Transparency
How JPALM.AI Scores Work
The JPALM.AI Score is built from one thing only: experiences that real contributors — defendants, formerly incarcerated people, family members, and legal representatives — report about their actual interactions with an attorney. Every experience passes through moderation and verification before it counts, and this page explains exactly how the math works.
"The JPALM.AI Score is an experience-based representation score derived from qualifying contributor reports about their interactions with an attorney."
"The JPALM.AI Score does not measure credentials, prestige, win rate, disciplinary history, advertising, popularity, profile completeness, or JPALM.AI's personal opinion of an attorney."
What the score never measures
The score measures reported representation experiences only. It is never influenced by any of the following:
- —Credentials
- —Law school prestige
- —Years in practice
- —Advertising
- —Peer ratings
- —Win/loss record
- —Disciplinary history
- —Profile completeness
- —Evidence confidence
- —Identity confidence
- —Attorney participation on JPALM.AI
- —Whether the attorney responds to reviews
- —Number of profile sources
These items may appear elsewhere on a profile, but they never mathematically change the score. Attorneys who respond to reviews receive a separate engagement badge ("Attorney Responds to Reviews") — no score points are awarded or removed for responding.
The six public score categories
JPALM.AI allows contributors to describe attorney representation in their own words. JPALM.AI analyzes the narrative for information relevant to JPALM.AI's published experience categories — six public categories, each with a fixed weight that together total 100%:
| Category | Based on | Weight |
|---|---|---|
| Communication & Access | Communication, responsiveness, ability to ask questions, access to counsel — as described in the narrative | 25% |
| Understanding & Legal Explanation | Explanation of charges, plea options, sentencing exposure, and legal options — as described in the narrative | 20% |
| Preparation & Court Readiness | Attorney preparation for court, hearings, motions, and trial — as described in the narrative | 20% |
| Client Participation | Whether the contributor's account, witnesses, evidence, and defense strategy were discussed — as described in the narrative | 15% |
| Professional Treatment | Professionalism, respect, and expectation management — as described in the narrative | 10% |
| Overall Experience | The contributor's overall experience and satisfaction — as described in the narrative | 10% |
If a category is not addressed by the narrative, it is excluded and its weight is proportionally redistributed across the supported categories — missing information never lowers a score. At least 4 of the 6 categories must be supported before a narrative experience can affect the aggregate score; below that, the experience may still be published and is labeled "Insufficient information for a complete JPALM.AI Experience Score." Family communication is displayed on profiles as a supporting signal where reported, but it is not automatically scored, because family communication is not equally applicable to every representation.
Normalization and the rating scale
Contributors describe their experience in their own words. JPALM.AI organizes the narrative into the published experience categories. The contributor reviews and confirms that interpretation before it becomes eligible for scoring. Each narrative-supported JPALM.AI category receives a structured 1–5 experience rating based on the contributor-confirmed JPALM.AI analysis:
- ·1 = Strongly Negative / Very Poor
- ·2 = Negative / Poor
- ·3 = Mixed / Average
- ·4 = Positive / Good
- ·5 = Strongly Positive / Excellent
Each supported category is normalized to a 0–100 score using ((rating − 1) ÷ 4) × 100. The individual experience score is calculated from the weighted combination of the supported categories.
If a category is not sufficiently addressed in the contributor's narrative, it is marked Not Enough Information and is excluded from that individual experience calculation. Its weight is proportionally redistributed among the supported categories. Missing information is never treated as zero or as a negative experience.
Narrative Experience Analysis
Contributors may describe their experience in their own words — up to 10,000 characters, as a complaint, a compliment, a mixed experience, or a neutral account. JPALM.AI may organize information contained in that narrative into JPALM.AI's experience categories. You tell the story; JPALM.AI organizes the experience; your words remain your words.
- ·JPALM.AI only evaluates categories reasonably supported by the contributor's narrative. An unmentioned topic is treated as insufficient information — never as a negative rating.
- ·The contributor is given an opportunity to review and correct JPALM.AI's interpretation before the experience becomes eligible for public scoring.
- ·The contributor's original words and JPALM.AI's structured analysis are displayed separately on every public experience.
- ·The contributor-confirmed analysis — not an AI-generated legal conclusion — becomes the structured experience data used by the JPALM.AI scoring engine.
- ·For every supported category, JPALM.AI records a 1–5 rating, the exact narrative language that supports it, and its confidence; the rating is normalized to 0–100 using ((rating − 1) / 4) × 100.
- ·Score Coverage reflects how much of JPALM.AI's experience framework was addressed by the contributor's narrative. It is not an attorney-quality score.
- ·JPALM.AI organizes what a contributor reports. It does not independently establish that reported events occurred and does not determine whether an attorney committed malpractice, misconduct, ineffective assistance, or a constitutional violation.
