COSMOS-CQAPublic Research Portal

Research Experiment

What the COSMOS-CQA workbench is testing.

COSMOS-CQA asks whether a browser-native, public, human-in-the-loop workbench can make cosmology artifact review more reproducible, inspectable, accessible, and scientifically careful.

Open Demo Workbench Repository Experiment Note

What experiment is this workbench modeling?

The modeled experiment is not a claim that the current demo detects cosmological signals. It is a research-infrastructure experiment: can visual tiles, sidecar views, labels, evidence bundles, provenance hashes, and deterministic replay help researchers and public contributors review artifact candidates in a way that is auditable after the session ends?

The hosted demo uses a synthetic Core Pack contract fixture to exercise the workflow safely. It models the kind of quality-assurance surface that could support future artifact review for public cosmology data cultures around CMB maps, weak-lensing products, survey images, and catalog release evidence. Public-source candidates are tracked separately until rights, attribution, CORS, rate-limit, and replay requirements are reviewed.

What does the current prototype demonstrate?

Core Pack intakeA manifest-driven lane for loading tile passports, evidence references, SBOM references, diagnostics, and provenance context.
Human review loopTile navigation, overlays, palettes, labels, undo, bookmarks, calibration, captions, and public truth-label hiding.
Evidence exportsValidation report JSON, session JSON, evidence bundle JSON, SBOM JSON, and replayable bookmark state.
Deterministic replayGolden fixtures and browser tests verify the sample workflow, report surfaces, and public/dev display boundaries.

Core engine fundamentals exposed publicly

The public project exposes the fundamentals needed to audit the research workflow without disclosing or implying proprietary future diagnostic claims.

Data contract

Core Pack manifests identify tiles, passports, evidence references, SBOM references, diagnostic placeholders, and release stewardship.

Tile passport

Each tile can carry dataset context, coordinates, checksum, sidecars, provenance, and truth-policy language.

Review sidecars

Overlays, palettes, captions, and sonic loops provide alternate ways to inspect the same review object.

Evidence chain

Labels, bookmarks, reports, hashes, sessions, and bundles preserve what happened during review.

How the sonic loop must be explained

The sonic loop is a deterministic sonification sidecar. In the current workbench, it turns selected tile-derived values into a short repeatable auditory contour so a reviewer can listen for changes in rhythm, pitch, or texture while still using visual overlays, captions, and written evidence.

It must be described as a perception, accessibility, attention, and training aid. It must not be described as an independent detector, classifier, scientific measurement, or validated diagnostic. Any future scientific use would require a documented mapping, controlled evaluation, accessibility review, bias/error analysis, and reproducible evidence.

It must also be explained as optional, user initiated, and bounded. Browser code can limit frequency and software gain, but it cannot control device volume, headphones, room acoustics, or individual sensitivity.

Public explanation: the sound is another way to inspect the same tile-sidecar data. It helps people notice patterns, compare states, and include contributors who benefit from auditory review. It does not decide the label.

How researchers and citizens benefit

For researchersA lightweight way to prototype artifact QA protocols, inspect evidence trails, test public review workflows, and discuss limitations before scaling.
For public contributorsA transparent path into research participation that teaches evidence structure, uncertainty, reproducibility, and responsible claim boundaries.
For open researchA reusable model for pairing citizen science participation with contracts, provenance, validation reports, accessibility, and local-first data handling.
For educationA hands-on demonstration of how scientific review surfaces can separate observation, label, evidence, caveat, and claim.

What is explicitly not claimed yet?

  • The current public demo does not validate a cosmology diagnostic or publish scientific results.
  • The sonic loop does not classify artifacts, discover signals, or replace visual/statistical review.
  • The synthetic Core Pack fixture is evidence-plumbing material, not public observational data or a restricted third-party science dataset.
  • Public mode hides truth labels; dev truth display is only for explicit local development review.
  • COSMOS-CQA is research-only public infrastructure, not an OSI open-source release and not a production decision system.

Public references

These references ground the public science, citizen science, and sonification context. They do not imply endorsement, data redistribution permission, scientific validation, or production fitness.

  1. ESA Planck Legacy Archive, public Planck mission product archive and conceptual reference for CMB map-release workflows.
  2. NASA LAMBDA WMAP data products, public CMB mission data reference for reproducible map products.
  3. NASA LAMBDA ACT products, public Atacama Cosmology Telescope data reference for modern CMB survey products.
  4. Dark Energy Survey Data Release 2, public survey release reference for imaging, catalog, and weak-lensing review cultures.
  5. Galaxy Zoo / Zooniverse, citizen-science reference for large-scale public classification and research participation.
  6. NASA Chandra Data Sonification, public astronomy sonification reference showing how astronomical data can be represented through sound.
  7. The Sonification Handbook, open reference on representing data with non-speech audio and the care required when explaining auditory mappings.