{
  "@context": "https://schema.org",
  "@type": "ScholarlyArticle",
  "id": "background-without-profiling-2026-10-01",
  "title": "Background Without Profiling",
  "subtitle": "Human-Bound Persistent State for Agentic Systems",
  "headline": "Background Without Profiling: Human-Bound Persistent State for Agentic Systems",
  "author": {
    "@type": "Person",
    "name": "Raynor Eissens",
    "url": "https://research.madeby.re/"
  },
  "datePublished": "2026-10-01",
  "dateModified": "2026-10-01",
  "canonicalUrl": "https://research.madeby.re/background-without-profiling/",
  "language": "en",
  "publicationStatus": "independent web article",
  "doi": null,
  "articleType": [
    "conceptual architecture",
    "terminological note",
    "governance abstraction"
  ],
  "abstract": "Defines Background as persistent human-bound state from which replaceable agents, models, services, devices and artifacts receive bounded context, and distinguishes it from profiles, agent memory, personal data stores and AI twins.",
  "concept": {
    "name": "Background",
    "aliases": [
      "AI Background",
      "human-bound persistent state"
    ],
    "canonicalDefinition": "Background is persistent human-bound state from which replaceable agents, models, services, devices, and artifacts can receive purpose-bounded context. It is not itself a profile, agent memory, or simulation of the person.",
    "continuityRoot": "human principal",
    "accessPattern": "bounded projection",
    "designPrinciple": "Background Without Profiling",
    "coreRule": "Carry explicit and provenance-bearing personal state without making broad inference about the person a prerequisite for useful assistance."
  },
  "distinctions": [
    {
      "concept": "Profile",
      "function": "describes or classifies the person",
      "contrast": "Background preserves durable state for the person rather than primarily compressing the person for prediction or categorization."
    },
    {
      "concept": "User model",
      "function": "adapts one system to a user",
      "contrast": "Background is not subordinate to one application or service."
    },
    {
      "concept": "Agent memory",
      "function": "maintains continuity for an agent or agentic system",
      "contrast": "Background survives replacement of the consuming agent and can serve multiple agents."
    },
    {
      "concept": "Personal data store",
      "function": "stores and controls access to personal data",
      "contrast": "A personal data store can implement Background storage, but Background names the continuity and governance layer rather than a storage mechanism."
    },
    {
      "concept": "Personal knowledge graph",
      "function": "structures personal entities, facts and relationships",
      "contrast": "A knowledge graph can be one representation inside Background but does not define its complete state, lifecycle or access model."
    },
    {
      "concept": "AI twin",
      "function": "simulates or reproduces aspects of a specific person",
      "contrast": "Background need not simulate the person or act as the person."
    },
    {
      "concept": "Human Context Layer",
      "function": "structured permissioned context enabling personal agents",
      "contrast": "Strong overlap; Background makes human-bound continuity and replaceable consuming systems the explicit organizing principle."
    }
  ],
  "architecture": {
    "humanBound": "human -> Background -> bounded projection -> consuming system",
    "agentBoundContrast": "human -> agent -> agent memory",
    "formalProjection": "pi(B_t, Q_i,t, P_i)",
    "controlledWriteBack": "S_i -> proposed delta B_t -> validation/policy -> B_t+1"
  },
  "properties": [
    "persistent across sessions",
    "independent of a single model",
    "independent of a single agent",
    "usable across applications",
    "device-independent in principle",
    "portable across providers in the strong form",
    "selectively disclosed",
    "permissioned",
    "provenance-aware",
    "supports controlled write-back",
    "supports revocation and deletion",
    "can distinguish declared, observed, derived and inferred state"
  ],
  "substitutionTests": [
    "model replacement",
    "agent replacement",
    "application replacement",
    "device replacement",
    "provider migration",
    "multiple consumers",
    "temporary consumers",
    "revocation",
    "provenance preservation",
    "partial deletion"
  ],
  "nonClaims": [
    "No claim that persistent personal context is new.",
    "No claim that personal data stores or Solid-like architectures are new.",
    "No claim that selective disclosure, capability security, or verifiable credentials are new.",
    "No claim that agent memory or personal knowledge graphs are new.",
    "No claim that Background requires one universal database or identifier.",
    "No claim that all personalization should avoid inference."
  ],
  "noveltyBoundary": "Conceptual and governance abstraction built from substantial prior art; not claimed as a new storage, identity, memory, or cryptographic primitive.",
  "relatedWork": [
    "https://www.bcg.com/x/the-multiplier/personal-ai-agents-compete-on-trust",
    "https://solidproject.org/about",
    "https://www.w3.org/TR/vc-data-model/",
    "https://arxiv.org/abs/2504.19413",
    "https://arxiv.org/abs/2402.07540",
    "https://www.sciencedirect.com/science/article/pii/S2212473X2600088X",
    "https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/"
  ],
  "relatedInternal": [
    "https://research.madeby.re/ambient-phone/",
    "https://research.madeby.re/agent-offload-ambient-field/",
    "https://research.madeby.re/artifact-layer-intelligence-stack/",
    "https://carryinglayer.com/",
    "https://superartifact.si/"
  ],
  "encodings": {
    "html": "https://research.madeby.re/background-without-profiling/",
    "plainText": "https://research.madeby.re/background-without-profiling/article.txt",
    "json": "https://research.madeby.re/background-without-profiling/article.json"
  },
  "keywords": [
    "AI Background",
    "Background Without Profiling",
    "human-bound state",
    "persistent personal state",
    "agent memory",
    "Human Context Layer",
    "personal data store",
    "personal knowledge graph",
    "AI twin",
    "bounded projection",
    "data portability",
    "selective disclosure",
    "personal AI",
    "agentic systems"
  ]
}