[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"portal-settings:stajic:de":3,"public-menus:all":37,"post:how-far-is-an-ai-claim-from-the-evidence-evidence-distance:de":220,"related:post:how-far-is-an-ai-claim-from-the-evidence-evidence-distance:de:1":1555},{"statusCode":4,"data":5,"message":36},200,{"tenantId":6,"lang":7,"defaultLang":7,"siteUrl":8,"contactEmail":9,"brandName":10,"logoUrl":11,"siteName":10,"siteDescription":12,"ogImage":9,"robotsIndex":13,"socialLinks":9,"reservedSlugs":9,"seoPolicy":14},"stajic","de","https:\u002F\u002Fstajic.de",null,"Stajic Platform","\u002FLogo_Planet.svg","Stajic Portal",true,{"branding":15,"relatedContent":16,"crossDomainLinks":17},{"logoUrl":11},{"enabled":13},[18,21,24,27,30,33],{"url":19,"label":20,"isActive":13,"showInFooter":13,"includeInSameAs":13},"https:\u002F\u002Ffigure.rocks","figure.rocks",{"url":22,"label":23,"isActive":13,"showInFooter":13,"includeInSameAs":13},"https:\u002F\u002Floving.rocks","loving.rocks",{"url":25,"label":26,"isActive":13,"showInFooter":13,"includeInSameAs":13},"https:\u002F\u002Fbazify.com","bazify.com",{"url":28,"label":29,"isActive":13,"showInFooter":13,"includeInSameAs":13},"https:\u002F\u002Fbazify.de","bazify.de",{"url":31,"label":32,"isActive":13,"showInFooter":13,"includeInSameAs":13},"https:\u002F\u002Fbazify.at","bazify.at",{"url":34,"label":35,"isActive":13,"showInFooter":13,"includeInSameAs":13},"https:\u002F\u002Fbazify.ba","bazify.ba","Portal settings resolved",[38,44],{"id":39,"name":40,"location":41,"isActive":13,"isDefault":42,"items":43},1,"main-navigation","header",false,[],{"id":45,"name":46,"location":47,"isActive":13,"isDefault":13,"items":48},4,"main-menu","sidebar",[49,65,78,92,102,117,132],{"id":50,"title":51,"url":59,"target":60,"icon":61,"isActive":13,"type":62,"productId":9,"categoryId":9,"shopCategoryId":9,"articleId":9,"pageId":63,"portfolioId":9,"children":64},"item-18",{"de":52,"en":53,"es":54,"fr":55,"it":53,"ru":56,"sr":57,"zh":58},"Startseite","Home","Inicio","Accueil","Главная","Почетна","首页","\u002Ffull-stack-web-developer-munich-performance-seo-and-maintainable-builds","_self","i-lucide-home","page",111,[],{"id":66,"title":67,"url":74,"target":60,"icon":75,"isActive":13,"type":62,"productId":9,"categoryId":9,"shopCategoryId":9,"articleId":9,"pageId":76,"portfolioId":9,"children":77},"item-22",{"de":68,"en":68,"es":69,"fr":68,"it":70,"ru":71,"sr":72,"zh":73},"Vision","Visión","Visione","Видение","Визија","想象","\u002Fueber-uns-webdesign-muenchen-webaplikation","i-lucide-eye",113,[],{"id":79,"title":80,"url":88,"target":60,"icon":89,"isActive":13,"type":62,"productId":9,"categoryId":9,"shopCategoryId":9,"articleId":9,"pageId":90,"portfolioId":9,"children":91},"item-19",{"de":81,"en":82,"es":83,"fr":82,"it":84,"ru":85,"sr":86,"zh":87},"Leistungen","Services","Servicios","Servizi","Услуги","Услуге","服务","\u002Fservices-dienstleistungen-muenchen","i-lucide-wrench",116,[],{"id":93,"title":94,"url":98,"target":60,"icon":99,"isActive":13,"type":62,"productId":9,"categoryId":9,"shopCategoryId":9,"articleId":9,"pageId":100,"portfolioId":9,"children":101},"item-23",{"de":95,"en":95,"es":95,"fr":95,"it":95,"ru":96,"sr":96,"zh":97},"Blog","Блог","博客","\u002Fblog","i-lucide-book-open",112,[],{"id":103,"title":104,"url":113,"target":60,"icon":114,"isActive":13,"type":62,"productId":9,"categoryId":9,"shopCategoryId":9,"articleId":9,"pageId":115,"portfolioId":9,"children":116},"item-32",{"de":105,"en":106,"es":107,"fr":108,"it":109,"ru":110,"sr":111,"zh":112},"Neue Technologien","New Technologies","Nuevas tecnologías","Nouvelles technologies","Nuove tecnologie","Новые технологии","Нове технологије","新技术！","\u002Fneue-webtechnologien","i-lucide-sparkles",122,[],{"id":118,"title":119,"url":128,"target":60,"icon":129,"isActive":13,"type":62,"productId":9,"categoryId":9,"shopCategoryId":9,"articleId":9,"pageId":130,"portfolioId":9,"children":131},"item-20",{"de":120,"en":121,"es":122,"fr":123,"it":124,"ru":125,"sr":126,"zh":127},"Kontakt","Contact us!","Contacto","Contact","Contatto","Контакт","Контактирајте нас","联系我们！","\u002Fcontact","i-lucide-mail",115,[],{"id":133,"title":134,"url":143,"target":60,"icon":144,"isActive":13,"type":62,"productId":9,"categoryId":9,"shopCategoryId":9,"articleId":9,"pageId":145,"portfolioId":9,"children":146},"item-21",{"de":135,"en":136,"es":137,"fr":138,"it":139,"ru":140,"sr":141,"zh":142},"Unsere Arbeit","Our Work","Nuestro trabajo","Nos réalisations","I nostri lavori","Наши работы","Наши радови","文件夹","\u002Fportfolio","i-lucide-briefcase",114,[147,160,174,190,205],{"id":148,"title":149,"url":143,"target":60,"icon":158,"isActive":13,"type":62,"productId":9,"categoryId":9,"shopCategoryId":9,"articleId":9,"pageId":145,"portfolioId":9,"children":159},"item-24",{"de":150,"en":151,"es":152,"fr":153,"it":154,"ru":155,"sr":156,"zh":157},"Alle Projekte","All Projects","Todos los proyectos","Tous les projets","Tutti i progetti","Все проекты","Сви пројекти","所有项目","i-lucide-grid-3x3",[],{"id":161,"title":162,"url":170,"target":60,"icon":171,"isActive":13,"type":172,"productId":9,"categoryId":9,"shopCategoryId":9,"articleId":9,"pageId":9,"portfolioId":9,"children":173},"item-29",{"de":163,"en":164,"es":165,"fr":166,"it":167,"ru":168,"sr":169,"zh":142},"Local Roots, Global Reach","Local Roots - Global Reach","Empresa local ","Entreprise locale","Azienda locale","Местная компания","Локално предузеће глобално тржиште","\u002Fportfolio\u002Flocal-roots-global-reach-communication-media-systems-for-modern-business","i-lucide-globe-2","custom",[],{"id":175,"title":176,"url":185,"target":60,"icon":186,"isActive":13,"type":187,"productId":9,"categoryId":9,"shopCategoryId":9,"articleId":9,"pageId":9,"portfolioId":188,"children":189},"menu-2916d060-8681-482a-af9a-c107a8bc30d8",{"de":177,"en":178,"es":179,"fr":180,"it":181,"ru":182,"sr":183,"zh":184},"Von Rohdaten zu verifizierbarem KI-Kontext: Eine praktische Architektur für agentenbereite Informationen","From Raw Data to Verifiable AI Context:","De datos sin procesar a contexto de IA verificable: una arquitectura práctica para información lista para agentes","Des données brutes au contexte d'IA vérifiable : une architecture pratique pour des informations prêtes pour les agents","Dai dati grezzi a un contesto IA verificabile: Un'architettura pratica per informazioni pronte per gli agenti","От необработанных данных к верифицируемому контексту ИИ: практическая архитектура информации, готовой для агентов","Od sirovih podataka do proverljivog AI konteksta: Praktična arhitektura za informacije spremne za agente","从原始数据到可验证的 AI 上下文：面向智能体就绪信息的实用架构","\u002Fportfolio\u002Ffrom-raw-data-to-verifiable-ai-context-a-practical-architecture-for-agent-ready-information","i-lucide-database","portfolio",22,[],{"id":191,"title":192,"url":201,"target":60,"icon":202,"isActive":13,"type":187,"productId":9,"categoryId":9,"shopCategoryId":9,"articleId":9,"pageId":9,"portfolioId":203,"children":204},"menu-e2061cd7-ff53-49d5-a0e8-3d49090577c0",{"de":193,"en":194,"es":195,"fr":196,"it":197,"ru":198,"sr":199,"zh":200},"Präsenzplattformen — Unternehmenskommunikation","Presence Platforms — Enterprise Communication","Plataformas de presencia — Comunicación empresarial","Plateformes de présence — Communication d'entreprise","Presence Platforms — Comunicazione aziendale","Платформы присутствия — Корпоративные коммуникации","Presence platforme — Poslovna komunikacija","Presence 平台 — 企业通信","\u002Fportfolio\u002Fpresence-platforms-enterprise-communication-that-wins-without-competing","i-lucide-badge-percent",21,[],{"id":206,"title":207,"url":216,"target":60,"icon":217,"isActive":13,"type":187,"productId":9,"categoryId":9,"shopCategoryId":9,"articleId":9,"pageId":9,"portfolioId":218,"children":219},"menu-a781582e-e90d-418e-b8c9-b070bc37c37e",{"de":208,"en":209,"es":210,"fr":211,"it":212,"ru":213,"sr":214,"zh":215},"Maximierung des POS-ROI durch Konsistenz","Maximizing POS ROI Through Consistent","Maximizar el ROI del POS a través de constante","Maximiser le ROI du PDV grâce à la cohérence","Massimizzare il ROI del POS attraverso costante","Максимизация ROI POS за счет последовательного","Maksimizacija POS ROI-ja kroz dosledno","通过持续最大化 POS 投资回报率","\u002Fportfolio\u002Ffrom-vision-to-value-maximizing-pos-roi-through-consistent-ci-creation-and-strategic-branding","i-lucide-briefcase-business",11,[],{"statusCode":4,"data":221,"message":1554},{"id":222,"title":223,"slug":224,"content":225,"contentJson":226,"excerpt":740,"featuredImage":741,"featuredImageAlt":742,"featuredImageCaption":9,"featuredImageTitle":9,"featuredImageCopyright":9,"featuredImageAuthor":9,"featuredImageSourceUrl":9,"featuredImageLicense":9,"featuredImageIsAiGenerated":42,"status":743,"publishedAt":744,"createdAt":745,"updatedAt":746,"seoLocalePaths":747,"categories":756,"author":765,"translations":769},"497","Wie weit ist eine KI-Behauptung von den Beweisen entfernt? — Evidenzabstand","how-far-is-an-ai-claim-from-the-evidence-evidence-distance","{\"time\":1792096889655,\"blocks\":[{\"id\":\"pBKmj2rADA\",\"type\":\"header\",\"data\":{\"text\":\"Frage\",\"level\":2},\"tunes\":{}},{\"id\":\"vUM_NFoPgQ\",\"type\":\"paragraph\",\"data\":{\"text\":\"Zwei KI-Antworten können beide Quellen zitieren und sich dennoch stark darin unterscheiden, wie direkt diese Quellen die getroffenen Aussagen stützen. Wie können wir diesen Unterschied beschreiben?\"},\"tunes\":{}},{\"id\":\"RVy4DRebuB\",\"type\":\"paragraph\",\"data\":{\"text\":\"Dieser Artikel verwendet den Begriff Evidenzdistanz für den Abstand zwischen einer Aussage und der direkt überprüfbaren Evidenz, die sie stützt.\"},\"tunes\":{}},{\"id\":\"2BLCxDL7A6\",\"type\":\"quote\",\"data\":{\"text\":\"Evidenzdistanz ist die Anzahl und Stärke von Transformationen, Annahmen und Inferenzschritten, die eine Aussage von der Evidenz trennen, die sie direkt verifizieren kann.\",\"caption\":\"Arbeitsdefinition — ein hier eingeführtes Denkkonzept, keine standardisierte Metrik.\",\"alignment\":\"left\"},\"tunes\":{}},{\"id\":\"wjo5iUIK34\",\"type\":\"tableOfContents\",\"data\":{\"title\":\"Contents\",\"minLevel\":2,\"maxLevel\":3},\"tunes\":{}},{\"id\":\"eWlrm2ZYBz\",\"type\":\"header\",\"data\":{\"text\":\"Was das wirklich bedeutet\",\"level\":2},\"tunes\":{}},{\"id\":\"nSTGfBRYGB\",\"type\":\"paragraph\",\"data\":{\"text\":\"Ein Zitat beweist, dass eine Quelle referenziert wurde. Es beweist nicht automatisch, dass die Quelle den Satz neben dem Zitat direkt stützt.\"},\"tunes\":{}},{\"id\":\"FGv6LYgd6R\",\"type\":\"paragraph\",\"data\":{\"text\":\"Zwischen einer ursprünglichen Beobachtung und einer endgültigen KI-Antwort können Extraktion, Parsing, Chunking, Retrieval, Zusammenfassung, Interpretation, Aggregation und Inferenz liegen.\"},\"tunes\":{}},{\"id\":\"nMnqK_QXbX\",\"type\":\"code\",\"data\":{\"code\":\"Primary evidence\\n   ↓\\nExtracted data\\n   ↓\\nRetrieved passage\\n   ↓\\nSummary\\n   ↓\\nInterpretation\\n   ↓\\nInference\\n   ↓\\nFinal claim\"},\"tunes\":{}},{\"id\":\"4N0XSdEmOI\",\"type\":\"paragraph\",\"data\":{\"text\":\"Jeder Schritt kann legitim sein. Aber jeder Schritt kann auch Verzerrung einführen.\"},\"tunes\":{}},{\"id\":\"nPHs66wwFR\",\"type\":\"header\",\"data\":{\"text\":\"Einfachstes Beispiel\",\"level\":2},\"tunes\":{}},{\"id\":\"Xcc8O821lv\",\"type\":\"paragraph\",\"data\":{\"text\":\"Angenommen, ein Temperatursensor meldet 38,2 °C.\"},\"tunes\":{}},{\"id\":\"dlmDyh6NOP\",\"type\":\"table\",\"data\":{\"withHeadings\":true,\"stretched\":false,\"content\":[[\"Aussage\",\"Beziehung zur Evidenz\",\"Veranschaulichende Distanz\"],[\"Der Sensor meldet 38,2 °C.\",\"Direkte Messung\",\"Gering\"],[\"Das System läuft heißer als gestern.\",\"Erfordert Vergleich mit der gestrigen Messung\",\"Mittel\"],[\"Das Kühlsystem versagt.\",\"Erfordert kausale Inferenz und alternative Erklärungen\",\"Hoch\"]]},\"tunes\":{}},{\"id\":\"F623zQhLQB\",\"type\":\"paragraph\",\"data\":{\"text\":\"Der ursprüngliche Sensorwert stützt die erste Aussage direkt. Er beweist für sich genommen nicht die dritte.