JPALM.AI organizes the contributor's narrative into the JPALM.AI experience categories. Before an AI-derived experience can affect a public score, the contributor is given an opportunity to review and correct the analysis. The contributor-confirmed analysis becomes the structured scoring source, subject to JPALM.AI's verification and moderation requirements. The extraction methodology is versioned separately from the score as JPALM_NARRATIVE_EXTRACTION_V2 and is never changed silently.
Verification and weighting
Not every experience counts equally. Verification determines how much weight an experience carries — it is about the strength of the connection between the contributor, the case, and the attorney, never about whether the opinion is positive or negative:
"Verification determines how confidently JPALM.AI can connect an experience to the contributor, case, and attorney. Verification does not determine whether an opinion is positive or negative."
Institutional Submission is tracked as a source type, not a verification tier — it does not automatically make an experience more truthful. A contributor's incarceration status never changes their weight; weight reflects verification of the contributor-case-attorney relationship.
"Experiences reflect contributors' reported perceptions and experiences. Verification does not mean JPALM.AI independently established that every subjective statement contained in a review is objectively true."
Recency
Recent representations carry slightly more weight than older ones, so the score reflects current practice without erasing history:
- ·0–3 years ago: weight 1.00
- ·More than 3 through 6 years ago: weight 0.95
- ·More than 6 through 10 years ago: weight 0.90
- ·More than 10 years ago: weight 0.80
Old experiences are never deleted — historical experience remains part of the public record.
Minimum data thresholds
A score is only shown when there is enough qualifying data to support it:
- ·0–2 qualifying experiences: "Not Yet Rated" — JPALM.AI does not yet have enough qualifying experience data to calculate a public score.
- ·3–4 qualifying experiences: the score is shown with a visible "Limited Experience Data" label.
- ·5–9 qualifying experiences: the score is shown with the sample size.
- ·10+ qualifying experiences: the score is shown with a stronger sample-size indicator.
Experience Data Confidence
Experience Data Confidence is a separate supporting signal shown beside the score:
- ·3–4 qualifying experiences: Emerging
- ·5–9: Developing
- ·10–24: Established
- ·25+: High Data Confidence
Experience Data Confidence reflects the quantity and verification strength of experience data supporting the score. It is not a rating of attorney quality.
Family experiences and clustering
Family members may submit valuable experiences, but their observations can differ from firsthand representation. Every experience is tagged with a perspective (firsthand client, family, authorized representative, or other participant). For the same attorney, defendant, and case, one firsthand experience stays individually weighted, while multiple family experiences are capped so they can never outweigh it. This prevents several relatives reporting on the same representation from artificially dominating an attorney's score.
Defense Participation Index
The Defense Participation Index is a separate supporting metric that measures how involved contributors report being in their own defense — whether their version of events, witnesses, evidence, motions, plea offers, and strategy were discussed with them.
It is scaled 0–100 from the defense topics the contributor reported discussing: for narrative experiences, the topics JPALM.AI identified as clearly discussed (Yes = 1) or clearly not discussed (No = 0) — topics the narrative does not address are excluded entirely. It is shown with its qualifying-experience count.
The index reports contributor experiences only. It never states that an attorney violated the Sixth Amendment, and it never classifies an attorney as effective, ineffective, or constitutionally deficient.
Manipulation protections
JPALM.AI runs automated checks that flag submissions for admin review, including:
- ·Duplicate contributor identifiers
- ·Duplicate defendant/case combinations
- ·Repeated narratives
- ·Unusually coordinated submission activity
- ·Attorney employee, opposing counsel, or competitor suspicion
- ·Bot or spam activity
- ·Multiple relatives rating the same representation
- ·Suspicious rapid review activity
Flags trigger admin review — they never automatically suppress a review. Positive and negative experiences follow identical moderation and verification standards, and legitimate negative reviews are never suppressed merely because they are negative.
Metrics JPALM.AI keeps separate
JPALM.AI profiles show several distinct metrics. They are never averaged together:
- JPALM.AI Score= representation experiences
- Defense Participation Index= reported client participation
- Profile Completeness= how much information JPALM.AI has
- Evidence Confidence= how strongly factual profile information is sourced
- Identity Confidence= how certain JPALM.AI is that records belong to the correct attorney
- Experience Data Confidence= how much qualifying experience evidence supports the score
- Disciplinary / Public Record Information= separately sourced public-record information
Profile Health is an admin-only operational metric and never appears in public scoring.
Versioning and auditability
The scoring methodology is versioned. The current production methodology is JPALM_SCORE_V2: the contributor-confirmed JPALM.AI analysis (JPALM_NARRATIVE_EXTRACTION_V2) is the production source for individual experience scores, while the aggregate weighting mathematics are unchanged from V1. Experiences scored under the earlier V1 questionnaire methodology keep their existing records — nothing is retroactively rescored without a new calculation record. Every score calculation writes an immutable audit record listing exactly which experiences were included and excluded, their verification and recency weights, and the resulting category scores. The mathematics are never changed silently — any change becomes a new version (JPALM_SCORE_V3, …) with the methodology on this page updated to match.