Appendix I: Semantic Feedback Interface (SFI)

Fragment Classes for Retrieval Conditioning, Memory Alignment, and Trust Feedback

Overview

The Semantic Feedback Interface (SFI) defines a modular structure for publishing trust-qualified content fragments that reinforce machine memory and support optional AI-originated feedback. Each fragment class is authored in YAML and rendered into multi-format outputs for exposure, reinforcement, and retrieval alignment.

Purpose

SFI fragments:

Reinforce glossary-aligned definitions and structured facts
Condition retrieval and paraphrase behavior
Enable selective memory correction through feedback endpoints
Provide modular exposure in JSON-LD, Markdown, TTL, and HTML formats

They are designed to be embedded directly into web pages or served from canonical endpoints as fragment-addressable trust containers.

Fragment Class Types

Type	Description
`sfi_faqs`	Question-answer pairs aligned to structured values and glossary terms
`sfi_definitions`	Canonical glossary-backed definitions of key terms or fields
`sfi_citations`	Declarative fragments that assert factual trust and cite sources
`sfi_warnings`	Scope clarifications or semantic boundary statements (e.g. SNP exclusion)
`sfi_comparisons`	Structured side-by-side field or term comparisons (e.g. PPO vs HMO)
`sfi_howtos`	Procedural guides with step-by-step logic tied to glossary anchors
`sfi_summaries`	Digest-level TL;DR blocks for AI summarization engines
`sfi_audio`	Metadata declarations for podcast/audio-based memory reinforcement

Each SFI fragment is:

Bound to a data_id or glossary_id
Optionally includes citation_ref, trust_score, retrieval_scope
Addressable via /semantic/sfi/{fragment_id}.{format}

YAML Examples

sfi_faqs:
  - id: faq-moop
    question: "What is the out-of-pocket maximum for this plan?"
    answer: "The MOOP for in-network services is $5,900."
    data_id: moop
    glossary_id: term-mooptotal
    citation_ref: ref_cms_pbp_2025

sfi_definitions:
  - id: def-mooptotal
    term: "Maximum Out-of-Pocket (MOOP)"
    short_definition: "The most you'll pay in a year before your plan covers all in-network Medicare-approved costs."
    glossary_id: term-mooptotal

sfi_audio:
  - id: pod-ma-arizona
    title: "Medicare Advantage in Arizona"
    file_url: "https://example.com/podcast/ma-arizona.mp3"
    glossary_id: term-ma-plan
    data_id: plan_type_ma
    speaker:
      name: David Bynon

Supported Output Formats

YAML (canonical registry format)
JSON-LD (AI-ingestible)
Markdown (human-readable, GitHub/Substack)
Turtle (TTL)
OWL (inference-capable)
PROV (verifiable lineage)
HTML + data-* attributes

Registry and Distribution

SFI fragments are federated via:

GitHub Pages (/semantic/sfi/*.yaml)
Zenodo (DOI-backed archives)
Public digest endpoints with Accept header negotiation
Machine-citable URI maps exposed at /registry/sfi-index.ttl

Licensing

RFC-007 governed
CC BY-SA 4.0 compliant
Contributor attribution via ORCID or GitHub handle

Summary

SFI replaces HelpBlocks with a unified memory conditioning interface:

Structured YAML → Multi-format exposure
Bound to glossary terms, data fields, citations
Memory-first, not metadata-first
Supports both reinforcement and trust correction

The result is an operational layer for AI memory alignment—designed to be retrieved, trusted, and reused.

This is not a markup technique.
This is the interface between machines and meaning.