Correctly Drawn LawMassachusetts legislative drafting & interpretation

Bias Spectrum Audit · an engagement

Bias you can see is bias you can fix.

Where does this law sit? Modern statutes rarely discriminate by naming anyone. They do it through thresholds, discretion, maps, and carve-outs that look neutral one clause at a time. This audit makes that structure visible by contrast — and hands the drafter the language that closes the gap.

Hand me one section of the General Laws, or one bill. You get back a two-page briefing that shows, for each drafting decision in it, how close today's choice sits to the worst version a skilled drafter could write today — and how far from the most protective one.

The comparison is never with the openly discriminatory laws of the past. Every modern statute looks humane beside a 1950 covenant, and an audit that says "look how far we've come" has manufactured complacency. The comparison that matters is with the rigged version: the same statute, every drafting decision pushed toward exclusion, no protected class ever named, every clause defensible on its face. When part of today's law is built the way the rigged version would be built, that resemblance is the finding.

The audit examines eight drafting decisions that recur in law after law — who decides, who is left out entirely, where the numeric line falls, whether "I didn't know" counts, whether the penalty is automatic or missing, whether a protection comes with the piece that makes it real, who does the paperwork — and one more that the text never states: a duty with no one assigned to enforce it, so that the real decisions are made off the books.

Five audits, published

The same method on five real texts.

A split statute, an anti-exclusionary reform with a weak enforcement seam, a quiet statute whose problem is silence, a home rule petition that asked for more than its sponsor needed, and a sub-minimum wage that wears the language of protection. Read the dots first.

G.L. c. 164, § 34B · statute

Double poles

near the rigged end, by silence
  1. Automatic penalty, judgment call — or nothing? (L6)
  2. A rule with no referee (L1b)
  3. Half a protection (L7)
  4. Measuring by map (L2)

Today's law sits near the rigged end — not because anyone wrote a harsh rule on purpose, but because no consequence was ever written in, and a duty with no referee is most of what the rigged version would have had to draft deliberately.

Read the briefing G.L. c. 94C, § 32J · statute

Drug-free school zones

split
  1. Measuring by map (L2)
  2. Does "I didn't know" count? — the zone (L5a)
  3. Automatic penalty, judgment call — or nothing? (L6)
  4. Does "I didn't know" count? — the trigger (L5b)
  5. Who decides? (L1)

Today's § 32J sits split: what it takes to trigger the add-on sentence has moved well toward the most protective version (since 2018 it reaches only aggravated conduct, not mere presence near a school), while the map-based zone, the no-knowledge clause, and the mandatory consecutive sentence still sit at or near the rigged end — a statute whose who-it-catches was reformed but whose how-it-punishes was not.

Read the briefing G.L. c. 54, § 24 · home rule petition

Fall River polling places

near the rigged end; over-scoped
  1. Who decides? (L1)
  2. Measuring by map (L2)
  3. Where the line is drawn (L3)
  4. Who's left out entirely (L4)
  5. Automatic penalty, judgment call — or nothing? (L6)
  6. Half a protection (L7)
  7. Who does the paperwork? (L8)

The bill as drafted sits hard toward the rigged end on the four drafting decisions that decide who can actually get to the polls, its impact report is a paper protection that earns no credit toward the protective end, and it grants far more authority than the city asked for — narrowing it to the stated purpose is the largest single fix available.

Read the briefing G.L. c. 151, § 2A · statute

Farmworker sub-minimum wage

near the rigged end
  1. Who's left out entirely (L4)
  2. Half a protection (L7)
  3. Where the line is drawn (L3)
  4. Who does the paperwork? (L8)

Today's law sits near the rigged end on every decision that carries weight: it is a freestanding sub-minimum tier for farm work (L4) whose one protection is a wage floor frozen at roughly half the general minimum and cut off from overtime, indexing, and every raise the general minimum gets (L7), hollowed further by exits for family labor and minors (L3) and a remedy — the contract is simply void — that is the least enforceable form there is (L8).

