Plain Language Is Not Enough: Reference, Time and Category as Barriers to Finding and Understanding Law

Slides from an invited talk given at the international conference "The Labyrinth of (Social) Rights: Linguistic, Systematic and Other Barriers to Access to Legal Protection" (ARIS project J5-50169), Faculty of Law, University of Ljubljana, 11 September 2026, panel 2 "AI and Legal Text Understanding". No conference proceedings are published. Plain-language reform makes legal sentences easier to read. It does not remove three barriers that remain standing after the text has been simplified: reference (where the applicable rule is actually written), time (whether what one is reading is still in force), and category (whether one falls under the rule or under one of its exceptions). The talk argues that these barriers are structural rather than stylistic, and therefore measurable. It presents measurements taken over the full corpus of Slovenian legislation, case law and legislative materials, using a knowledge graph built from open public sources and an AI legal assistant developed by the author. Findings presented include: the reference topology of social legislation (one article of the Exercise of Rights to Public Funds Act requires resolving sixteen descriptive cross-references, three of them to repealed acts; following references from that act reaches 30 acts in one hop and 371 in three, against a median of five for Slovenian acts); more than 1,200 references in valid acts that resolve to repealed legislation; the "shadow of a statute", that is, articles set aside by a notwithstanding clause in some other act while the official register records no change at all (389 articles across 113 acts, one article in eight in health legislation, arriving in waves in 2009-2012, 2020-2021 and 2023); a census of Slovenian judgments applying exceptio illegalis (113 judgments read individually, the power actually exercised in 70, more than half of all applications in the last four years), where the dominant pattern is secondary legislation adding a condition or category the statute does not know; and a first reading of official explanatory notes to draft articles as structured data (about 13,000 notes from two parliamentary terms, roughly one in ten explaining nothing beyond restating the article, one in seven for entirely new acts). The presentation distinguishes textual complexity from structural complexity and argues that the two are complementary research programmes: neither alone measures comprehensibility. Stated limits: published case law is a selection rather than a register, so all litigation counts are lower bounds; association is not causation, despite within-act comparison and placebo terms; and the exceptio illegalis count covers only judgments naming the doctrine explicitly. The views expressed are the author's own and do not represent official positions of the Government Office for Legislation.