[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-llms-apply-rules-like-humans-study-finds":10,"sections":35},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},5502,"llms-apply-rules-like-humans-study-finds","LLMs Apply Rules Like Humans, Study Finds","A new psychology-inspired study finds large language models mimic human rule-following, even under time pressure, raising legal and philosophical questions.","A new study finds that large language models apply rules the way people actually do, not just by pattern-matching text.\n\nResearchers used psychology experiment designs, not just benchmark tests, to probe how large language models handle rule-based decisions. In one study, several LLMs mirrored the same patterns humans show when applying rules to new scenarios, whether the test cases were written before or after each model's training cutoff. Humans also showed some unexpected quirks depending on how a scenario was worded, and remarkably, the models reproduced those same quirks. A second study looked at what happens under time pressure: humans tend to lean more on a rule's literal wording than its underlying purpose when rushed, and Gemini Pro and Claude 3 showed the same shift, while GPT-4o and Llama 3.2 90B did not.\n\nThat gap between models is the real story. If some systems track the human habit of leaning on literal text under pressure and others don't, that's a meaningful difference for anyone building AI into legal or compliance workflows, where the letter versus the spirit of a rule is often the whole argument. The researchers frame this as evidence that LLMs have something like a genuine grasp of what a rule is, not just a knack for mimicking rule-following text.\n\nConvenient as that sounds for AI vendors, grasping a concept and reliably applying it under real-world stakes are two different things, and this study only tested the former.","[\"llms\",\"ai research\",\"legal tech\",\"psychology\"]","2026-08-18T04:00:00.000Z","2026-08-18T22:55:53.872Z","2026-08-18T22:56:05.787Z","published",null,[],"ai",[26,27,28,29],"llms","ai research","legal tech","psychology",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2503.00992",0,{"sections":36},[37,41,45,50,55,60,65,70,75,79,84,89,94,99],{"name":38,"slug":24,"count":39,"latest_published_at":40},"AI",3293,"2026-08-20T04:00:00.000Z",{"name":42,"slug":43,"count":44,"latest_published_at":40},"Security","security",435,{"name":46,"slug":47,"count":48,"latest_published_at":49},"Policy","policy",210,"2026-08-19T09:32:27.000Z",{"name":51,"slug":52,"count":53,"latest_published_at":54},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Hardware","hardware",140,"2026-08-19T18:25:42.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Consumer Tech","consumer-tech",95,"2026-08-18T16:05:00.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Science","science",90,"2026-08-19T18:41:02.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Software","software",73,"2026-08-18T07:51:50.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":18},"Dev Tools","dev-tools",69,{"name":80,"slug":81,"count":82,"latest_published_at":83},"Startups","startups",47,"2026-08-19T19:13:46.000Z",{"name":85,"slug":86,"count":87,"latest_published_at":88},"Gaming","gaming",41,"2026-07-09T04:00:00.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"General","general",33,"2026-08-18T22:18:13.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]