Democracy Monitor

Monitoring democratic institutions through public records

← Back to overview

Information Availability

infoAvailability

Public access to government data, FOIA compliance, and publication of mandated reports form the transparency infrastructure that enables democratic accountability. Removal of datasets, website takedowns, or suppression of required disclosures reduces the public's ability to monitor government conduct.

Public data is how you check the government’s claims — whichever side you are checking. Why this matters →

Know of a government action in this category that we missed? Tell us.View the AI prompts used to assess this category

Week of Sep 11, 2017

Notable departure from norms

AI content assessment elevated

Confirmed evidence: 2 actions · 0 discussions

The two-pass document review flags departures from baseline practice, corroborated by the second pass. Monitoring increased.

For the week of September 11, 2017, the Information Availability category registered a status consistent with notable departure, driven solely by the L2 AI content assessment layer. Of 9 documents collected, 2 were flagged at initial screening (P1) and both were assessed as potentially concerning at detailed review (P2), yielding a 100.0% P2 concern rate against a baseline of 11.0%. No documents were classified as clearly concerning.

This elevated concern rate may indicate shifts in how federal agencies were handling public information obligations during this period, which could affect the transparency infrastructure that enables citizens, journalists, and oversight bodies to monitor government activity. The structural context is minimal — only 9 documents were collected — and small compositional shifts (e.g., administrative procedure documents rising from 3.5% to 11.1%) are unreliable signals at this volume, where a single document entering or leaving the sample can dramatically alter percentages.

However, no P2-confirmed documents with full reasoning are available for this assessment, which significantly limits the ability to characterize specific government actions driving the elevated signal. The two potentially concerning documents were not elevated to "clearly concerning," meaning the AI reviewer identified features worth noting but did not find strong evidence of information availability restrictions in either case. Without access to the specific flagging rationale or document titles, it is not possible to describe which agencies, datasets, or transparency mechanisms were implicated.

Several counter-arguments deserve consideration, ranked by plausibility:

  1. Small-sample statistical artifact (most plausible). With only 2 documents reaching P2 review, a 100.0% concern rate carries extremely low statistical confidence. A single document assessed differently would have halved the rate to 50.0%. The elevated status may reflect sample noise rather than a meaningful pattern in government behavior.

  2. Threshold sensitivity at low volume. The P1 screening flagged 2 of 9 documents (22.2%), which is a modest absolute number. If the screening criteria captured documents that were ambiguous rather than clearly problematic — as suggested by both receiving "potentially concerning" rather than "clearly concerning" ratings — the elevated status may reflect cautious categorization rather than substantive transparency erosion.

  3. Routine administrative activity with transparency-adjacent language. Federal Register publications in this period may have included rulemaking or procedural notices that touch on information handling without representing any departure from longstanding practice. The shift toward administrative procedure documents could indicate routine process updates rather than policy changes restricting access.

  4. Seasonal or cyclical effects. The week falls near the end of the federal fiscal year (September 30), when agencies often publish notices related to program changes, funding reallocations, and regulatory updates. Elevated procedural activity during this period is a recurring pattern that could generate false positives in information availability screening.

Limitations: This assessment rests on 9 total documents, none of which have P2-confirmed detailed reasoning available for review. The 100.0% P2 concern rate, while arithmetically elevated, is derived from only 2 documents, making it a weak basis for trend identification. No specific government actions restricting information access can be identified from the available data.

View weekly summary for Sep 11, 2017 →

Week Archive#81 weeks with narratives