local hotlines and offers to switch to the Safety Screen. AI responses are not stored on Anthropic's servers for training. See our Privacy Policy for details.

5. "Your next three steps" recommendation engine

Deterministic, not AI. Given your latest scores across all assessments, a fixed decision table picks up to three highest-leverage next actions. Rule source: backend/next_steps_routes.py. This means the same inputs always produce the same recommendations — an important transparency guarantee.

Where we lack empirical validation

SA CoParents is a new tool. The Mediation Readiness, Special Needs 7-domain, and Communication Readiness scores are not yet externally validated by a peer-reviewed study. We are actively collecting anonymized outcome signals (with opt-in consent) to publish reliability + concurrent-validity work in future.

In the meantime, use every score as a conversation starter with your mediator or therapist — not as the last word.

Change log & scoring history

Every meaningful change to a scoring rule is recorded in the public repository at github.com/mattsossi-cmyk/sacoparents-app. If a rule change would materially shift your historical scores, we recompute in a way that keeps your longitudinal chart honest: we snapshot the version tag alongside each stored score, and only recompute forward.

Questions or peer review requests

SA CoParents · 16607 Blanco #703, San Antonio, Texas 78232