\"},\"tunes\":{}},{\"id\":\"AtPqdtQ3tK\",\"type\":\"code\",\"data\":{\"code\":\"Evidence exists\\n        ≠\\nEvidence directly supports every derived claim\"},\"tunes\":{}},{\"id\":\"3fWULI4ZZa\",\"type\":\"header\",\"data\":{\"text\":\"Wo das Beispiel nicht mehr funktioniert\",\"level\":2},\"tunes\":{}},{\"id\":\"lWg8fQjCbR\",\"type\":\"paragraph\",\"data\":{\"text\":\"Evidenzdistanz ist nicht einfach eine Zählung von Sätzen oder Verarbeitungsschritten.\"},\"tunes\":{}},{\"id\":\"UwXXyS_xnp\",\"type\":\"paragraph\",\"data\":{\"text\":\"Eine mathematisch gültige Transformation kann Evidenz nahezu perfekt bewahren, während eine einzige spekulative Interpretation einen viel größeren epistemischen Sprung erzeugen kann.\"},\"tunes\":{}},{\"id\":\"qOtlSqbZmt\",\"type\":\"paragraph\",\"data\":{\"text\":\"Das Konzept umfasst daher sowohl Distanz durch Transformation als auch Distanz durch Inferenz.\"},\"tunes\":{}},{\"id\":\"z0tE1scu1w\",\"type\":\"header\",\"data\":{\"text\":\"Direkte Antwort\",\"level\":2},\"tunes\":{}},{\"id\":\"kazpS6DqUn\",\"type\":\"paragraph\",\"data\":{\"text\":\"Die Evidenzdistanz nimmt zu, wenn eine Behauptung von mehr Transformationen, mehr Annahmen, schwächerer Provenienz oder stärkerer Inferenz abhängt, als die zugrunde liegende Evidenz direkt rechtfertigt.\"},\"tunes\":{}},{\"id\":\"64c0vwT55C\",\"type\":\"code\",\"data\":{\"code\":\"Short distance:\\nsource → claim\\n\\nLonger distance:\\nsource → extraction → summary → assumption → inference → claim\"},\"tunes\":{}},{\"id\":\"MhOixLLuV7\",\"type\":\"paragraph\",\"data\":{\"text\":\"Eine größere Distanz macht eine Behauptung nicht automatisch falsch. Sie bedeutet, dass die Argumentationskette mehr Validierung erfordert.\"},\"tunes\":{}},{\"id\":\"wVsphcBSqw\",\"type\":\"header\",\"data\":{\"text\":\"Warum das für KI wichtig ist\",\"level\":2},\"tunes\":{}},{\"id\":\"MkFG6A0RsC\",\"type\":\"paragraph\",\"data\":{\"text\":\"LLMs sind besonders gut darin, flüssige Brücken zwischen Informationen zu erzeugen. Diese Fähigkeit ist nützlich, aber Flüssigkeit kann verbergen, wie weit sich die Schlussfolgerung von der direkt abgerufenen Evidenz entfernt hat.\"},\"tunes\":{}},{\"id\":\"4LXQyZ_nP2\",\"type\":\"paragraph\",\"data\":{\"text\":\"Eine Antwort kann daher gut belegt erscheinen, während sie Behauptungen enthält, deren Stützung nur durch mehrere unausgesprochene Inferenzschritte besteht.\"},\"tunes\":{}},{\"id\":\"d0XgOLAuI8\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fstajic.de\u002Fde\u002Fblog\u002Ffrom-research-protocol-to-a-general-ai-reasoning-framework\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Vom Forschungsprotokoll zu einem allgemeinen KI-Reasoning-Framework\",\"description\":\"Zuverlässiges Schlussfolgern erfordert eine explizite Trennung von Evidenz, Annahmen, Hypothesen und Schlussfolgerungen.\"}},\"tunes\":{}},{\"id\":\"PjYzIFYT1U\",\"type\":\"header\",\"data\":{\"text\":\"Vier Komponenten der Evidenzdistanz\",\"level\":2},\"tunes\":{}},{\"id\":\"0x9QYuuH-l\",\"type\":\"header\",\"data\":{\"text\":\"1. Transformationsdistanz\",\"level\":3},\"tunes\":{}},{\"id\":\"gJ4DxD4oGt\",\"type\":\"paragraph\",\"data\":{\"text\":\"Wie oft wurde die ursprüngliche Evidenz konvertiert, zusammengefasst, normalisiert oder umformatiert, bevor sie das Modell erreicht hat?\"},\"tunes\":{}},{\"id\":\"MWNx-GiITU\",\"type\":\"header\",\"data\":{\"text\":\"2. Inferenzdistanz\",\"level\":3},\"tunes\":{}},{\"id\":\"9iJhSbpMrd\",\"type\":\"paragraph\",\"data\":{\"text\":\"Wie viele Argumentationsschritte sind erforderlich, bevor die Evidenz die endgültige Schlussfolgerung stützt?\"},\"tunes\":{}},{\"id\":\"FSrQHI2HdF\",\"type\":\"header\",\"data\":{\"text\":\"3. Provenienzdistanz\",\"level\":3},\"tunes\":{}},{\"id\":\"i_5MJVmw6r\",\"type\":\"paragraph\",\"data\":{\"text\":\"Kann die abgerufene Aussage noch auf eine identifizierbare Originalquelle zurückgeführt werden, oder ist sie durch Zusammenfassungen und Kopien gegangen, deren Ursprung unklar ist?\"},\"tunes\":{}},{\"id\":\"dhyb_9AYdc\",\"type\":\"header\",\"data\":{\"text\":\"4. Anwendbarkeitsdistanz\",\"level\":3},\"tunes\":{}},{\"id\":\"SDYDfH21Nz\",\"type\":\"paragraph\",\"data\":{\"text\":\"Wie viel Argumentation ist erforderlich, um festzustellen, dass die Evidenz auf genau diese Version, diesen Nutzer, diese Gerichtsbarkeit, diesen Zeitraum oder diese Situation zutrifft?\"},\"tunes\":{}},{\"id\":\"aonb2aApom\",\"type\":\"header\",\"data\":{\"text\":\"Evidenzdistanz in RAG\",\"level\":2},\"tunes\":{}},{\"id\":\"c5i9oX23SF\",\"type\":\"paragraph\",\"data\":{\"text\":\"RAG kann eine wichtige Distanz verkürzen: die Distanz zwischen einem Modell und externem Wissen.\"},\"tunes\":{}},{\"id\":\"bRgk5469WZ\",\"type\":\"paragraph\",\"data\":{\"text\":\"Es verkürzt nicht automatisch die Distanz zwischen einer abgerufenen Passage und der endgültigen Behauptung.\"},\"tunes\":{}},{\"id\":\"x6Ji-N-lry\",\"type\":\"code\",\"data\":{\"code\":\"RAG solves:\\nmodel ↔ external information\\n\\nRAG does not automatically solve:\\nevidence ↔ justified conclusion\"},\"tunes\":{}},{\"id\":\"gRrOi5JDnn\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fstajic.de\u002Fde\u002Fblog\u002Fwhere-does-an-llm-get-its-data-rag-data-sources-in-python\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Woher bezieht ein LLM seine Daten? RAG-Datenquellen in Python\",\"description\":\"Retrieval bringt externe Informationen in den Kontext. Evidence Distance fragt, was zwischen diesen Informationen und der schließlich erzeugten Behauptung passiert.\"}},\"tunes\":{}},{\"id\":\"Or0p1x40Fc\",\"type\":\"header\",\"data\":{\"text\":\"Verbindung zum Retrieval Trigger\",\"level\":2},\"tunes\":{}},{\"id\":\"H6s1-2OI7y\",\"type\":\"paragraph\",\"data\":{\"text\":\"Der Retrieval Trigger fragt, ob das interne Modellwissen ausreichend ist.\"},\"tunes\":{}},{\"id\":\"pbGaZ-mRzL\",\"type\":\"paragraph\",\"data\":{\"text\":\"Evidence Distance stellt eine spätere Frage: Wie direkt stützt es, nachdem Evidenz abgerufen wurde, das, was das System behaupten möchte?\"},\"tunes\":{}},{\"id\":\"XCegS6de5e\",\"type\":\"code\",\"data\":{\"code\":\"Need evidence?\\n   ↓\\nRetrieval Trigger\\n   ↓\\nRetrieve\\n   ↓\\nHow directly does evidence support the claim?\\n   ↓\\nEvidence Distance\"},\"tunes\":{}},{\"id\":\"Diu2wdjbJz\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fstajic.de\u002Fde\u002Fblog\u002Fwhen-should-an-ai-stop-trusting-its-own-knowledge-the-retrieval-trigger\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Wann sollte eine KI aufhören, ihrem eigenen Wissen zu vertrauen? — Der Retrieval Trigger\",\"description\":\"Der Trigger entscheidet, wann Evidenz benötigt wird. Evidence Distance bewertet, wie weit die schließliche Behauptung über diese Evidenz hinausgeht.\"}},\"tunes\":{}},{\"id\":\"dHdNuaHXOS\",\"type\":\"header\",\"data\":{\"text\":\"Eine praktische Diagnose\",\"level\":2},\"tunes\":{}},{\"id\":\"zxTVFZIVpw\",\"type\":\"paragraph\",\"data\":{\"text\":\"Rekonstruieren Sie für eine wichtige KI-generierte Behauptung die Stützkette rückwärts.\"},\"tunes\":{}},{\"id\":\"X8sRRGeav4\",\"type\":\"code\",\"data\":{\"code\":\"CLAIM\\n  ↑\\nWhich inference produced it?\\n  ↑\\nWhich retrieved evidence supports that inference?\\n  ↑\\nWhich source produced that evidence?\\n  ↑\\nCan the original source still be inspected?\"},\"tunes\":{}},{\"id\":\"7irzpI0oMV\",\"type\":\"paragraph\",\"data\":{\"text\":\"Wenn die Kette bricht, ist das System auf ein Problem der Evidenznachverfolgbarkeit gestoßen.\"},\"tunes\":{}},{\"id\":\"kKNI-ESAqP\",\"type\":\"paragraph\",\"data\":{\"text\":\"Wenn die Kette intakt bleibt, aber mehrere starke Annahmen erfordert, ist die Evidence Distance immer noch hoch.\"},\"tunes\":{}},{\"id\":\"vn-vbQGoIC\",\"type\":\"header\",\"data\":{\"text\":\"Beispiel: Produktdaten\",\"level\":2},\"tunes\":{}},{\"id\":\"9bigaWbg_3\",\"type\":\"paragraph\",\"data\":{\"text\":\"Angenommen, eine Herstellerspezifikation besagt, dass ein Laptop 1,48 kg wiegt.\"},\"tunes\":{}},{\"id\":\"IfKrukuOp9\",\"type\":\"code\",\"data\":{\"code\":\"Manufacturer specification\\n→ 1.48 kg\\n→ \\\"The laptop weighs 1.48 kg\\\"\\n\\nLow Evidence Distance\"},\"tunes\":{}},{\"id\":\"uIwRyMs4U8\",\"type\":\"paragraph\",\"data\":{\"text\":\"Betrachten Sie nun:\"},\"tunes\":{}},{\"id\":\"ih07pbVjtb\",\"type\":\"code\",\"data\":{\"code\":\"Manufacturer specification\\n→ 1.48 kg\\n→ comparison with three competitors\\n→ assumption about travel behavior\\n→ \\\"This is the best laptop for frequent travelers\\\"\\n\\nMuch higher Evidence Distance\"},\"tunes\":{}},{\"id\":\"qKO4_aR6vR\",\"type\":\"paragraph\",\"data\":{\"text\":\"Die Spezifikation ist real. Die Empfehlung ist noch mehrere Argumentationsschritte davon entfernt.\"},\"tunes\":{}},{\"id\":\"90zOwAjmvt\",\"type\":\"header\",\"data\":{\"text\":\"Beispiel: Wissenschaftliche Evidenz\",\"level\":2},\"tunes\":{}},{\"id\":\"gju8RS56u4\",\"type\":\"paragraph\",\"data\":{\"text\":\"Eine Studie kann eine beobachtete Assoziation berichten. Eine KI-Antwort kann diese fälschlicherweise zu einer Kausalaussage verdichten.\"},\"tunes\":{}},{\"id\":\"LgAJrZoaBe\",\"type\":\"code\",\"data\":{\"code\":\"Observed association\\n        ↓\\nInterpretation\\n        ↓\\nUnstated causal assumption\\n        ↓\\n\\\"X causes Y\\\"\"},\"tunes\":{}},{\"id\":\"-BnXVHW_7R\",\"type\":\"paragraph\",\"data\":{\"text\":\"Das Problem ist nicht, dass keine Evidenz existiert. Das Problem ist, dass die abschließende Aussage eine größere Evidenzdistanz überbrückt, als die Studie direkt stützt.\"},\"tunes\":{}},{\"id\":\"sKywq0xlRb\",\"type\":\"header\",\"data\":{\"text\":\"Verbindung zu Agent-Ready Context\",\"level\":2},\"tunes\":{}},{\"id\":\"a1-X0w-8Pn\",\"type\":\"paragraph\",\"data\":{\"text\":\"Strukturierte Informationsobjekte können vermeidbare Evidenzdistanz reduzieren, indem sie Quellenidentität, Einheiten, Zeitstempel, Konflikte und Gültigkeit bewahren, anstatt das Modell zu zwingen, sie aus abgeflachtem Text zu rekonstruieren.\"},\"tunes\":{}},{\"id\":\"EZywuvf7qq\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fstajic.de\u002Fde\u002Fportfolio\u002Ffrom-raw-data-to-verifiable-ai-context-a-practical-architecture-for-agent-ready-information\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Von Rohdaten zu verifizierbarem KI-Kontext: Eine praktische Architektur für agentenbereite Informationen\",\"description\":\"Die Bewahrung von Herkunft und Gültigkeit vor dem Abruf reduziert die Anzahl der später erforderlichen ungestützten Rekonstruktionsschritte.\"}},\"tunes\":{}},{\"id\":\"7JLsOQVOfL\",\"type\":\"header\",\"data\":{\"text\":\"Eine einfache Arbeitsskala\",\"level\":2},\"tunes\":{}},{\"id\":\"WLdbzx22ZH\",\"type\":\"table\",\"data\":{\"withHeadings\":true,\"stretched\":false,\"content\":[[\"Stufe\",\"Beschreibung\"],[\"ED0\",\"Direkt berichtete Quellentatsache\"],[\"ED1\",\"Einfache deterministische Transformation oder direkter Vergleich\"],[\"ED2\",\"Ein bedeutungsvoller interpretativer Schritt\"],[\"ED3\",\"Mehrere Annahmen oder kombinierte Quellen\"],[\"ED4\",\"Starke kausale, prädiktive oder präskriptive Schlussfolgerung über die direkte Evidenz hinaus\"]]},\"tunes\":{}},{\"id\":\"9FvtJjfCLQ\",\"type\":\"paragraph\",\"data\":{\"text\":\"Diese Skala ist illustrativ, kein wissenschaftlicher Standard. Ihr Zweck ist es, eine explizite Frage zu erzwingen: Wie viel mussten wir zwischen der Evidenz und der Behauptung hinzufügen?\"},\"tunes\":{}},{\"id\":\"6hCnJtbh1y\",\"type\":\"header\",\"data\":{\"text\":\"Missverständnisse\",\"level\":2},\"tunes\":{}},{\"id\":\"g0kaPh4HxV\",\"type\":\"paragraph\",\"data\":{\"text\":\"Eine Zitation bedeutet, dass die Evidenzdistanz gering ist. Nein. Eine Zitation kann neben einer Behauptung stehen, die das, was die zitierte Quelle belegt, erheblich überschreitet.\"},\"tunes\":{}},{\"id\":\"3vnmNlD5Mk\",\"type\":\"paragraph\",\"data\":{\"text\":\"Eine lange Evidenzdistanz bedeutet, dass die Behauptung falsch ist. Nein. Komplexe Schlussfolgerungen können gut gestützt sein, erfordern aber eine stärkere Argumentationskette.