Living Requirement Documents: Beyond Surveys and Polls

The Problem with Snapshots

A survey is a photograph. It captures opinion at one moment in time, from one sample of people, about one framing of the question. By the time results are published, the community has moved on. New information has emerged. People's views have evolved.

Worse, a survey's questions are designed by the researcher, not the community. The framing constrains the answers. "Do you support or oppose the development?" doesn't allow for "I support it if the affordable housing percentage is increased" or "I need more information about the traffic study."

The Living Requirement Document is not a photograph. It's a living map — re-synthesized as input arrives, version-controlled, and structured to capture the full dimensionality of community perspective.

Anatomy of an LRD

A Living Requirement Document has four sections, each serving a different purpose for decision-makers:

Section 1: Strong Consensus (70%+ Agreement)

These are the items where participants' underlying concerns converge. Not "70% voted yes" — but "at least 70% of participants expressed this concern or need in their interview." The share is counted from each participant's extracted summary, not estimated by the model.

Hypothetical example:

Business access on Main Street must be preserved or improved

Raised by 143 of 188 participants (76%)

This tells a decision-maker: whatever you do, this concern is widely shared among the people who took part.

Section 2: Conditional Consensus

These are items where agreement exists but is contingent on specific conditions being met. This is the most valuable section for decision-makers because it reveals how to unlock support. It counts how many participants support something today versus how many would support it with a named condition.

Hypothetical example:

Bike lane is acceptable IF alternative parking is provided within 2 blocks

Support today: 41% | Support with condition: 63% | Delta: +22 points

A poll would have shown 41% support. The LRD shows majority support may be available — if you address the parking condition. That's a different decision landscape.

Section 3: Genuine Divergence

These are real value conflicts between sizeable groups that cannot be resolved through better communication. They require democratic resolution — voting, compromise, or executive decision.

Hypothetical example:

Funding source: New dedicated tax (44%) vs. Reallocate existing budget (41%) vs. Undecided (15%)

Note: This is a genuine value divergence requiring democratic resolution.

The LRD is honest about what synthesis can and cannot solve. Not every disagreement is a communication failure.

Section 4: Confidence Gaps

This section is designed to flag where the data is weak — which communities are underrepresented and where targeted outreach is needed before the LRD should be treated as representative. It works when participants choose to share a few coarse details (age band, housing tenure, area) and the deploying organization supplies population figures for those groups. Where both exist, the LRD also shows the stance split reweighted to the population. Groups of fewer than five people are never shown.

Example:

⚠ Renters under 25: 8% of population, 2% of sample

Recommendation: Targeted outreach through university channels

In the meantime, the LRD shows a margin of error — but participation is self-selected, and that margin assumes random sampling, so treat it as a lower bound on the real uncertainty. Being explicit about data quality is something surveys rarely do.

Version Control for Community Intent

LRDs are versioned like software. As new interviews come in, the synthesis is re-run, producing a new version. Decision-makers can see how consensus has evolved over time and what new concerns have emerged. Each version publishes a Merkle root computed from the perspectives it used, and every participant gets a receipt they can check in the browser to confirm their perspective was included. (That proves what a published version was computed from; on its own it can't prove nothing was left out before publication — which is why receipts are in participants' hands.)

This means an LRD is never "done." It's a living document that can become more nuanced as more perspectives enter the system. No synthesis is published below 10 perspectives, and no statement is published with fewer than 5 supporters.

How LRDs Compare

Who Could Use LRDs?

Living Requirement Documents are designed for any entity that needs to understand what a population actually wants. Civic Filament is in beta with no live deployments yet, but these are the uses we're designing for:

• ◇City planners could use them to design proposals that have community support before the first public hearing
• ◇School boards could use them to allocate budgets in alignment with parent and teacher priorities
• ◇DAO governance could use them to write proposals that reflect member values, not just token-holder votes
• ◇Corporate HR could use them to understand employee needs with more fidelity than annual engagement surveys
• ◇Nonprofits could use them for community needs assessments that capture conditional support and genuine barriers

The Shift from "What Do You Think?" to "What Do You Need?"

The fundamental innovation of the LRD is that it captures needs rather than opinions. An opinion is brittle — it changes with framing, mood, and information. A need is stable — it reflects what a person actually requires to thrive.

When governance is built on needs rather than opinions, decisions become more durable, more legitimate, and more likely to actually solve the problems they're meant to address.

That's not an incremental improvement on the survey. It's a different paradigm for understanding collective intent.