Vis Lens Uncertainty

Uncertainty Representation Visualization Lens

Philosophical Mode: Statistical Primary Question: "How is uncertainty honestly represented?" Focus: Error Bar Definitions, Distribution-Aware Alternatives, Multi-Seed Variance Protocols

Arguments

/autoskillit:vis-lens-uncertainty [context_path] [experiment_plan_path]

• context_path (optional positional arg 1) — Absolute path to a lens context file containing IV/DV tables, H0/H1 hypotheses, controlled variables, and success criteria. If provided, read this file before beginning analysis to obtain structured context. If omitted, discover context by exploring the CWD.
• experiment_plan_path (optional positional arg 2) — Absolute path to the full experiment plan. If provided, read for complete experimental methodology and design. If omitted, locate the experiment plan by exploring the CWD.

When to Use

• Reviewing figures that show means, scores, or aggregate metrics without uncertainty
• Checking whether error bars are correctly defined (SD vs SE vs CI is not interchangeable)
• Evaluating whether n_seeds is sufficient to quantify variance
• Planning uncertainty overlays for distribution-heavy results (RL reward curves, ablations)
• User invokes /autoskillit:vis-lens-uncertainty

Critical Constraints

NEVER:

• Modify any source code files
• Do not litter the codebase with useless comments, TODO markers, or explanatory annotations — the skill output and diagram speak for themselves
• Create files outside {{AUTOSKILLIT_TEMP}}/vis-lens-uncertainty/
• Conflate SD, SE, CI95, and PI95 — they have fundamentally different interpretations
• Omit the CRITICAL flag when n_seeds == 1; single-seed variance is unquantifiable

ALWAYS:

• CRITICAL: if n_seeds == 1, flag the figure as CRITICAL — single-seed variance is unquantifiable and must be reported

• If n_seeds >= 10, prefer violin/box/strip over error bars to show the full distribution shape

• Label every uncertainty overlay with its exact measure (SD, SE, CI95, PI95) and n in the figure caption

• BEFORE creating any diagram, LOAD the /autoskillit:mermaid skill using the Skill tool - this is MANDATORY

• If the Skill tool cannot be used (disable-model-invocation) or refuses this invocation, do NOT proceed with diagram creation. Abort this step and omit the diagram from output.

• Write output to {{AUTOSKILLIT_TEMP}}/vis-lens-uncertainty/vis_spec_uncertainty_{YYYY-MM-DD_HHMMSS}.md (relative to the current working directory)

• After writing the file, emit the structured output token as literal plain text with no markdown formatting on the token name (the adjudicator performs a regex match):

  diagram_path = /absolute/path/to/{{AUTOSKILLIT_TEMP}}/vis-lens-uncertainty/vis_spec_uncertainty_{...}.md
  

Analysis Workflow

Step 0: Parse optional arguments

If positional arg 1 (context_path) is provided and the file exists, read it to obtain IV/DV tables, H0/H1 hypotheses, controlled variables, and success criteria. If positional arg 2 (experiment_plan_path) is provided and exists, read the experiment plan for full methodology. Use this structured context as the foundation for Steps 1–4; skip the CWD exploration for these fields if the context file supplies them.

Step 1: Inventory Figures and Seeds

Scan experiment plan, context file, and codebase for:

Figures with Error-Bearing Quantities

• Find all figures that aggregate over multiple runs, seeds, or samples
• Look for: mean, average, std, stderr, ci, error_bar, errorbar, fill_between

Seed Count

• Find the number of random seeds or independent runs used
• Look for: n_seeds, num_seeds, seeds, SEEDS, random_state, seed_list, runs

Existing Uncertainty Representation

• Find whether any uncertainty overlay exists at all
• Look for: plt.fill_between, ax.errorbar, capsize, yerr, xerr, ci=, sd=, ci_band

Claims About Variance

• Find claims that assert robustness, stability, or statistical significance
• Look for: significant, robust, stable, p-value, consistent, reproducible

Step 2: Determine Correct Uncertainty Measure Per Figure

For each figure that shows an aggregated quantity, determine the correct measure:

SD — Standard Deviation

• Definition: spread of the data distribution (population variability), not inference error
• When to use: when the claim is about spread of the distribution, not precision of the mean
• Formula: σ = sqrt(1/(n-1) × Σ(xᵢ - x̄)²)

SE — Standard Error of the Mean

• Definition: SD / √n — sampling error of the estimated mean
• When to use: when the claim is about how precisely the mean is estimated
• Formula: SE = SD / √n