\"},\"tunes\":{}},{\"id\":\"nLnebZgawL\",\"type\":\"paragraph\",\"data\":{\"text\":\"Primärquellen eliminieren Inferenz. Nein. Primäre Evidenz kann dennoch Interpretation erfordern.\"},\"tunes\":{}},{\"id\":\"ypGitVy9x3\",\"type\":\"paragraph\",\"data\":{\"text\":\"Mehr Quellen reduzieren automatisch die Distanz. Zehn schwache Wiederholungen derselben ungestützten Behauptung erzeugen keine direkte Evidenz.\"},\"tunes\":{}},{\"id\":\"W-M4nnnzEg\",\"type\":\"header\",\"data\":{\"text\":\"Grenzfall: Abgeleitete Daten\",\"level\":2},\"tunes\":{}},{\"id\":\"YEzmltN56F\",\"type\":\"paragraph\",\"data\":{\"text\":\"Einige abgeleitete Werte sind zuverlässiger als rohe menschliche Beschreibungen. Eine korrekt berechnete Summe aus verifizierten Transaktionen kann äußerst verlässlich sein, obwohl sie nur eine Transformation von den Originalaufzeichnungen entfernt ist.\"},\"tunes\":{}},{\"id\":\"FB4n6LMnDW\",\"type\":\"paragraph\",\"data\":{\"text\":\"Die Evidenzdistanz misst daher die epistemische Trennung, nicht eine vereinfachte Vorliebe für unberührte Daten.\"},\"tunes\":{}},{\"id\":\"mge41Ld_cD\",\"type\":\"header\",\"data\":{\"text\":\"Einschränkungen\",\"level\":2},\"tunes\":{}},{\"id\":\"85Agxl7u5w\",\"type\":\"paragraph\",\"data\":{\"text\":\"Die Evidenzdistanz ist eine konzeptionelle Diagnose, kein universeller quantitativer Wert.\"},\"tunes\":{}},{\"id\":\"b-p7FbBY8Y\",\"type\":\"paragraph\",\"data\":{\"text\":\"Verschiedene Domänen erfordern unterschiedliche Behandlung von Inferenz. Mathematische Beweise, klinische Evidenz, Softwareprotokolle, Marktanalysen und historische Interpretation können keinen gemeinsamen universellen numerischen Schwellenwert teilen.\"},\"tunes\":{}},{\"id\":\"EksjnuYWwq\",\"type\":\"header\",\"data\":{\"text\":\"Was würde diese Antwort ändern?\",\"level\":2},\"tunes\":{}},{\"id\":\"gnIKkYxGVY\",\"type\":\"paragraph\",\"data\":{\"text\":\"Zukünftige KI-Systeme könnten explizite Provenienzgraphen, maschinenprüfbare Behauptungs-Evidenz-Verknüpfungen und typisierte Argumentationsspuren generieren. Das könnte Teile der Evidenzdistanz automatisch messbar machen.\"},\"tunes\":{}},{\"id\":\"NDNOC8F5BM\",\"type\":\"paragraph\",\"data\":{\"text\":\"Das zugrunde liegende Problem bleibt bestehen: Eine Behauptung kann sich weiter von ihrer Evidenz entfernen, selbst wenn jeder Zwischensatz plausibel klingt.\"},\"tunes\":{}},{\"id\":\"N9WxXZXU9K\",\"type\":\"header\",\"data\":{\"text\":\"Fazit\",\"level\":2},\"tunes\":{}},{\"id\":\"B0nu1n-lr8\",\"type\":\"paragraph\",\"data\":{\"text\":\"Die wichtige Unterscheidung liegt nicht nur zwischen belegten und unbelegten Antworten.\"},\"tunes\":{}},{\"id\":\"dUA5AMapMG\",\"type\":\"paragraph\",\"data\":{\"text\":\"Sie liegt zwischen Behauptungen, die nahe an direkt überprüfbarer Evidenz bleiben, und Behauptungen, die von zunehmend langen Ketten der Interpretation und Annahme abhängen.\"},\"tunes\":{}},{\"id\":\"YMLENvLaug\",\"type\":\"code\",\"data\":{\"code\":\"Evidence\\n   ↓\\nEvidence Distance\\n   ↓\\nReasoning\\n   ↓\\nAnswer Validity Boundary\"},\"tunes\":{}},{\"id\":\"Z_Cu0srfiO\",\"type\":\"paragraph\",\"data\":{\"text\":\"Die Evidenzdistanz gibt uns eine Sprache, um diesen Raum zu beschreiben.\"},\"tunes\":{}},{\"id\":\"AUEeDCNKRA\",\"type\":\"header\",\"data\":{\"text\":\"Primärquellen\",\"level\":2},\"tunes\":{}},{\"id\":\"eHeLgVtehs\",\"type\":\"paragraph\",\"data\":{\"text\":\"Patrick Lewis et al., \u003Ca href=\\\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.11401\\\" target=\\\"_blank\\\">Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks\u003C\u002Fa> (2020).\"},\"tunes\":{}},{\"id\":\"i--E80s2nB\",\"type\":\"paragraph\",\"data\":{\"text\":\"Nelson F. Liu et al., \u003Ca href=\\\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2307.03172\\\" target=\\\"_blank\\\">Lost in the Middle: How Language Models Use Long Contexts\u003C\u002Fa> (2023).\"},\"tunes\":{}},{\"id\":\"U2ByVS6VcD\",\"type\":\"paragraph\",\"data\":{\"text\":\"W3C, \u003Ca href=\\\"https:\u002F\u002Fwww.w3.org\u002FTR\u002Fprov-overview\u002F\\\" target=\\\"_blank\\\">PROV Overview\u003C\u002Fa>. Ein standardisiertes Modell zur Darstellung von Provenienzbeziehungen.\"},\"tunes\":{}}],\"version\":\"2.31.6\"}",{"time":227,"blocks":228,"version":739},1792096889655,[229,235,241,246,254,261,266,271,276,282,287,292,297,319,324,329,334,339,344,349,354,359,364,369,374,379,384,395,400,405,410,415,420,425,430,435,440,445,450,455,460,469,474,479,484,489,498,503,508,513,518,523,528,533,538,543,548,553,558,563,568,573,578,583,591,596,619,624,629,634,639,644,649,654,659,664,669,674,679,684,689,694,699,704,709,714,719,724,729,734],{"id":230,"data":231,"type":41,"tunes":234},"pBKmj2rADA",{"text":232,"level":233},"Frage",2,{},{"id":236,"data":237,"type":239,"tunes":240},"vUM_NFoPgQ",{"text":238},"Zwei KI-Antworten können beide Quellen zitieren und sich dennoch stark darin unterscheiden, wie direkt diese Quellen die getroffenen Aussagen stützen. Wie können wir diesen Unterschied beschreiben?","paragraph",{},{"id":242,"data":243,"type":239,"tunes":245},"RVy4DRebuB",{"text":244},"Dieser Artikel verwendet den Begriff Evidenzdistanz für den Abstand zwischen einer Aussage und der direkt überprüfbaren Evidenz, die sie stützt.",{},{"id":247,"data":248,"type":252,"tunes":253},"2BLCxDL7A6",{"text":249,"caption":250,"alignment":251},"Evidenzdistanz ist die Anzahl und Stärke von Transformationen, Annahmen und Inferenzschritten, die eine Aussage von der Evidenz trennen, die sie direkt verifizieren kann.","Arbeitsdefinition — ein hier eingeführtes Denkkonzept, keine standardisierte Metrik.","left","quote",{},{"id":255,"data":256,"type":259,"tunes":260},"wjo5iUIK34",{"title":257,"maxLevel":258,"minLevel":233},"Contents",3,"tableOfContents",{},{"id":262,"data":263,"type":41,"tunes":265},"eWlrm2ZYBz",{"text":264,"level":233},"Was das wirklich bedeutet",{},{"id":267,"data":268,"type":239,"tunes":270},"nSTGfBRYGB",{"text":269},"Ein Zitat beweist, dass eine Quelle referenziert wurde. Es beweist nicht automatisch, dass die Quelle den Satz neben dem Zitat direkt stützt.",{},{"id":272,"data":273,"type":239,"tunes":275},"FGv6LYgd6R",{"text":274},"Zwischen einer ursprünglichen Beobachtung und einer endgültigen KI-Antwort können Extraktion, Parsing, Chunking, Retrieval, Zusammenfassung, Interpretation, Aggregation und Inferenz liegen.",{},{"id":277,"data":278,"type":280,"tunes":281},"nMnqK_QXbX",{"code":279},"Primary evidence\n   ↓\nExtracted data\n   ↓\nRetrieved passage\n   ↓\nSummary\n   ↓\nInterpretation\n   ↓\nInference\n   ↓\nFinal claim","code",{},{"id":283,"data":284,"type":239,"tunes":286},"4N0XSdEmOI",{"text":285},"Jeder Schritt kann legitim sein. Aber jeder Schritt kann auch Verzerrung einführen.",{},{"id":288,"data":289,"type":41,"tunes":291},"nPHs66wwFR",{"text":290,"level":233},"Einfachstes Beispiel",{},{"id":293,"data":294,"type":239,"tunes":296},"Xcc8O821lv",{"text":295},"Angenommen, ein Temperatursensor meldet 38,2 °C.",{},{"id":298,"data":299,"type":317,"tunes":318},"dlmDyh6NOP",{"content":300,"stretched":42,"withHeadings":13},[301,305,309,313],[302,303,304],"Aussage","Beziehung zur Evidenz","Veranschaulichende Distanz",[306,307,308],"Der Sensor meldet 38,2 °C.","Direkte Messung","Gering",[310,311,312],"Das System läuft heißer als gestern.","Erfordert Vergleich mit der gestrigen Messung","Mittel",[314,315,316],"Das Kühlsystem versagt.","Erfordert kausale Inferenz und alternative Erklärungen","Hoch","table",{},{"id":320,"data":321,"type":239,"tunes":323},"F623zQhLQB",{"text":322},"Der ursprüngliche Sensorwert stützt die erste Aussage direkt. Er beweist für sich genommen nicht die dritte.",{},{"id":325,"data":326,"type":280,"tunes":328},"AtPqdtQ3tK",{"code":327},"Evidence exists\n        ≠\nEvidence directly supports every derived claim",{},{"id":330,"data":331,"type":41,"tunes":333},"3fWULI4ZZa",{"text":332,"level":233},"Wo das Beispiel nicht mehr funktioniert",{},{"id":335,"data":336,"type":239,"tunes":338},"lWg8fQjCbR",{"text":337},"Evidenzdistanz ist nicht einfach eine Zählung von Sätzen oder Verarbeitungsschritten.",{},{"id":340,"data":341,"type":239,"tunes":343},"UwXXyS_xnp",{"text":342},"Eine mathematisch gültige Transformation kann Evidenz nahezu perfekt bewahren, während eine einzige spekulative Interpretation einen viel größeren epistemischen Sprung erzeugen kann.",{},{"id":345,"data":346,"type":239,"tunes":348},"qOtlSqbZmt",{"text":347},"Das Konzept umfasst daher sowohl Distanz durch Transformation als auch Distanz durch Inferenz.",{},{"id":350,"data":351,"type":41,"tunes":353},"z0tE1scu1w",{"text":352,"level":233},"Direkte Antwort",{},{"id":355,"data":356,"type":239,"tunes":358},"kazpS6DqUn",{"text":357},"Die Evidenzdistanz nimmt zu, wenn eine Behauptung von mehr Transformationen, mehr Annahmen, schwächerer Provenienz oder stärkerer Inferenz abhängt, als die zugrunde liegende Evidenz direkt rechtfertigt.",{},{"id":360,"data":361,"type":280,"tunes":363},"64c0vwT55C",{"code":362},"Short distance:\nsource → claim\n\nLonger distance:\nsource → extraction → summary → assumption → inference → claim",{},{"id":365,"data":366,"type":239,"tunes":368},"MhOixLLuV7",{"text":367},"Eine größere Distanz macht eine Behauptung nicht automatisch falsch. Sie bedeutet, dass die Argumentationskette mehr Validierung erfordert.",{},{"id":370,"data":371,"type":41,"tunes":373},"wVsphcBSqw",{"text":372,"level":233},"Warum das für KI wichtig ist",{},{"id":375,"data":376,"type":239,"tunes":378},"MkFG6A0RsC",{"text":377},"LLMs sind besonders gut darin, flüssige Brücken zwischen Informationen zu erzeugen. Diese Fähigkeit ist nützlich, aber Flüssigkeit kann verbergen, wie weit sich die Schlussfolgerung von der direkt abgerufenen Evidenz entfernt hat.",{},{"id":380,"data":381,"type":239,"tunes":383},"4LXQyZ_nP2",{"text":382},"Eine Antwort kann daher gut belegt erscheinen, während sie Behauptungen enthält, deren Stützung nur durch mehrere unausgesprochene Inferenzschritte besteht.",{},{"id":385,"data":386,"type":393,"tunes":394},"d0XgOLAuI8",{"link":387,"meta":388},"https:\u002F\u002Fstajic.de\u002Fde\u002Fblog\u002Ffrom-research-protocol-to-a-general-ai-reasoning-framework",{"image":389,"title":391,"description":392},{"url":390},"","Vom Forschungsprotokoll zu einem allgemeinen KI-Reasoning-Framework","Zuverlässiges Schlussfolgern erfordert eine explizite Trennung von Evidenz, Annahmen, Hypothesen und Schlussfolgerungen.","linkTool",{},{"id":396,"data":397,"type":41,"tunes":399},"PjYzIFYT1U",{"text":398,"level":233},"Vier Komponenten der Evidenzdistanz",{},{"id":401,"data":402,"type":41,"tunes":404},"0x9QYuuH-l",{"text":403,"level":258},"1. Transformationsdistanz",{},{"id":406,"data":407,"type":239,"tunes":409},"gJ4DxD4oGt",{"text":408},"Wie oft wurde die ursprüngliche Evidenz konvertiert, zusammengefasst, normalisiert oder umformatiert, bevor sie das Modell erreicht hat?",{},{"id":411,"data":412,"type":41,"tunes":414},"MWNx-GiITU",{"text":413,"level":258},"2. Inferenzdistanz",{},{"id":416,"data":417,"type":239,"tunes":419},"9iJhSbpMrd",{"text":418},"Wie viele Argumentationsschritte sind erforderlich, bevor die Evidenz die endgültige Schlussfolgerung stützt?",{},{"id":421,"data":422,"type":41,"tunes":424},"FSrQHI2HdF",{"text":423,"level":258},"3. Provenienzdistanz",{},{"id":426,"data":427,"type":239,"tunes":429},"i_5MJVmw6r",{"text":428},"Kann die abgerufene Aussage noch auf eine identifizierbare Originalquelle zurückgeführt werden, oder ist sie durch Zusammenfassungen und Kopien gegangen, deren Ursprung unklar ist?",{},{"id":431,"data":432,"type":41,"tunes":434},"dhyb_9AYdc",{"text":433,"level":258},"4. Anwendbarkeitsdistanz",{},{"id":436,"data":437,"type":239,"tunes":439},"SDYDfH21Nz",{"text":438},"Wie viel Argumentation ist erforderlich, um festzustellen, dass die Evidenz auf genau diese Version, diesen Nutzer, diese Gerichtsbarkeit, diesen Zeitraum oder diese Situation zutrifft?",{},{"id":441,"data":442,"type":41,"tunes":444},"aonb2aApom",{"text":443,"level":233},"Evidenzdistanz in RAG",{},{"id":446,"data":447,"type":239,"tunes":449},"c5i9oX23SF",{"text":448},"RAG kann eine wichtige Distanz verkürzen: die Distanz zwischen einem Modell und externem Wissen.",{},{"id":451,"data":452,"type":239,"tunes":454},"bRgk5469WZ",{"text":453},"Es verkürzt nicht automatisch die Distanz zwischen einer abgerufenen Passage und der endgültigen Behauptung.",{},{"id":456,"data":457,"type":280,"tunes":459},"x6Ji-N-lry",{"code":458},"RAG solves:\nmodel ↔ external information\n\nRAG does not automatically solve:\nevidence ↔ justified conclusion",{},{"id":461,"data":462,"type":393,"tunes":468},"gRrOi5JDnn",{"link":463,"meta":464},"https:\u002F\u002Fstajic.de\u002Fde\u002Fblog\u002Fwhere-does-an-llm-get-its-data-rag-data-sources-in-python",{"image":465,"title":466,"description":467},{"url":390},"Woher bezieht ein LLM seine Daten? RAG-Datenquellen in Python","Retrieval bringt externe Informationen in den Kontext. Evidence Distance fragt, was zwischen diesen Informationen und der schließlich erzeugten Behauptung passiert.",{},{"id":470,"data":471,"type":41,"tunes":473},"Or0p1x40Fc",{"text":472,"level":233},"Verbindung zum Retrieval Trigger",{},{"id":475,"data":476,"type":239,"tunes":478},"H6s1-2OI7y",{"text":477},"Der Retrieval Trigger fragt, ob das interne Modellwissen ausreichend ist.",{},{"id":480,"data":481,"type":239,"tunes":483},"pbGaZ-mRzL",{"text":482},"Evidence Distance stellt eine spätere Frage: Wie direkt stützt es, nachdem Evidenz abgerufen wurde, das, was das System behaupten möchte?",{},{"id":485,"data":486,"type":280,"tunes":488},"XCegS6de5e",{"code":487},"Need evidence?