Skip to main content
Eligible orders over US$80 ship free by sea No minimum order US · UK · CA · AU
P074 · AI transparency layer · index, follow
P074 · Model Card / Validation / Limits

AI Tongue Methodology & Limitations

See what the model is designed to do, what data supports it, how performance is tested, where it fails and which metrics are not yet public.

Missing evidence is displayed as missing. A bare “95% accurate” claim is not acceptable without the task, model, dataset, split, metric and limitations.

No invented accuracyDataset counts separatedExternal validation explicit
SystemAI Tongue Explorer
DiagnosisNo
PrescriptionNo
Robotsindex, follow

Transparency Principle

A methodology page is trustworthy when it makes missing evidence visible. “No audited public metric is currently published” is more credible than a fabricated accuracy percentage.

Current model card

What Is Public Today?

Unknown model and validation values are displayed as Not Published / Evidence Pending rather than invented.

ModelNot Published
VersionNot Published
Validation StatusEvidence Pending
External ValidationNot Yet Validated / Not Published
Field
Current Public Value
Status
Release Date
Not Published
Dynamic
Intended Task
Educational visible-image observations / constitution-oriented context; exact production task pending model card
Scoped
Output Classes
Not Published
Dynamic
Input Type
Not Published
Dynamic
Quality Requirements
Not Published
Dynamic
Disease Diagnosis
No
Boundary
Intended use

What the Model May Do—and What It Must Not Do

The literal production pipeline must be verified before it is published as architecture.

Intended Educational Pipeline

  1. Receive a suitable image.
  2. Evaluate image quality.
  3. Detect/segment the supported visible region.
  4. Extract supported visible features.
  5. Map supported outputs to traditional constitution-oriented context.
  6. Return confidence and limitations.

Not Intended

  • disease diagnosis or screening
  • cancer / diabetes / organ-function diagnosis
  • anemia or infection detection
  • prescription or medication recommendations
  • emergency triage
Image quality gate

Input Quality Is Part of Performance

Lighting, exposure, white balance, blur, camera processing, distance, tongue position, obstruction and cropping can affect an image-based system.

1

Acceptable

Proceed under the current tested capture specification.

2

Borderline

Warn or lower confidence only if the current method supports that state.

3

Unacceptable

Reject rather than invent a result.

4

Report Rejection Rate

Not Published

Dataset accounting

Images Are Not People

Unique people, original images, usable images and augmented images must be counted separately.

Unique PeopleNot Published
Original ImagesNot Published
Usable ImagesNot Published
Augmented ImagesNot Published
Never say “trained on 100,000 patients” merely because augmentation produced 100,000 image samples. Person-level independence and duplicate control matter.
Split & ground truth

Labels, Duplicates & Train/Test Separation

Performance can be overstated if the label definition is weak or the same person's images leak across splits.

Ground Truth

Define the Label

Document who assigned it, framework source, annotator process, adjudication and uncertain-label handling.

Duplicate Control

Control Repeated Images / People

Exact and near duplicates plus same-person grouping matter when person-level generalization is the intended test.

Split

Separate Development from Evaluation

Document splitting unit, proportions/counts, leakage prevention and tuning policy.

Metrics

A Single “Accuracy” Number Is Not Enough

Every performance statement should name the task, model version, dataset, unit of analysis, split, metric, threshold, subgroup and limitations.

Public Validation MetricsNo audited public performance metric is currently published.
CalibrationNot Published
Subgroup ReportNot Published
External ValidationUse explicit status above
Where appropriate, report point estimate, uncertainty/confidence interval and sample count. Avoid false precision.
Failure modes

Where the Model Can Fail

Failure conditions should be public and versioned.

Capture

Lighting / Blur / Crop

Colored lighting, blur, partial tongue and heavy filtering are obvious failure candidates.

Distribution

New Cameras / Populations

Performance can shift across country, device and capture workflow.

Interpretation

Ambiguity / Unsupported Cases

Coating/color ambiguity, unusual presentation and out-of-distribution inputs need explicit handling.

Verified failure modes: Not Published
Version control

Model Changes Require New Evidence

Do not silently retrain and keep the same methodology page/version.

Change Record
What Must Be Tracked
Status
Version / Date
What changed and why
Required
Training Data
Changed?
Required
Labels / Thresholds
Changed?
Required
Validation
Re-run and reviewer
Required
Current change log: Not Published
Three separate validity questions

Traditional Framework ≠ AI Performance ≠ Medical Diagnostic Validity

P071 defines constitution concepts. P074 explains the narrow digital model task. Neither one should be used to manufacture medical diagnostic validity.

Question
Owner
What It Answers
Traditional definitions
P071
What the nine constitution categories mean in the chosen traditional framework
Technical performance
P074
How model vX performs under protocol Y
User-data privacy
P072
How current user images/results are collected, stored and controlled
Questions

Frequently Asked Questions

Short answers preserve the page boundary and route deeper safety or product questions to the correct owner.

How accurate is AI tongue analysis?

Use the published model/task metrics. If no audited metrics exist, no public validated percentage should be stated.

How many images trained the AI?

See audited original and usable image counts when published.

How many people?

See the separate unique-person count.

Has it been externally validated?

Use the explicit external-validation status.

Does lighting affect results?

It can; current capture and robustness tests should quantify the effect.

Is confidence the same as accuracy?

No.

Is this a disease diagnostic model?

No.

Can an advisor replace validation?

No.

Can a model update change accuracy?

Yes; material updates require new validation and version disclosure.