Read the briefing G.L. c. 40A, § 3A · statute

MBTA Communities Act

split
  1. Measuring by map (L2)
  2. Who's left out entirely (L4)
  3. Who decides? (L1)
  4. Where the line is drawn (L3)
  5. Automatic penalty, judgment call — or nothing? (L6)
  6. Who does the paperwork? (L8)
  7. An affordability requirement (none)

Today's law is strong on what it requires — near the most protective end on substance — but sits near the rigged end on whether it bites: the text names no enforcer and states only a penalty the wealthiest towns can shrug off.

Read the briefing

What you receive

Two parts. The first is the product.

A legislator reads Part I. Counsel and staff who want to check the work read Part II.

Part I — the briefing

Two pages or less, no methodology. Where today's law sits: the chart, one row per drafting decision. What we found: the trouble spots resting on the quoted statutory words, what's missing, any correction to how the text has been described, any paper protection that could backfire. Options for the drafter: provisions worth a second look and protective language to weigh — with, for each, whether it works by itself or only if an agency acts, funds, and files correctly.

Part II — the evidence

The whole-chapter scan (every section, with a disposition — an unread section is a visible blank, never a silent one). The drafting decisions with the statutory words that make them. The same law written four ways, kept parallel so it can be compared clause by clause — including two deliberately harmful versions, each under a warning label, each paired with how to spot the same pattern in a live bill. Words versus effect. Three ratings for each finding: fix or patch, how strong the proof is, and how urgently the gap in the evidence needs to be closed. The validation record.

Version history

Where a real lineage exists — a predecessor bill, an executive-branch proposal, a competing draft — each is scored on the same scale, and the briefing says in one sentence a legislator can repeat whether the bill is trending protective, regressive, or mixed. No lineage is ever invented for consistency's sake.

How it is done

Built as a Claude skill. Run by a drafter.

The method is written down as a structured skill — eight drafting decisions, four versions, the licensing rule that says a trouble spot must rest on a quoted passage and a stated evidence rating, and a governor that stops a quiet statute from being dressed in alarming language. Claude runs it. I run Claude, and I read every page before it leaves.

  1. The operative text

    The section or bill is pulled from the authoritative source and checked against what you sent. Three audits in this method's history rested on stale or mis-cited text; the check is not optional. Every section of the surrounding chapter gets a one-line disposition.

  2. The drafting decisions

    Each of the eight is located in the text with the words that make it, and the silences are recorded as decisions too. If a sponsor's purpose document exists, the bill's scope is measured against it — an act that grants standing authority to do what was asked for once has its bias case in the excess.

  3. The four versions

    The same law written at four points on the scale, subject matter held constant. The most protective version is stress-tested for durability: the most aggressive equity mechanism is often the one a court strikes, and a struck protection leaves people worse off than the status quo.

  4. Placement and rating

    Every dot on the chart is licensed by a quote and an evidence rating or it does not go on the chart. The rhetoric is checked against the ratings before anything is written.

  5. Validation

    On any audit bound for a legislator, three checks are blocking: a second, blind read of the dots in a fresh session; a mechanical lint of every placement; and a plain-language sweep of every sentence a reader will see.

  6. My read

    I read the briefing as the client will. This is the step that cannot be delegated, and the reason this is a service rather than a paste box.

What it is not

An analytic lens, not a determination of unlawfulness.

Whether a disparate impact is illegal is a legal conclusion for counsel, made on a full record. The audit surfaces structure and evidence; it does not prosecute motives, and its first page says so. It also does not sort out which decisions were deliberate, which were copied from an old form, and which were never thought through — the consequences are the same either way, and the fix is usually easier when it was not deliberate.

The two deliberately harmful drafts in Part II are teaching tools, labelled as such on every page. They exist so a reviewer can recognize the same structure in the next bill that crosses the desk. They are never proposals, and the terms of engagement cover where they may go.

Request an audit. Tell me the section or bill, the stage it is at, whether a purpose document exists, and who will read the result — that last answer decides which validation checks are blocking. Send the text.

Request an audit Read the five published audits first

Correspondence does not create an attorney-client relationship. Nothing on this page is legal advice.