\n   ↓\nRetrieval Trigger\n   ↓\nRetrieve\n   ↓\nHow directly does evidence support the claim?\n   ↓\nEvidence Distance",{},{"id":490,"data":491,"type":393,"tunes":497},"Diu2wdjbJz",{"link":492,"meta":493},"https:\u002F\u002Fstajic.de\u002Fde\u002Fblog\u002Fwhen-should-an-ai-stop-trusting-its-own-knowledge-the-retrieval-trigger",{"image":494,"title":495,"description":496},{"url":390},"Wann sollte eine KI aufhören, ihrem eigenen Wissen zu vertrauen? — Der Retrieval Trigger","Der Trigger entscheidet, wann Evidenz benötigt wird. Evidence Distance bewertet, wie weit die schließliche Behauptung über diese Evidenz hinausgeht.",{},{"id":499,"data":500,"type":41,"tunes":502},"dHdNuaHXOS",{"text":501,"level":233},"Eine praktische Diagnose",{},{"id":504,"data":505,"type":239,"tunes":507},"zxTVFZIVpw",{"text":506},"Rekonstruieren Sie für eine wichtige KI-generierte Behauptung die Stützkette rückwärts.",{},{"id":509,"data":510,"type":280,"tunes":512},"X8sRRGeav4",{"code":511},"CLAIM\n  ↑\nWhich inference produced it?\n  ↑\nWhich retrieved evidence supports that inference?\n  ↑\nWhich source produced that evidence?\n  ↑\nCan the original source still be inspected?",{},{"id":514,"data":515,"type":239,"tunes":517},"7irzpI0oMV",{"text":516},"Wenn die Kette bricht, ist das System auf ein Problem der Evidenznachverfolgbarkeit gestoßen.",{},{"id":519,"data":520,"type":239,"tunes":522},"kKNI-ESAqP",{"text":521},"Wenn die Kette intakt bleibt, aber mehrere starke Annahmen erfordert, ist die Evidence Distance immer noch hoch.",{},{"id":524,"data":525,"type":41,"tunes":527},"vn-vbQGoIC",{"text":526,"level":233},"Beispiel: Produktdaten",{},{"id":529,"data":530,"type":239,"tunes":532},"9bigaWbg_3",{"text":531},"Angenommen, eine Herstellerspezifikation besagt, dass ein Laptop 1,48 kg wiegt.",{},{"id":534,"data":535,"type":280,"tunes":537},"IfKrukuOp9",{"code":536},"Manufacturer specification\n→ 1.48 kg\n→ \"The laptop weighs 1.48 kg\"\n\nLow Evidence Distance",{},{"id":539,"data":540,"type":239,"tunes":542},"uIwRyMs4U8",{"text":541},"Betrachten Sie nun:",{},{"id":544,"data":545,"type":280,"tunes":547},"ih07pbVjtb",{"code":546},"Manufacturer specification\n→ 1.48 kg\n→ comparison with three competitors\n→ assumption about travel behavior\n→ \"This is the best laptop for frequent travelers\"\n\nMuch higher Evidence Distance",{},{"id":549,"data":550,"type":239,"tunes":552},"qKO4_aR6vR",{"text":551},"Die Spezifikation ist real. Die Empfehlung ist noch mehrere Argumentationsschritte davon entfernt.",{},{"id":554,"data":555,"type":41,"tunes":557},"90zOwAjmvt",{"text":556,"level":233},"Beispiel: Wissenschaftliche Evidenz",{},{"id":559,"data":560,"type":239,"tunes":562},"gju8RS56u4",{"text":561},"Eine Studie kann eine beobachtete Assoziation berichten. Eine KI-Antwort kann diese fälschlicherweise zu einer Kausalaussage verdichten.",{},{"id":564,"data":565,"type":280,"tunes":567},"LgAJrZoaBe",{"code":566},"Observed association\n        ↓\nInterpretation\n        ↓\nUnstated causal assumption\n        ↓\n\"X causes Y\"",{},{"id":569,"data":570,"type":239,"tunes":572},"-BnXVHW_7R",{"text":571},"Das Problem ist nicht, dass keine Evidenz existiert. Das Problem ist, dass die abschließende Aussage eine größere Evidenzdistanz überbrückt, als die Studie direkt stützt.",{},{"id":574,"data":575,"type":41,"tunes":577},"sKywq0xlRb",{"text":576,"level":233},"Verbindung zu Agent-Ready Context",{},{"id":579,"data":580,"type":239,"tunes":582},"a1-X0w-8Pn",{"text":581},"Strukturierte Informationsobjekte können vermeidbare Evidenzdistanz reduzieren, indem sie Quellenidentität, Einheiten, Zeitstempel, Konflikte und Gültigkeit bewahren, anstatt das Modell zu zwingen, sie aus abgeflachtem Text zu rekonstruieren.",{},{"id":584,"data":585,"type":393,"tunes":590},"EZywuvf7qq",{"link":586,"meta":587},"https:\u002F\u002Fstajic.de\u002Fde\u002Fportfolio\u002Ffrom-raw-data-to-verifiable-ai-context-a-practical-architecture-for-agent-ready-information",{"image":588,"title":177,"description":589},{"url":390},"Die Bewahrung von Herkunft und Gültigkeit vor dem Abruf reduziert die Anzahl der später erforderlichen ungestützten Rekonstruktionsschritte.",{},{"id":592,"data":593,"type":41,"tunes":595},"7JLsOQVOfL",{"text":594,"level":233},"Eine einfache Arbeitsskala",{},{"id":597,"data":598,"type":317,"tunes":618},"WLdbzx22ZH",{"content":599,"stretched":42,"withHeadings":13},[600,603,606,609,612,615],[601,602],"Stufe","Beschreibung",[604,605],"ED0","Direkt berichtete Quellentatsache",[607,608],"ED1","Einfache deterministische Transformation oder direkter Vergleich",[610,611],"ED2","Ein bedeutungsvoller interpretativer Schritt",[613,614],"ED3","Mehrere Annahmen oder kombinierte Quellen",[616,617],"ED4","Starke kausale, prädiktive oder präskriptive Schlussfolgerung über die direkte Evidenz hinaus",{},{"id":620,"data":621,"type":239,"tunes":623},"9FvtJjfCLQ",{"text":622},"Diese Skala ist illustrativ, kein wissenschaftlicher Standard. Ihr Zweck ist es, eine explizite Frage zu erzwingen: Wie viel mussten wir zwischen der Evidenz und der Behauptung hinzufügen?",{},{"id":625,"data":626,"type":41,"tunes":628},"6hCnJtbh1y",{"text":627,"level":233},"Missverständnisse",{},{"id":630,"data":631,"type":239,"tunes":633},"g0kaPh4HxV",{"text":632},"Eine Zitation bedeutet, dass die Evidenzdistanz gering ist. Nein. Eine Zitation kann neben einer Behauptung stehen, die das, was die zitierte Quelle belegt, erheblich überschreitet.",{},{"id":635,"data":636,"type":239,"tunes":638},"3vnmNlD5Mk",{"text":637},"Eine lange Evidenzdistanz bedeutet, dass die Behauptung falsch ist. Nein. Komplexe Schlussfolgerungen können gut gestützt sein, erfordern aber eine stärkere Argumentationskette.",{},{"id":640,"data":641,"type":239,"tunes":643},"nLnebZgawL",{"text":642},"Primärquellen eliminieren Inferenz. Nein. Primäre Evidenz kann dennoch Interpretation erfordern.",{},{"id":645,"data":646,"type":239,"tunes":648},"ypGitVy9x3",{"text":647},"Mehr Quellen reduzieren automatisch die Distanz. Zehn schwache Wiederholungen derselben ungestützten Behauptung erzeugen keine direkte Evidenz.",{},{"id":650,"data":651,"type":41,"tunes":653},"W-M4nnnzEg",{"text":652,"level":233},"Grenzfall: Abgeleitete Daten",{},{"id":655,"data":656,"type":239,"tunes":658},"YEzmltN56F",{"text":657},"Einige abgeleitete Werte sind zuverlässiger als rohe menschliche Beschreibungen. Eine korrekt berechnete Summe aus verifizierten Transaktionen kann äußerst verlässlich sein, obwohl sie nur eine Transformation von den Originalaufzeichnungen entfernt ist.",{},{"id":660,"data":661,"type":239,"tunes":663},"FB4n6LMnDW",{"text":662},"Die Evidenzdistanz misst daher die epistemische Trennung, nicht eine vereinfachte Vorliebe für unberührte Daten.",{},{"id":665,"data":666,"type":41,"tunes":668},"mge41Ld_cD",{"text":667,"level":233},"Einschränkungen",{},{"id":670,"data":671,"type":239,"tunes":673},"85Agxl7u5w",{"text":672},"Die Evidenzdistanz ist eine konzeptionelle Diagnose, kein universeller quantitativer Wert.",{},{"id":675,"data":676,"type":239,"tunes":678},"b-p7FbBY8Y",{"text":677},"Verschiedene Domänen erfordern unterschiedliche Behandlung von Inferenz. Mathematische Beweise, klinische Evidenz, Softwareprotokolle, Marktanalysen und historische Interpretation können keinen gemeinsamen universellen numerischen Schwellenwert teilen.",{},{"id":680,"data":681,"type":41,"tunes":683},"EksjnuYWwq",{"text":682,"level":233},"Was würde diese Antwort ändern?",{},{"id":685,"data":686,"type":239,"tunes":688},"gnIKkYxGVY",{"text":687},"Zukünftige KI-Systeme könnten explizite Provenienzgraphen, maschinenprüfbare Behauptungs-Evidenz-Verknüpfungen und typisierte Argumentationsspuren generieren. Das könnte Teile der Evidenzdistanz automatisch messbar machen.",{},{"id":690,"data":691,"type":239,"tunes":693},"NDNOC8F5BM",{"text":692},"Das zugrunde liegende Problem bleibt bestehen: Eine Behauptung kann sich weiter von ihrer Evidenz entfernen, selbst wenn jeder Zwischensatz plausibel klingt.",{},{"id":695,"data":696,"type":41,"tunes":698},"N9WxXZXU9K",{"text":697,"level":233},"Fazit",{},{"id":700,"data":701,"type":239,"tunes":703},"B0nu1n-lr8",{"text":702},"Die wichtige Unterscheidung liegt nicht nur zwischen belegten und unbelegten Antworten.",{},{"id":705,"data":706,"type":239,"tunes":708},"dUA5AMapMG",{"text":707},"Sie liegt zwischen Behauptungen, die nahe an direkt überprüfbarer Evidenz bleiben, und Behauptungen, die von zunehmend langen Ketten der Interpretation und Annahme abhängen.",{},{"id":710,"data":711,"type":280,"tunes":713},"YMLENvLaug",{"code":712},"Evidence\n   ↓\nEvidence Distance\n   ↓\nReasoning\n   ↓\nAnswer Validity Boundary",{},{"id":715,"data":716,"type":239,"tunes":718},"Z_Cu0srfiO",{"text":717},"Die Evidenzdistanz gibt uns eine Sprache, um diesen Raum zu beschreiben.",{},{"id":720,"data":721,"type":41,"tunes":723},"AUEeDCNKRA",{"text":722,"level":233},"Primärquellen",{},{"id":725,"data":726,"type":239,"tunes":728},"eHeLgVtehs",{"text":727},"Patrick Lewis et al., \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.11401\" target=\"_blank\">Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks\u003C\u002Fa> (2020).",{},{"id":730,"data":731,"type":239,"tunes":733},"i--E80s2nB",{"text":732},"Nelson F. Liu et al., \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2307.03172\" target=\"_blank\">Lost in the Middle: How Language Models Use Long Contexts\u003C\u002Fa> (2023).",{},{"id":735,"data":736,"type":239,"tunes":738},"U2ByVS6VcD",{"text":737},"W3C, \u003Ca href=\"https:\u002F\u002Fwww.w3.org\u002FTR\u002Fprov-overview\u002F\" target=\"_blank\">PROV Overview\u003C\u002Fa>. Ein standardisiertes Modell zur Darstellung von Provenienzbeziehungen.",{},"2.31.6","Nicht alle unterstützten Antworten stehen ihren Belegen gleich nahe. Die Evidenzdistanz ist eine praktische Möglichkeit zu beschreiben, wie viele Transformationen, Zusammenfassungen, Annahmen und Inferenzschritte eine KI-Behauptung von direkt überprüfbaren Belegen trennen.","\u002Fuploads\u002F2026\u002F10\u002Fhow-far-is-an-ai-claim-from-the-evidence-evidence-distance-1792095863209-3c1mog.webp","how-far-is-an-ai-claim-from-the-evidence-evidence-distance-1792095863209-3c1mog","PUBLISHED","2026-10-15T00:23:00.000Z","2026-10-15T20:23:08.634Z","2026-10-15T20:43:52.893Z",{"en":748,"de":749,"sr":750,"es":751,"fr":752,"it":753,"ru":754,"zh":755},"\u002Fblog\u002Fhow-far-is-an-ai-claim-from-the-evidence-evidence-distance","\u002Fde\u002Fblog\u002Fhow-far-is-an-ai-claim-from-the-evidence-evidence-distance","\u002Fsr\u002Fblog\u002Fhow-far-is-an-ai-claim-from-the-evidence-evidence-distance","\u002Fes\u002Fblog\u002Fhow-far-is-an-ai-claim-from-the-evidence-evidence-distance","\u002Ffr\u002Fblog\u002Fhow-far-is-an-ai-claim-from-the-evidence-evidence-distance","\u002Fit\u002Fblog\u002Fhow-far-is-an-ai-claim-from-the-evidence-evidence-distance","\u002Fru\u002Fblog\u002Fhow-far-is-an-ai-claim-from-the-evidence-evidence-distance","\u002Fzh\u002Fblog\u002Fhow-far-is-an-ai-claim-from-the-evidence-evidence-distance",[757,761],{"id":758,"name":759,"slug":760},68,"Risiken, Kontrollen & Nachweise","risks-and-controls",{"id":762,"name":763,"slug":764},49,"Kontrollen & Nachweise","controls",{"id":766,"login":767,"email":390,"displayName":768},"20","rooth8233","Aleksandar Stajić",[770,1063],{"lang":7,"title":223,"content":225,"contentJson":771,"excerpt":740},{"time":227,"blocks":772,"version":739},[773,776,779,782,785,788,791,794,797,800,803,806,809,817,820,823,826,829,832,835,838,841,844,847,850,853,856,861,864,867,870,873,876,879,882,885,888,891,894,897,900,905,908,911,914,917,922,925,928,931,934,937,940,943,946,949,952,955,958,961,964,967,970,973,978,981,991,994,997,1000,1003,1006,1009,1012,1015,1018,1021,1024,1027,1030,1033,1036,1039,1042,1045,1048,1051,1054,1057,1060],{"id":230,"data":774,"type":41,"tunes":775},{"text":232,"level":233},{},{"id":236,"data":777,"type":239,"tunes":778},{"text":238},{},{"id":242,"data":780,"type":239,"tunes":781},{"text":244},{},{"id":247,"data":783,"type":252,"tunes":784},{"text":249,"caption":250,"alignment":251},{},{"id":255,"data":786,"type":259,"tunes":787},{"title":257,"maxLevel":258,"minLevel":233},{},{"id":262,"data":789,"type":41,"tunes":790},{"text":264,"level":233},{},{"id":267,"data":792,"type":239,"tunes":793},{"text":269},{},{"id":272,"data":795,"type":239,"tunes":796},{"text":274},{},{"id":277,"data":798,"type":280,"tunes":799},{"code":279},{},{"id":283,"data":801,"type":239,"tunes":802},{"text":285},{},{"id":288,"data":804,"type":41,"tunes":805},{"text":290,"level":233},{},{"id":293,"data":807,"type":239,"tunes":808},{"text":295},{},{"id":298,"data":810,"type":317,"tunes":816},{"content":811,"stretched":42,"withHeadings":13},[812,813,814,815],[302,303,304],[306,307,308],[310,311,312],[314,315,316],{},{"id":320,"data":818,"type":239,"tunes":819},{"text":322},{},{"id":325,"data":821,"type":280,"tunes":822},{"code":327},{},{"id":330,"data":824,"type":41,"tunes":825},{"text":332,"level":233},{},{"id":335,"data":827,"type":239,"tunes":828},{"text":337},{},{"id":340,"data":830,"type":239,"tunes":831},{"text":342},{},{"id":345,"data":833,"type":239,"tunes":834},{"text":347},{},{"id":350,"data":836,"type":41,"tunes":837},{"text":352,"level":233},{},{"id":355,"data":839,"type":239,"tunes":840},{"text":357},{},{"id":360,"data":842,"type":280,"tunes":843},{"code":362},{},{"id":365,"data":845,"type":239,"tunes":846},{"text":367},{},{"id":370,"data":848,"type":41,"tunes":849},{"text":372,"level":233},{},{"id":375,"data":851,"type":239,"tunes":852},{"text":377},{},{"id":380,"data":854,"type":239,"tunes":855},{"text":382},{},{"id":385,"data":857,"type":393,"tunes":860},{"link":387,"meta":858},{"image":859,"title":391,"description":392},{"url":390},{},{"id":396,"data":862,"type":41,"tunes":863},{"text":398,"level":233},{},{"id":401,"data":865,"type":41,"tunes":866},{"text":403,"level":258},{},{"id":406,"data":868,"type":239,"tunes":869},{"text":408},{},{"id":411,"data":871,"type":41,"tunes":872},{"text":413,"level":258},{},{"id":416,"data":874,"type":239,"tunes":875},{"text":418},{},{"id":421,"data":877,"type":41,"tunes":878},{"text":423,"level":258},{},{"id":426,"data":880,"type":239,"tunes":881},{"text":428},{},{"id":431,"data":883,"type":41,"tunes":884},{"text":433,"level":258},{},{"id":436,"data":886,"type":239,"tunes":887},{"text":438},{},{"id":441,"data":889,"type":41,"tunes":890},{"text":443,"level":233},{},{"id":446,"data":892,"type":239,"tunes":893},{"text":448},{},{"id":451,"data":895,"type":239,"tunes":896},{"text":453},{},{"id":456,"data":898,"type":280,"tunes":899},{"code":458},{},{"id":461,"data":901,"type":393,"tunes":904},{"link":463,"meta":902},{"image":903,"title":466,"description":467},{"url":390},{},{"id":470,"data":906,"type":41,"tunes":907},{"text":472,"level":233},{},{"id":475,"data":909,"type":239,"tunes":910},{"text":477},{},{"id":480,"data":912,"type":239,"tunes":913},{"text":482},{},{"id":485,"data":915,"type":280,"tunes":916},{"code":487},{},{"id":490,"data":918,"type":393,"tunes":921},{"link":492,"meta":919},{"image":920,"title":495,"description":496},{"url":390},{},{"id":499,"data":923,"type":41,"tunes":924},{"text":501,"level":233},{},{"id":504,"data":926,"type":239,"tunes":927},{"text":506},{},{"id":509,"data":929,"type":280,"tunes":930},{"code":511},{},{"id":514,"data":932,"type":239,"tunes":933},{"text":516},{},{"id":519,"data":935,"type":239,"tunes":936},{"text":521},{},{"id":524,"data":938,"type":41,"tunes":939},{"text":526,"level":233},{},{"id":529,"data":941,"type":239,"tunes":942},{"text":531},{},{"id":534,"data":944,"type":280,"tunes":945},{"code":536},{},{"id":539,"data":947,"type":239,"tunes":948},{"text":541},{},{"id":544,"data":950,"type":280,"tunes":951},{"code":546},{},{"id":549,"data":953,"type":239,"tunes":954},{"text":551},{},{"id":554,"data":956,"type":41,"tunes":957},{"text":556,"level":233},{},{"id":559,"data":959,"type":239,"tunes":960},{"text":561},{},{"id":564,"data":962,"type":280,"tunes":963},{"code":566},{},{"id":569,"data":965,"type":239,"tunes":966},{"text":571},{},{"id":574,"data":968,"type":41,"tunes":969},{"text":576,"level":233},{},{"id":579,"data":971,"type":239,"tunes":972},{"text":581},{},{"id":584,"data":974,"type":393,"tunes":977},{"link":586,"meta":975},{"image":976,"title":177,"description":589},{"url":390},{},{"id":592,"data":979,"type":41,"tunes":980},{"text":594,"level":233},{},{"id":597,"data":982,"type":317,"tunes":990},{"content":983,"stretched":42,"withHeadings":13},[984,985,986,987,988,989],[601,602],[604,605],[607,608],[610,611],[613,614],[616,617],{},{"id":620,"data":992,"type":239,"tunes":993},{"text":622},{},{"id":625,"data":995,"type":41,"tunes":996},{"text":627,"level":233},{},{"id":630,"data":998,"type":239,"tunes":999},{"text":632},{},{"id":635,"data":1001,"type":239,"tunes":1002},{"text":637},{},{"id":640,"data":1004,"type":239,"tunes":1005},{"text":642},{},{"id":645,"data":1007,"type":239,"tunes":1008},{"text":647},{},{"id":650,"data":1010,"type":41,"tunes":1011},{"text":652,"level":233},{},{"id":655,"data":1013,"type":239,"tunes":1014},{"text":657},{},{"id":660,"data":1016,"type":239,"tunes":1017},{"text":662},{},{"id":665,"data":1019,"type":41,"tunes":1020},{"text":667,"level":233},{},{"id":670,"data":1022,"type":239,"tunes":1023},{"text":672},{},{"id":675,"data":1025,"type":239,"tunes":1026},{"text":677},{},{"id":680,"data":1028,"type":41,"tunes":1029},{"text":682,"level":233},{},{"id":685,"data":1031,"type":239,"tunes":1032},{"text":687},{},{"id":690,"data":1034,"type":239,"tunes":1035},{"text":692},{},{"id":695,"data":1037,"type":41,"tunes":1038},{"text":697,"level":233},{},{"id":700,"data":1040,"type":239,"tunes":1041},{"text":702},{},{"id":705,"data":1043,"type":239,"tunes":1044},{"text":707},{},{"id":710,"data":1046,"type":280,"tunes":1047},{"code":712},{},{"id":715,"data":1049,"type":239,"tunes":1050},{"text":717},{},{"id":720,"data":1052,"type":41,"tunes":1053},{"text":722,"level":233},{},{"id":725,"data":1055,"type":239,"tunes":1056},{"text":727},{},{"id":730,"data":1058,"type":239,"tunes":1059},{"text":732},{},{"id":735,"data":1061,"type":239,"tunes":1062},{"text":737},{},{"lang":1064,"title":1065,"content":1066,"contentJson":1067,"excerpt":1553},"en","How Far Is an AI Claim From the Evidence? — Evidence Distance","{\"time\":1792096795538,\"blocks\":[{\"id\":\"g-2Ciyj8LJ\",\"type\":\"header\",\"data\":{\"text\":\"Question\",\"level\":2},\"tunes\":{}},{\"id\":\"SQWD7UraiX\",\"type\":\"paragraph\",\"data\":{\"text\":\"Two AI answers can both cite sources and still differ greatly in how directly those sources support the claims being made. How can we describe that difference?\"},\"tunes\":{}},{\"id\":\"Sj1fx2865j\",\"type\":\"paragraph\",\"data\":{\"text\":\"This article uses the term Evidence Distance for the distance between a claim and the directly inspectable evidence that supports it.\"},\"tunes\":{}},{\"id\":\"d5th-A-mq5\",\"type\":\"quote\",\"data\":{\"text\":\"Evidence Distance is the number and strength of transformations, assumptions and inferential steps separating a claim from the evidence that can directly verify it.\",\"caption\":\"Working definition — a reasoning concept introduced here, not a standardized metric.\",\"alignment\":\"left\"},\"tunes\":{}},{\"id\":\"X81jXC72FC\",\"type\":\"tableOfContents\",\"data\":{\"title\":\"Contents\",\"minLevel\":2,\"maxLevel\":3},\"tunes\":{}},{\"id\":\"_pZcnPEoil\",\"type\":\"header\",\"data\":{\"text\":\"What This Really Means\",\"level\":2},\"tunes\":{}},{\"id\":\"F5odO7xyha\",\"type\":\"paragraph\",\"data\":{\"text\":\"A citation proves that a source was referenced. It does not automatically prove that the source directly supports the sentence beside the citation.\"},\"tunes\":{}},{\"id\":\"8Gqlg-ArJS\",\"type\":\"paragraph\",\"data\":{\"text\":\"Between an original observation and a final AI answer there may be extraction, parsing, chunking, retrieval, summarization, interpretation, aggregation and inference.\"},\"tunes\":{}},{\"id\":\"lXkL5LV9ab\",\"type\":\"code\",\"data\":{\"code\":\"Primary evidence\\n   ↓\\nExtracted data\\n   ↓\\nRetrieved passage\\n   ↓\\nSummary\\n   ↓\\nInterpretation\\n   ↓\\nInference\\n   ↓\\nFinal claim\"},\"tunes\":{}},{\"id\":\"Z-8GPQIpCI\",\"type\":\"paragraph\",\"data\":{\"text\":\"Every step can be legitimate. But every step can also introduce distortion.\"},\"tunes\":{}},{\"id\":\"aXMwTGZ8Ia\",\"type\":\"header\",\"data\":{\"text\":\"Simplest Example\",\"level\":2},\"tunes\":{}},{\"id\":\"bZlU7KD3M2\",\"type\":\"paragraph\",\"data\":{\"text\":\"Suppose a temperature sensor reports 38.2 °C.\"},\"tunes\":{}},{\"id\":\"SP87UUPaU3\",\"type\":\"table\",\"data\":{\"withHeadings\":true,\"stretched\":false,\"content\":[[\"Claim\",\"Relationship to evidence\",\"Illustrative distance\"],[\"The sensor reports 38.2 °C.\",\"Direct reading\",\"Low\"],[\"The system is running hotter than yesterday.\",\"Requires comparison with yesterday's reading\",\"Moderate\"],[\"The cooling system is failing.\",\"Requires causal inference and alternative explanations\",\"High\"]]},\"tunes\":{}},{\"id\":\"eYar059d69\",\"type\":\"paragraph\",\"data\":{\"text\":\"The original sensor value directly supports the first claim. It does not by itself prove the third.\"},\"tunes\":{}},{\"id\":\"eCnFlByvHt\",\"type\":\"code\",\"data\":{\"code\":\"Evidence exists\\n        ≠\\nEvidence directly supports every derived claim\"},\"tunes\":{}},{\"id\":\"aXGkK6BpkF\",\"type\":\"header\",\"data\":{\"text\":\"Where the Example Stops Working\",\"level\":2},\"tunes\":{}},{\"id\":\"lAwb9QZ_yB\",\"type\":\"paragraph\",\"data\":{\"text\":\"Evidence Distance is not simply a count of sentences or processing steps.\"},\"tunes\":{}},{\"id\":\"-SzYko9oIH\",\"type\":\"paragraph\",\"data\":{\"text\":\"One mathematically valid transformation may preserve evidence almost perfectly, while a single speculative interpretation may create a much larger epistemic jump.\"},\"tunes\":{}},{\"id\":\"XOQloBndLw\",\"type\":\"paragraph\",\"data\":{\"text\":\"The concept therefore includes both distance by transformation and distance by inference.\"},\"tunes\":{}},{\"id\":\"EOXbS4TAOR\",\"type\":\"header\",\"data\":{\"text\":\"Direct Answer\",\"level\":2},\"tunes\":{}},{\"id\":\"HcVL8vGM74\",\"type\":\"paragraph\",\"data\":{\"text\":\"Evidence Distance increases when a claim depends on more transformations, more assumptions, weaker provenance or stronger inference than the underlying evidence directly warrants.\"},\"tunes\":{}},{\"id\":\"_f7WXjUOsS\",\"type\":\"code\",\"data\":{\"code\":\"Short distance:\\nsource → claim\\n\\nLonger distance:\\nsource → extraction → summary → assumption → inference → claim\"},\"tunes\":{}},{\"id\":\"zUEC-cDhqm\",\"type\":\"paragraph\",\"data\":{\"text\":\"A longer distance does not automatically make a claim false. It means the reasoning chain requires more validation.