Datasets
FARS: What a Fatality Census Cannot Support
The three inclusion filters that decide what the federal fatality census holds.
Written for people who have to defend a method rather than describe one. Nothing below asks to be taken on trust: each piece opens with the documents it rests on, and says when it last watched them resolve.
Saturday, 29 August 2026
Litigation data
The 15-Month Year in the Federal Case Database
Reporting-period changes, unit shifts, and vanishing fields in the FJC Integrated Database.
Litigation data
What the RECAP Archive's Coverage Skews Toward
Why a contributed archive of federal filings is not a sample of federal litigation.
Explainable AI
Using the NIST AI RMF to Question a Vendor
The AI RMF read as a purchasing checklist rather than as a governance document.
Explainable AI
Who Classifies an AI System Under the EU AI Act
Provider and deployer duties, and the three ways a buyer inherits the provider role.
Datasets
Reconciling a Texas Fatal Crash Count
Why the state and federal fatal crash counts disagree, and how to reconcile them.
Friday, 28 August 2026
Datasets
TxDOT CRIS Crash Records: What They Support
The scope and the limits of the Texas Crash Records Information System.
Thursday, 16 July 2026
Datasets
FMCSA Crash and Carrier Data: A Method Guide
Which federal source answers which question about commercial vehicle crashes.
Thursday, 11 June 2026
Litigation data
How Case Outcomes Are Coded in Public Data
Disposition and nature-of-suit coding in the FJC Integrated Data Base.
Thursday, 14 May 2026
Litigation data
Reading a Federal Civil Docket as Data
What a federal civil docket does and does not record, from a data point of view.
Thursday, 9 April 2026
Explainable AI
Evaluating Accuracy Claims in Legal AI Tools
Why a bare accuracy percentage tells you almost nothing, and what to ask instead.
Thursday, 12 March 2026
Explainable AI
What Explainability Means in Document Review
The four NIST principles, applied to tools that classify legal documents.