Sources & influences
The criteria and weights are LexGrade's own. The method used to express them stands on prior work, credited here.
A standard should be candid about where it comes from. The sources below are credited because the method was genuinely drawn from them. Ideas and methods are free to build on, and the acknowledgment is owed regardless.
Scholarship
LexGrade's rubric-based approach (weighted components, capped point values, and narrative performance bands that reduce a grader's subjectivity to a shared vocabulary) builds on an established tradition in legal-writing assessment, in particular:
- Jessica Clark & Christy DeSanctis, Toward a Unified Grading Vocabulary: Using Rubrics in Legal Writing Courses, 63 J. Legal Educ. 3 (2013).
Empirical grounding
The hallucination figures cited on the homepage and in the Standard's ethics-gate rationale come from:
- Matthew Dahl, Varun Magesh, Mirac Suzgun & Daniel E. Ho, Large Legal Fictions: Profiling Legal Hallucinations in Large Language Models, 16 J. Legal Analysis 64 (2024), doi:10.1093/jla/laae003. The study measured hallucination rates in model answers to verifiable questions about federal court cases: 58% for ChatGPT-4, up to 88% for Llama 2.
LexGrade adapts that pedagogy from the law-school classroom to motion practice. The criteria, weights, and examples are its own. No rubric text or appendix from the cited work is reproduced here; LexGrade states its criteria in its own words.
Calibration authority
The calibration is measured against the rules that govern the practice it scores: the New York Civil Practice Law and Rules (CPLR) and the Uniform Civil Rules for the Supreme and County Courts (22 NYCRR Part 202). These are primary legal authority, cited here as calibration references. No endorsement of LexGrade is implied.
Building on a source does not imply that source endorses LexGrade. LexGrade confers no authority, and nothing here is legal advice. If you believe a source is used improperly or should be credited, tell us.