\"},\"tunes\":{}},{\"id\":\"ix7TUGNz_W\",\"type\":\"header\",\"data\":{\"text\":\"Why This Matters for AI\",\"level\":2},\"tunes\":{}},{\"id\":\"wrAy5PLtLB\",\"type\":\"paragraph\",\"data\":{\"text\":\"LLMs are particularly good at producing fluent bridges between pieces of information. That capability is useful, but fluency can hide how far the conclusion has moved beyond the directly retrieved evidence.\"},\"tunes\":{}},{\"id\":\"jJseLuTWdv\",\"type\":\"paragraph\",\"data\":{\"text\":\"An answer can therefore look well sourced while containing claims whose support exists only through several unstated inferential steps.\"},\"tunes\":{}},{\"id\":\"ox8fok9dzs\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fstajic.de\u002Fblog\u002Ffrom-research-protocol-to-a-general-ai-reasoning-framework\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"From Research Protocol to a General AI Reasoning Framework\",\"description\":\"Reliable reasoning requires explicit separation of evidence, assumptions, hypotheses and conclusions.\"}},\"tunes\":{}},{\"id\":\"6ipse6YG1U\",\"type\":\"header\",\"data\":{\"text\":\"Four Components of Evidence Distance\",\"level\":2},\"tunes\":{}},{\"id\":\"P4xQ7o96dR\",\"type\":\"header\",\"data\":{\"text\":\"1. Transformation Distance\",\"level\":3},\"tunes\":{}},{\"id\":\"dvN8TUNEYG\",\"type\":\"paragraph\",\"data\":{\"text\":\"How many times has the original evidence been converted, summarized, normalized or reformatted before reaching the model?\"},\"tunes\":{}},{\"id\":\"dXofACF9_x\",\"type\":\"header\",\"data\":{\"text\":\"2. Inferential Distance\",\"level\":3},\"tunes\":{}},{\"id\":\"kkEzV3ydId\",\"type\":\"paragraph\",\"data\":{\"text\":\"How many reasoning steps are required before the evidence supports the final conclusion?\"},\"tunes\":{}},{\"id\":\"PPYB0RuX5_\",\"type\":\"header\",\"data\":{\"text\":\"3. Provenance Distance\",\"level\":3},\"tunes\":{}},{\"id\":\"cAVBYebJr7\",\"type\":\"paragraph\",\"data\":{\"text\":\"Can the retrieved statement still be traced to an identifiable original source, or has it passed through summaries and copies whose origin is unclear?\"},\"tunes\":{}},{\"id\":\"254N_99Z7X\",\"type\":\"header\",\"data\":{\"text\":\"4. Applicability Distance\",\"level\":3},\"tunes\":{}},{\"id\":\"to3KWv5WSK\",\"type\":\"paragraph\",\"data\":{\"text\":\"How much reasoning is required to establish that the evidence applies to this exact version, user, jurisdiction, time period or situation?\"},\"tunes\":{}},{\"id\":\"3WnVPQ6wqw\",\"type\":\"header\",\"data\":{\"text\":\"Evidence Distance in RAG\",\"level\":2},\"tunes\":{}},{\"id\":\"r8s-75EBhA\",\"type\":\"paragraph\",\"data\":{\"text\":\"RAG can shorten one important distance: the distance between a model and external knowledge.\"},\"tunes\":{}},{\"id\":\"I_8h52A3Yc\",\"type\":\"paragraph\",\"data\":{\"text\":\"It does not automatically shorten the distance between a retrieved passage and the final claim.\"},\"tunes\":{}},{\"id\":\"jDDRUu2FUK\",\"type\":\"code\",\"data\":{\"code\":\"RAG solves:\\nmodel ↔ external information\\n\\nRAG does not automatically solve:\\nevidence ↔ justified conclusion\"},\"tunes\":{}},{\"id\":\"rm14qIVB1M\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fstajic.de\u002Fblog\u002Fwhere-does-an-llm-get-its-data-rag-data-sources-in-python\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"Where Does an LLM Get Its Data? RAG Data Sources in Python\",\"description\":\"Retrieval brings external information into context. Evidence Distance asks what happens between that information and the claim eventually produced.\"}},\"tunes\":{}},{\"id\":\"CRUZDXYloa\",\"type\":\"header\",\"data\":{\"text\":\"Connection to the Retrieval Trigger\",\"level\":2},\"tunes\":{}},{\"id\":\"-TPiYJ569c\",\"type\":\"paragraph\",\"data\":{\"text\":\"The Retrieval Trigger asks whether internal model knowledge is sufficient.\"},\"tunes\":{}},{\"id\":\"_VjLi5fs4x\",\"type\":\"paragraph\",\"data\":{\"text\":\"Evidence Distance asks a later question: after evidence has been retrieved, how directly does it support what the system wants to claim?\"},\"tunes\":{}},{\"id\":\"EImPw6i617\",\"type\":\"code\",\"data\":{\"code\":\"Need evidence?\\n   ↓\\nRetrieval Trigger\\n   ↓\\nRetrieve\\n   ↓\\nHow directly does evidence support the claim?\\n   ↓\\nEvidence Distance\"},\"tunes\":{}},{\"id\":\"ZqyhB2bFtR\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fstajic.de\u002Fblog\u002Fwhen-should-an-ai-stop-trusting-its-own-knowledge-the-retrieval-trigger\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"When Should an AI Stop Trusting Its Own Knowledge? — The Retrieval Trigger\",\"description\":\"The trigger decides when evidence is needed. Evidence Distance evaluates how far the eventual claim moves beyond that evidence.\"}},\"tunes\":{}},{\"id\":\"ST2vnm6vwF\",\"type\":\"header\",\"data\":{\"text\":\"A Practical Diagnostic\",\"level\":2},\"tunes\":{}},{\"id\":\"qyoHcDcTKb\",\"type\":\"paragraph\",\"data\":{\"text\":\"For an important AI-generated claim, reconstruct the support chain backwards.\"},\"tunes\":{}},{\"id\":\"e3fDmLk0fe\",\"type\":\"code\",\"data\":{\"code\":\"CLAIM\\n  ↑\\nWhich inference produced it?\\n  ↑\\nWhich retrieved evidence supports that inference?\\n  ↑\\nWhich source produced that evidence?\\n  ↑\\nCan the original source still be inspected?\"},\"tunes\":{}},{\"id\":\"iQ60c7NRaE\",\"type\":\"paragraph\",\"data\":{\"text\":\"If the chain breaks, the system has encountered an evidence-traceability problem.\"},\"tunes\":{}},{\"id\":\"YQAKlNZ2gM\",\"type\":\"paragraph\",\"data\":{\"text\":\"If the chain remains intact but requires several strong assumptions, the Evidence Distance is still high.\"},\"tunes\":{}},{\"id\":\"clN7jFkUBf\",\"type\":\"header\",\"data\":{\"text\":\"Example: Product Data\",\"level\":2},\"tunes\":{}},{\"id\":\"CRF-A9O7Ot\",\"type\":\"paragraph\",\"data\":{\"text\":\"Suppose a manufacturer specification says a laptop weighs 1.48 kg.\"},\"tunes\":{}},{\"id\":\"JF3z6aF2q_\",\"type\":\"code\",\"data\":{\"code\":\"Manufacturer specification\\n→ 1.48 kg\\n→ \\\"The laptop weighs 1.48 kg\\\"\\n\\nLow Evidence Distance\"},\"tunes\":{}},{\"id\":\"vTMOOgkLPb\",\"type\":\"paragraph\",\"data\":{\"text\":\"Now consider:\"},\"tunes\":{}},{\"id\":\"Tuhfzor1yp\",\"type\":\"code\",\"data\":{\"code\":\"Manufacturer specification\\n→ 1.48 kg\\n→ comparison with three competitors\\n→ assumption about travel behavior\\n→ \\\"This is the best laptop for frequent travelers\\\"\\n\\nMuch higher Evidence Distance\"},\"tunes\":{}},{\"id\":\"JHFRGIgUJD\",\"type\":\"paragraph\",\"data\":{\"text\":\"The specification is real. The recommendation is still several reasoning steps away from it.\"},\"tunes\":{}},{\"id\":\"2lKERCrnW1\",\"type\":\"header\",\"data\":{\"text\":\"Example: Scientific Evidence\",\"level\":2},\"tunes\":{}},{\"id\":\"HA8DBYwUuK\",\"type\":\"paragraph\",\"data\":{\"text\":\"A paper may report an observed association. An AI answer can incorrectly compress that into a causal statement.\"},\"tunes\":{}},{\"id\":\"D9C2dVB7j_\",\"type\":\"code\",\"data\":{\"code\":\"Observed association\\n        ↓\\nInterpretation\\n        ↓\\nUnstated causal assumption\\n        ↓\\n\\\"X causes Y\\\"\"},\"tunes\":{}},{\"id\":\"CKNjK25pkL\",\"type\":\"paragraph\",\"data\":{\"text\":\"The problem is not that no evidence exists. The problem is that the final statement crosses a larger evidential distance than the study directly supports.\"},\"tunes\":{}},{\"id\":\"sR430N3cDd\",\"type\":\"header\",\"data\":{\"text\":\"Connection to Agent-Ready Context\",\"level\":2},\"tunes\":{}},{\"id\":\"v77YyG55ko\",\"type\":\"paragraph\",\"data\":{\"text\":\"Structured information objects can reduce avoidable Evidence Distance by preserving source identity, units, timestamps, conflicts and validity instead of forcing the model to reconstruct them from flattened text.\"},\"tunes\":{}},{\"id\":\"EX8darCIjd\",\"type\":\"linkTool\",\"data\":{\"link\":\"https:\u002F\u002Fstajic.de\u002Fportfolio\u002Ffrom-raw-data-to-verifiable-ai-context-a-practical-architecture-for-agent-ready-information\",\"meta\":{\"image\":{\"url\":\"\"},\"title\":\"From Raw Data to Verifiable AI Context: A Practical Architecture for Agent-Ready Information\",\"description\":\"Preserving provenance and validity before retrieval reduces the number of unsupported reconstruction steps required later.\"}},\"tunes\":{}},{\"id\":\"dsfuoURj1e\",\"type\":\"header\",\"data\":{\"text\":\"A Simple Working Scale\",\"level\":2},\"tunes\":{}},{\"id\":\"BLi7Yth64O\",\"type\":\"table\",\"data\":{\"withHeadings\":true,\"stretched\":false,\"content\":[[\"Level\",\"Description\"],[\"ED0\",\"Directly reported source fact\"],[\"ED1\",\"Simple deterministic transformation or direct comparison\"],[\"ED2\",\"One meaningful interpretive step\"],[\"ED3\",\"Multiple assumptions or combined sources\"],[\"ED4\",\"Strong causal, predictive or prescriptive conclusion beyond direct evidence\"]]},\"tunes\":{}},{\"id\":\"8y8bzCcHBs\",\"type\":\"paragraph\",\"data\":{\"text\":\"This scale is illustrative, not a scientific standard. Its purpose is to force an explicit question: How much did we have to add between the evidence and the claim?\"},\"tunes\":{}},{\"id\":\"aFK0YiAlYO\",\"type\":\"header\",\"data\":{\"text\":\"Misconceptions\",\"level\":2},\"tunes\":{}},{\"id\":\"07KSk6xDx-\",\"type\":\"paragraph\",\"data\":{\"text\":\"A citation means Evidence Distance is low. No. A citation can sit beside a claim that substantially exceeds what the cited source establishes.\"},\"tunes\":{}},{\"id\":\"t_cf-Zfan9\",\"type\":\"paragraph\",\"data\":{\"text\":\"Long Evidence Distance means the claim is wrong. No. Complex conclusions can be well supported, but they require a stronger reasoning chain.\"},\"tunes\":{}},{\"id\":\"Pu9PRX67Q6\",\"type\":\"paragraph\",\"data\":{\"text\":\"Primary sources eliminate inference. No. Primary evidence can still require interpretation.\"},\"tunes\":{}},{\"id\":\"oF1xNsOHHq\",\"type\":\"paragraph\",\"data\":{\"text\":\"More sources automatically reduce distance. Ten weak repetitions of the same unsupported claim do not create direct evidence.\"},\"tunes\":{}},{\"id\":\"IkolK5baph\",\"type\":\"header\",\"data\":{\"text\":\"Edge Case: Derived Data\",\"level\":2},\"tunes\":{}},{\"id\":\"jRH36fwdR3\",\"type\":\"paragraph\",\"data\":{\"text\":\"Some derived values are more reliable than raw human descriptions. A correctly calculated total from verified transactions can be highly dependable even though it is one transformation away from the original records.\"},\"tunes\":{}},{\"id\":\"SXbdKiNDst\",\"type\":\"paragraph\",\"data\":{\"text\":\"Evidence Distance therefore measures epistemic separation, not a simplistic preference for untouched data.\"},\"tunes\":{}},{\"id\":\"mUbvdnFChM\",\"type\":\"header\",\"data\":{\"text\":\"Limitations\",\"level\":2},\"tunes\":{}},{\"id\":\"N_EskxFOep\",\"type\":\"paragraph\",\"data\":{\"text\":\"Evidence Distance is a conceptual diagnostic, not a universal quantitative score.\"},\"tunes\":{}},{\"id\":\"nW_-E08So_\",\"type\":\"paragraph\",\"data\":{\"text\":\"Different domains require different treatment of inference. Mathematical proof, clinical evidence, software logs, market analysis and historical interpretation cannot share one universal numerical threshold.\"},\"tunes\":{}},{\"id\":\"_Bd95wXkri\",\"type\":\"header\",\"data\":{\"text\":\"What Would Change This Answer?\",\"level\":2},\"tunes\":{}},{\"id\":\"OqCaMc2_OX\",\"type\":\"paragraph\",\"data\":{\"text\":\"Future AI systems may generate explicit provenance graphs, machine-checkable claim-evidence links and typed reasoning traces. That could make parts of Evidence Distance measurable automatically.\"},\"tunes\":{}},{\"id\":\"58Hf3O9-Iz\",\"type\":\"paragraph\",\"data\":{\"text\":\"The underlying problem remains: a claim can move farther from its evidence even while every intermediate sentence sounds plausible.\"},\"tunes\":{}},{\"id\":\"8WOq5oE_e_\",\"type\":\"header\",\"data\":{\"text\":\"Conclusion\",\"level\":2},\"tunes\":{}},{\"id\":\"1pr1TfwnKc\",\"type\":\"paragraph\",\"data\":{\"text\":\"The important distinction is not merely between sourced and unsourced answers.\"},\"tunes\":{}},{\"id\":\"54CysT_NXa\",\"type\":\"paragraph\",\"data\":{\"text\":\"It is between claims that remain close to directly inspectable evidence and claims that depend on increasingly long chains of interpretation and assumption.\"},\"tunes\":{}},{\"id\":\"Qrjo3WEUSL\",\"type\":\"code\",\"data\":{\"code\":\"Evidence\\n   ↓\\nEvidence Distance\\n   ↓\\nReasoning\\n   ↓\\nAnswer Validity Boundary\"},\"tunes\":{}},{\"id\":\"6ir8aIAfd3\",\"type\":\"paragraph\",\"data\":{\"text\":\"Evidence Distance gives us a language for describing that space.\"},\"tunes\":{}},{\"id\":\"vPhupXcbeN\",\"type\":\"header\",\"data\":{\"text\":\"Primary Sources\",\"level\":2},\"tunes\":{}},{\"id\":\"BaO4JmWRQL\",\"type\":\"paragraph\",\"data\":{\"text\":\"Patrick Lewis et al., \u003Ca href=\\\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2005.11401\\\" target=\\\"_blank\\\">Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks\u003C\u002Fa> (2020).\"},\"tunes\":{}},{\"id\":\"eg4Z4G6Cgs\",\"type\":\"paragraph\",\"data\":{\"text\":\"Nelson F. Liu et al., \u003Ca href=\\\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2307.03172\\\" target=\\\"_blank\\\">Lost in the Middle: How Language Models Use Long Contexts\u003C\u002Fa> (2023).\"},\"tunes\":{}},{\"id\":\"BJH7zUd4T8\",\"type\":\"paragraph\",\"data\":{\"text\":\"W3C, \u003Ca href=\\\"https:\u002F\u002Fwww.w3.org\u002FTR\u002Fprov-overview\u002F\\\" target=\\\"_blank\\\">PROV Overview\u003C\u002Fa>. A standardized model for representing provenance relationships.\"},\"tunes\":{}}],\"version\":\"2.31.6\"}",{"time":1068,"blocks":1069,"version":739},1792096795538,[1070,1075,1080,1085,1091,1095,1100,1105,1110,1114,1119,1124,1129,1150,1155,1159,1164,1169,1174,1179,1184,1189,1193,1198,1203,1208,1213,1222,1227,1232,1237,1242,1247,1252,1257,1262,1267,1272,1277,1282,1286,1295,1300,1305,1310,1314,1323,1328,1333,1337,1342,1347,1352,1357,1361,1366,1370,1375,1380,1385,1389,1394,1399,1404,1413,1418,1436,1441,1446,1451,1456,1461,1466,1471,1476,1481,1486,1491,1496,1501,1506,1511,1516,1521,1526,1530,1535,1540,1544,1548],{"id":1071,"data":1072,"type":41,"tunes":1074},"g-2Ciyj8LJ",{"text":1073,"level":233},"Question",{},{"id":1076,"data":1077,"type":239,"tunes":1079},"SQWD7UraiX",{"text":1078},"Two AI answers can both cite sources and still differ greatly in how directly those sources support the claims being made. How can we describe that difference?",{},{"id":1081,"data":1082,"type":239,"tunes":1084},"Sj1fx2865j",{"text":1083},"This article uses the term Evidence Distance for the distance between a claim and the directly inspectable evidence that supports it.",{},{"id":1086,"data":1087,"type":252,"tunes":1090},"d5th-A-mq5",{"text":1088,"caption":1089,"alignment":251},"Evidence Distance is the number and strength of transformations, assumptions and inferential steps separating a claim from the evidence that can directly verify it.","Working definition — a reasoning concept introduced here, not a standardized metric.",{},{"id":1092,"data":1093,"type":259,"tunes":1094},"X81jXC72FC",{"title":257,"maxLevel":258,"minLevel":233},{},{"id":1096,"data":1097,"type":41,"tunes":1099},"_pZcnPEoil",{"text":1098,"level":233},"What This Really Means",{},{"id":1101,"data":1102,"type":239,"tunes":1104},"F5odO7xyha",{"text":1103},"A citation proves that a source was referenced. It does not automatically prove that the source directly supports the sentence beside the citation.",{},{"id":1106,"data":1107,"type":239,"tunes":1109},"8Gqlg-ArJS",{"text":1108},"Between an original observation and a final AI answer there may be extraction, parsing, chunking, retrieval, summarization, interpretation, aggregation and inference.",{},{"id":1111,"data":1112,"type":280,"tunes":1113},"lXkL5LV9ab",{"code":279},{},{"id":1115,"data":1116,"type":239,"tunes":1118},"Z-8GPQIpCI",{"text":1117},"Every step can be legitimate. But every step can also introduce distortion.",{},{"id":1120,"data":1121,"type":41,"tunes":1123},"aXMwTGZ8Ia",{"text":1122,"level":233},"Simplest Example",{},{"id":1125,"data":1126,"type":239,"tunes":1128},"bZlU7KD3M2",{"text":1127},"Suppose a temperature sensor reports 38.2 °C.",{},{"id":1130,"data":1131,"type":317,"tunes":1149},"SP87UUPaU3",{"content":1132,"stretched":42,"withHeadings":13},[1133,1137,1141,1145],[1134,1135,1136],"Claim","Relationship to evidence","Illustrative distance",[1138,1139,1140],"The sensor reports 38.2 °C.","Direct reading","Low",[1142,1143,1144],"The system is running hotter than yesterday.","Requires comparison with yesterday's reading","Moderate",[1146,1147,1148],"The cooling system is failing.","Requires causal inference and alternative explanations","High",{},{"id":1151,"data":1152,"type":239,"tunes":1154},"eYar059d69",{"text":1153},"The original sensor value directly supports the first claim. It does not by itself prove the third.",{},{"id":1156,"data":1157,"type":280,"tunes":1158},"eCnFlByvHt",{"code":327},{},{"id":1160,"data":1161,"type":41,"tunes":1163},"aXGkK6BpkF",{"text":1162,"level":233},"Where the Example Stops Working",{},{"id":1165,"data":1166,"type":239,"tunes":1168},"lAwb9QZ_yB",{"text":1167},"Evidence Distance is not simply a count of sentences or processing steps.",{},{"id":1170,"data":1171,"type":239,"tunes":1173},"-SzYko9oIH",{"text":1172},"One mathematically valid transformation may preserve evidence almost perfectly, while a single speculative interpretation may create a much larger epistemic jump.",{},{"id":1175,"data":1176,"type":239,"tunes":1178},"XOQloBndLw",{"text":1177},"The concept therefore includes both distance by transformation and distance by inference.",{},{"id":1180,"data":1181,"type":41,"tunes":1183},"EOXbS4TAOR",{"text":1182,"level":233},"Direct Answer",{},{"id":1185,"data":1186,"type":239,"tunes":1188},"HcVL8vGM74",{"text":1187},"Evidence Distance increases when a claim depends on more transformations, more assumptions, weaker provenance or stronger inference than the underlying evidence directly warrants.",{},{"id":1190,"data":1191,"type":280,"tunes":1192},"_f7WXjUOsS",{"code":362},{},{"id":1194,"data":1195,"type":239,"tunes":1197},"zUEC-cDhqm",{"text":1196},"A longer distance does not automatically make a claim false. It means the reasoning chain requires more validation.",{},{"id":1199,"data":1200,"type":41,"tunes":1202},"ix7TUGNz_W",{"text":1201,"level":233},"Why This Matters for AI",{},{"id":1204,"data":1205,"type":239,"tunes":1207},"wrAy5PLtLB",{"text":1206},"LLMs are particularly good at producing fluent bridges between pieces of information. That capability is useful, but fluency can hide how far the conclusion has moved beyond the directly retrieved evidence.",{},{"id":1209,"data":1210,"type":239,"tunes":1212},"jJseLuTWdv",{"text":1211},"An answer can therefore look well sourced while containing claims whose support exists only through several unstated inferential steps.",{},{"id":1214,"data":1215,"type":393,"tunes":1221},"ox8fok9dzs",{"link":1216,"meta":1217},"https:\u002F\u002Fstajic.de\u002Fblog\u002Ffrom-research-protocol-to-a-general-ai-reasoning-framework",{"image":1218,"title":1219,"description":1220},{"url":390},"From Research Protocol to a General AI Reasoning Framework","Reliable reasoning requires explicit separation of evidence, assumptions, hypotheses and conclusions.",{},{"id":1223,"data":1224,"type":41,"tunes":1226},"6ipse6YG1U",{"text":1225,"level":233},"Four Components of Evidence Distance",{},{"id":1228,"data":1229,"type":41,"tunes":1231},"P4xQ7o96dR",{"text":1230,"level":258},"1. Transformation Distance",{},{"id":1233,"data":1234,"type":239,"tunes":1236},"dvN8TUNEYG",{"text":1235},"How many times has the original evidence been converted, summarized, normalized or reformatted before reaching the model?",{},{"id":1238,"data":1239,"type":41,"tunes":1241},"dXofACF9_x",{"text":1240,"level":258},"2. Inferential Distance",{},{"id":1243,"data":1244,"type":239,"tunes":1246},"kkEzV3ydId",{"text":1245},"How many reasoning steps are required before the evidence supports the final conclusion?",{},{"id":1248,"data":1249,"type":41,"tunes":1251},"PPYB0RuX5_",{"text":1250,"level":258},"3. Provenance Distance",{},{"id":1253,"data":1254,"type":239,"tunes":1256},"cAVBYebJr7",{"text":1255},"Can the retrieved statement still be traced to an identifiable original source, or has it passed through summaries and copies whose origin is unclear?",{},{"id":1258,"data":1259,"type":41,"tunes":1261},"254N_99Z7X",{"text":1260,"level":258},"4. Applicability Distance",{},{"id":1263,"data":1264,"type":239,"tunes":1266},"to3KWv5WSK",{"text":1265},"How much reasoning is required to establish that the evidence applies to this exact version, user, jurisdiction, time period or situation?",{},{"id":1268,"data":1269,"type":41,"tunes":1271},"3WnVPQ6wqw",{"text":1270,"level":233},"Evidence Distance in RAG",{},{"id":1273,"data":1274,"type":239,"tunes":1276},"r8s-75EBhA",{"text":1275},"RAG can shorten one important distance: the distance between a model and external knowledge.",{},{"id":1278,"data":1279,"type":239,"tunes":1281},"I_8h52A3Yc",{"text":1280},"It does not automatically shorten the distance between a retrieved passage and the final claim.",{},{"id":1283,"data":1284,"type":280,"tunes":1285},"jDDRUu2FUK",{"code":458},{},{"id":1287,"data":1288,"type":393,"tunes":1294},"rm14qIVB1M",{"link":1289,"meta":1290},"https:\u002F\u002Fstajic.de\u002Fblog\u002Fwhere-does-an-llm-get-its-data-rag-data-sources-in-python",{"image":1291,"title":1292,"description":1293},{"url":390},"Where Does an LLM Get Its Data? RAG Data Sources in Python","Retrieval brings external information into context. Evidence Distance asks what happens between that information and the claim eventually produced.",{},{"id":1296,"data":1297,"type":41,"tunes":1299},"CRUZDXYloa",{"text":1298,"level":233},"Connection to the Retrieval Trigger",{},{"id":1301,"data":1302,"type":239,"tunes":1304},"-TPiYJ569c",{"text":1303},"The Retrieval Trigger asks whether internal model knowledge is sufficient.",{},{"id":1306,"data":1307,"type":239,"tunes":1309},"_VjLi5fs4x",{"text":1308},"Evidence Distance asks a later question: after evidence has been retrieved, how directly does it support what the system wants to claim?",{},{"id":1311,"data":1312,"type":280,"tunes":1313},"EImPw6i617",{"code":487},{},{"id":1315,"data":1316,"type":393,"tunes":1322},"ZqyhB2bFtR",{"link":1317,"meta":1318},"https:\u002F\u002Fstajic.de\u002Fblog\u002Fwhen-should-an-ai-stop-trusting-its-own-knowledge-the-retrieval-trigger",{"image":1319,"title":1320,"description":1321},{"url":390},"When Should an AI Stop Trusting Its Own Knowledge? — The Retrieval Trigger","The trigger decides when evidence is needed. Evidence Distance evaluates how far the eventual claim moves beyond that evidence.",{},{"id":1324,"data":1325,"type":41,"tunes":1327},"ST2vnm6vwF",{"text":1326,"level":233},"A Practical Diagnostic",{},{"id":1329,"data":1330,"type":239,"tunes":1332},"qyoHcDcTKb",{"text":1331},"For an important AI-generated claim, reconstruct the support chain backwards.",{},{"id":1334,"data":1335,"type":280,"tunes":1336},"e3fDmLk0fe",{"code":511},{},{"id":1338,"data":1339,"type":239,"tunes":1341},"iQ60c7NRaE",{"text":1340},"If the chain breaks, the system has encountered an evidence-traceability problem.",{},{"id":1343,"data":1344,"type":239,"tunes":1346},"YQAKlNZ2gM",{"text":1345},"If the chain remains intact but requires several strong assumptions, the Evidence Distance is still high.",{},{"id":1348,"data":1349,"type":41,"tunes":1351},"clN7jFkUBf",{"text":1350,"level":233},"Example: Product Data",{},{"id":1353,"data":1354,"type":239,"tunes":1356},"CRF-A9O7Ot",{"text":1355},"Suppose a manufacturer specification says a laptop weighs 1.48 kg.",{},{"id":1358,"data":1359,"type":280,"tunes":1360},"JF3z6aF2q_",{"code":536},{},{"id":1362,"data":1363,"type":239,"tunes":1365},"vTMOOgkLPb",{"text":1364},"Now consider:",{},{"id":1367,"data":1368,"type":280,"tunes":1369},"Tuhfzor1yp",{"code":546},{},{"id":1371,"data":1372,"type":239,"tunes":1374},"JHFRGIgUJD",{"text":1373},"The specification is real. The recommendation is still several reasoning steps away from it.",{},{"id":1376,"data":1377,"type":41,"tunes":1379},"2lKERCrnW1",{"text":1378,"level":233},"Example: Scientific Evidence",{},{"id":1381,"data":1382,"type":239,"tunes":1384},"HA8DBYwUuK",{"text":1383},"A paper may report an observed association. An AI answer can incorrectly compress that into a causal statement.",{},{"id":1386,"data":1387,"type":280,"tunes":1388},"D9C2dVB7j_",{"code":566},{},{"id":1390,"data":1391,"type":239,"tunes":1393},"CKNjK25pkL",{"text":1392},"The problem is not that no evidence exists. The problem is that the final statement crosses a larger evidential distance than the study directly supports.",{},{"id":1395,"data":1396,"type":41,"tunes":1398},"sR430N3cDd",{"text":1397,"level":233},"Connection to Agent-Ready Context",{},{"id":1400,"data":1401,"type":239,"tunes":1403},"v77YyG55ko",{"text":1402},"Structured information objects can reduce avoidable Evidence Distance by preserving source identity, units, timestamps, conflicts and validity instead of forcing the model to reconstruct them from flattened text.",{},{"id":1405,"data":1406,"type":393,"tunes":1412},"EX8darCIjd",{"link":1407,"meta":1408},"https:\u002F\u002Fstajic.de\u002Fportfolio\u002Ffrom-raw-data-to-verifiable-ai-context-a-practical-architecture-for-agent-ready-information",{"image":1409,"title":1410,"description":1411},{"url":390},"From Raw Data to Verifiable AI Context: A Practical Architecture for Agent-Ready Information","Preserving provenance and validity before retrieval reduces the number of unsupported reconstruction steps required later.",{},{"id":1414,"data":1415,"type":41,"tunes":1417},"dsfuoURj1e",{"text":1416,"level":233},"A Simple Working Scale",{},{"id":1419,"data":1420,"type":317,"tunes":1435},"BLi7Yth64O",{"content":1421,"stretched":42,"withHeadings":13},[1422,1425,1427,1429,1431,1433],[1423,1424],"Level","Description",[604,1426],"Directly reported source fact",[607,1428],"Simple deterministic transformation or direct comparison",[610,1430],"One meaningful interpretive step",[613,1432],"Multiple assumptions or combined sources",[616,1434],"Strong causal, predictive or prescriptive conclusion beyond direct evidence",{},{"id":1437,"data":1438,"type":239,"tunes":1440},"8y8bzCcHBs",{"text":1439},"This scale is illustrative, not a scientific standard. Its purpose is to force an explicit question: How much did we have to add between the evidence and the claim?",{},{"id":1442,"data":1443,"type":41,"tunes":1445},"aFK0YiAlYO",{"text":1444,"level":233},"Misconceptions",{},{"id":1447,"data":1448,"type":239,"tunes":1450},"07KSk6xDx-",{"text":1449},"A citation means Evidence Distance is low. No. A citation can sit beside a claim that substantially exceeds what the cited source establishes.",{},{"id":1452,"data":1453,"type":239,"tunes":1455},"t_cf-Zfan9",{"text":1454},"Long Evidence Distance means the claim is wrong. No. Complex conclusions can be well supported, but they require a stronger reasoning chain.",{},{"id":1457,"data":1458,"type":239,"tunes":1460},"Pu9PRX67Q6",{"text":1459},"Primary sources eliminate inference. No. Primary evidence can still require interpretation.",{},{"id":1462,"data":1463,"type":239,"tunes":1465},"oF1xNsOHHq",{"text":1464},"More sources automatically reduce distance. Ten weak repetitions of the same unsupported claim do not create direct evidence.",{},{"id":1467,"data":1468,"type":41,"tunes":1470},"IkolK5baph",{"text":1469,"level":233},"Edge Case: Derived Data",{},{"id":1472,"data":1473,"type":239,"tunes":1475},"jRH36fwdR3",{"text":1474},"Some derived values are more reliable than raw human descriptions. A correctly calculated total from verified transactions can be highly dependable even though it is one transformation away from the original records.",{},{"id":1477,"data":1478,"type":239,"tunes":1480},"SXbdKiNDst",{"text":1479},"Evidence Distance therefore measures epistemic separation, not a simplistic preference for untouched data.",{},{"id":1482,"data":1483,"type":41,"tunes":1485},"mUbvdnFChM",{"text":1484,"level":233},"Limitations",{},{"id":1487,"data":1488,"type":239,"tunes":1490},"N_EskxFOep",{"text":1489},"Evidence Distance is a conceptual diagnostic, not a universal quantitative score.",{},{"id":1492,"data":1493,"type":239,"tunes":1495},"nW_-E08So_",{"text":1494},"Different domains require different treatment of inference. Mathematical proof, clinical evidence, software logs, market analysis and historical interpretation cannot share one universal numerical threshold.",{},{"id":1497,"data":1498,"type":41,"tunes":1500},"_Bd95wXkri",{"text":1499,"level":233},"What Would Change This Answer?",{},{"id":1502,"data":1503,"type":239,"tunes":1505},"OqCaMc2_OX",{"text":1504},"Future AI systems may generate explicit provenance graphs, machine-checkable claim-evidence links and typed reasoning traces. That could make parts of Evidence Distance measurable automatically.",{},{"id":1507,"data":1508,"type":239,"tunes":1510},"58Hf3O9-Iz",{"text":1509},"The underlying problem remains: a claim can move farther from its evidence even while every intermediate sentence sounds plausible.",{},{"id":1512,"data":1513,"type":41,"tunes":1515},"8WOq5oE_e_",{"text":1514,"level":233},"Conclusion",{},{"id":1517,"data":1518,"type":239,"tunes":1520},"1pr1TfwnKc",{"text":1519},"The important distinction is not merely between sourced and unsourced answers.",{},{"id":1522,"data":1523,"type":239,"tunes":1525},"54CysT_NXa",{"text":1524},"It is between claims that remain close to directly inspectable evidence and claims that depend on increasingly long chains of interpretation and assumption.",{},{"id":1527,"data":1528,"type":280,"tunes":1529},"Qrjo3WEUSL",{"code":712},{},{"id":1531,"data":1532,"type":239,"tunes":1534},"6ir8aIAfd3",{"text":1533},"Evidence Distance gives us a language for describing that space.",{},{"id":1536,"data":1537,"type":41,"tunes":1539},"vPhupXcbeN",{"text":1538,"level":233},"Primary Sources",{},{"id":1541,"data":1542,"type":239,"tunes":1543},"BaO4JmWRQL",{"text":727},{},{"id":1545,"data":1546,"type":239,"tunes":1547},"eg4Z4G6Cgs",{"text":732},{},{"id":1549,"data":1550,"type":239,"tunes":1552},"BJH7zUd4T8",{"text":1551},"W3C, \u003Ca href=\"https:\u002F\u002Fwww.w3.org\u002FTR\u002Fprov-overview\u002F\" target=\"_blank\">PROV Overview\u003C\u002Fa>. A standardized model for representing provenance relationships.",{},"Not all supported answers are equally close to their evidence. Evidence Distance is a practical way to describe how many transformations, summaries, assumptions and inferential steps separate an AI claim from directly inspectable evidence.","Post erfolgreich abgerufen",{"items":1556,"source":1613,"manualIds":1614,"manualMatchedIds":1615},[1557,1564,1571,1578,1585,1592,1599,1606],{"id":1558,"slug":1559,"title":1560,"excerpt":1561,"featuredImage":1562,"publishedAt":1563},"491","ai-governance-models-data-permissions-risk-and-auditability","KI-Governance: Modelle, Daten, Berechtigungen, Risiko und Auditierbarkeit","KI-Governance definiert, wer KI-Systeme über Modelle, Anbieter, Daten, Berechtigungen, Risiken, Evaluierung und den gesamten Lebenszyklus hinweg genehmigen, betreiben, ändern und prüfen darf.","\u002Fuploads\u002F2026\u002F10\u002Fai-governance-models-data-permissions-risk-and-auditability-1791485901301-g60xyu.webp","2026-10-08T12:56:00.000Z",{"id":1565,"slug":1566,"title":1567,"excerpt":1568,"featuredImage":1569,"publishedAt":1570},"481","generative-ai-explained-models-retrieval-tools-and-applications-are-not-the-same-thing","Generative KI erklärt: Modelle, Retrieval, Tools und Anwendungen sind nicht dasselbe","Generative KI ist mehr als ein Modell. Erfahren Sie, wie Modelle, Retrieval, Tools, Kontext, Runtimes und Anwendungen in Produktions-KI-Systemen zusammenwirken.","\u002Fuploads\u002F2026\u002F10\u002Fgenerative-ai-explained-models-retrieval-tools-and-applications-are-not-the-same-thing-1791475411822-pp0dvz.webp","2026-10-08T12:00:00.000Z",{"id":1572,"slug":1573,"title":1574,"excerpt":1575,"featuredImage":1576,"publishedAt":1577},"494","air-gapped-ai-how-ai-systems-work-without-internet-or-cloud-access","Luftgetrennte KI: Wie KI-Systeme ohne Internet- oder Cloud-Zugriff funktionieren","Air-gapped AI führt Modelle, RAG und KI-Anwendungen innerhalb einer isolierten Sicherheitsdomäne ohne Internet- oder Cloud-Abhängigkeiten aus. Erfahren Sie, wie Modelle, Daten, Updates und Tools offline funktionieren.","\u002Fuploads\u002F2026\u002F10\u002Fair-gapped-ai-how-ai-systems-work-without-internet-or-cloud-access-1791487983978-e6xqf0.webp","2026-10-08T11:32:00.000Z",{"id":1579,"slug":1580,"title":1581,"excerpt":1582,"featuredImage":1583,"publishedAt":1584},"381","enterprise-grade-multi-tenant-architecture-for-an-international-platform","Unternehmensfähige mandantenfähige Architektur für eine internationale Plattform","Loving Rocks ist eine Hochzeitsplattform auf Unternehmensniveau, konzipiert mit einer echten Mehrmandantenarchitektur, isolierten Datenbanken pro Mandant und integrierter Internationalisierung für globale Skalierbarkeit, Sicherheit und langfristige Betriebsstabilität.","\u002Fuploads\u002F2026\u002F01\u002Fenterprise-grade-multi-tenant-architecture-for-an-international-platform-1769789121298-b6v7ak.webp","2026-01-30T12:04:00.000Z",{"id":1586,"slug":1587,"title":1588,"excerpt":1589,"featuredImage":1590,"publishedAt":1591},"495","sovereign-ai-control-of-models-data-infrastructure-and-dependencies","Souveräne KI: Kontrolle über Modelle, Daten, Infrastruktur und Abhängigkeiten","Souveräne KI bedeutet wirksame Kontrolle über Modelle, Daten, Infrastruktur, Software, Betrieb und strategische Abhängigkeiten – nicht einfach, wo ein KI-Modell gehostet wird.","\u002Fuploads\u002F2026\u002F10\u002Fsovereign-ai-control-of-models-data-infrastructure-and-dependencies-1791488833132-niy85x.webp","2026-10-08T15:45:00.000Z",{"id":1593,"slug":1594,"title":1595,"excerpt":1596,"featuredImage":1597,"publishedAt":1598},"498","can-evidence-survive-the-ai-pipeline-the-evidence-preservation-chain","Kann Beweismaterial die KI-Pipeline überleben? — Die Beweiserhaltungskette","Belege können an der Quelle korrekt sein und nach Parsing, Normalisierung, Chunking, Retrieval und Kontextzusammenstellung dennoch unzuverlässig werden. Die Evidence Preservation Chain verfolgt, ob Identität, Provenienz, Geltungsbereich und Bedeutung die gesamte KI-Pipeline überstehen.","\u002Fuploads\u002F2026\u002F10\u002Fcan-evidence-survive-the-ai-pipeline-the-evidence-preservation-chain-1792096396174-urdg9z.webp","2026-10-15T16:31:00.000Z",{"id":1600,"slug":1601,"title":1602,"excerpt":1603,"featuredImage":1604,"publishedAt":1605},"492","mcp-explained-what-it-connects-what-it-does-not-do-and-where-it-fits","MCP erklärt: Was es verbindet, was es nicht tut und wo es passt","Das Model Context Protocol verbindet KI-Anwendungen über eine standardisierte Client-Server-Grenze mit externen Tools, Ressourcen und Prompts. Erfahren Sie, was MCP tut, was es nicht tut und wo es in der Agentenarchitektur einzuordnen ist.","\u002Fuploads\u002F2026\u002F10\u002Fmcp-explained-what-it-connects-what-it-does-not-do-and-where-it-fits-1791486640275-7ub1cq.webp","2026-10-08T15:09:00.000Z",{"id":1607,"slug":1608,"title":1609,"excerpt":1610,"featuredImage":1611,"publishedAt":1612},"482","adr-vs-nfr-architecture-decisions-and-system-quality-are-not-the-same-thing","ADR vs. NFR: Architekturentscheidungen und Systemqualität sind nicht dasselbe","ADR vs. NFR erklärt: Erfahren Sie, wie Systemqualitätsanforderungen Architekturentscheidungen beeinflussen, wie ADRs Abwägungen dokumentieren und warum die Validierung getrennt bleibt.","\u002Fuploads\u002F2026\u002F10\u002Fadr-vs-nfr-architecture-decisions-and-system-quality-are-not-the-same-thing-1791475921511-6zgen1.webp","2026-10-08T12:11:00.000Z","fallback",[],[]]