CI 95% — Confidence Interval

• Definition: mean ± t(0.975, n−1) × SE — inferential, sample-size dependent
• When to use: when making statistical inference claims; shrinks with larger n
• Formula: CI = mean ± t(0.975, n−1) × (SD / √n)

PI 95% — Prediction Interval

• Definition: interval for a future single observation (wider than CI)
• When to use: when the claim is about where a new individual result will fall
• Formula: PI = mean ± t(0.975, n−1) × SD × sqrt(1 + 1/n)

Distribution-Aware Selection:

• n_seeds == 1 → CRITICAL: variance cannot be quantified; flag this immediately
• n_seeds 2–4 → CI95 or SE; warn that the interval is unreliable at low n
• n_seeds 5–9 → error_bar with CI95 is acceptable; consider box plot
• n_seeds ≥ 10 → prefer violin/box/strip over error bars to show full distribution shape

Step 3: Flag Critical Cases

For every figure where n_seeds == 1:

• Mark with severity: CRITICAL
• Reason: single-seed variance is unquantifiable; the result may not replicate
• Remediation: report results across ≥ 3 seeds minimum (5+ recommended)

For figures where the wrong measure is used (e.g., SD labeled as CI, or CI claimed without stating n):

• Mark with severity: WARNING
• Document the specific mislabeling and the correct interpretation

Step 4: Emit yaml:figure-spec Blocks

For each figure, emit one yaml:figure-spec fenced block (schema defined in vis-lens-chart-select) with stat_overlay filled in. Then LOAD /autoskillit:mermaid and create the mermaid diagram.

Output Template

# Uncertainty Representation Spec: {System / Experiment Name}

**Lens:** Uncertainty Representation (Statistical)
**Question:** How is uncertainty honestly represented?
**Date:** {YYYY-MM-DD}
**Scope:** {What was analyzed}
**n_seeds detected:** {N}

## Uncertainty Measure Summary

| Figure | n_seeds | Recommended Measure | Current Measure | Status |
|--------|---------|---------------------|-----------------|--------|
| {fig-01} | 1 | N/A — CRITICAL | none | CRITICAL |
| {fig-02} | 5 | CI95 | SE | WARNING — mislabeled |
| {fig-03} | 10 | violin | error_bar | WARNING — prefer distribution plot |
| {fig-04} | 5 | CI95 | CI95 | OK |

## Figure Specs

```yaml
# yaml:figure-spec — canonical schema (spec_version: "1.0")
figure_id: "fig-02-ablation-accuracy"
figure_title: "Ablation Study: Component Contribution"
spec_version: "1.0"
chart_type: "grouped-bar"
chart_type_fallback: "dot-plot"
perceptual_justification: "Position encoding for nominal × quantitative comparison."
data_source: "results/ablation.csv"
data_mapping:
  x: "variant"
  y: "accuracy"
  color: "component"
  size: ""
  facet: ""
layout:
  width_inches: 5.0
  height_inches: 3.5
  dpi: 300
stat_overlay:
  type: "error_bar"
  measure: "CI95"
  n_seeds: 5
annotations: ["n=5 seeds; CI95 shown"]
anti_patterns: ["ap-bar-no-error"]
palette: "okabe-ito"
format: "pdf"
target_dpi: 300
library: "matplotlib"
report_section: "Section 5 Ablation"
priority: "P1"
placement_tier: "main"
conflicts: []
metadata:
  created_by: "vis-lens-uncertainty"
  reviewed_by: ""
  last_updated: "{YYYY-MM-DD}"

Uncertainty Representation Diagram

%%{init: {'flowchart': {'nodeSpacing': 50, 'rankSpacing': 60, 'curve': 'basis'}}}%%
flowchart TB
    %% CLASS DEFINITIONS %%
    classDef cli fill:#1a237e,stroke:#7986cb,stroke-width:2px,color:#fff;
    classDef stateNode fill:#004d40,stroke:#4db6ac,stroke-width:2px,color:#fff;
    classDef handler fill:#e65100,stroke:#ffb74d,stroke-width:2px,color:#fff;
    classDef output fill:#00695c,stroke:#4db6ac,stroke-width:2px,color:#fff;
    classDef detector fill:#b71c1c,stroke:#ef5350,stroke-width:2px,color:#fff;
    classDef gap fill:#ff6f00,stroke:#ffa726,stroke-width:2px,color:#000;

    subgraph Seeds ["SEED COUNT"]
        N1["n_seeds = {N}<br/>━━━━━━━━━━<br/>{adequacy assessment}"]
    end

    subgraph Measure ["UNCERTAINTY MEASURE"]
        M1["SD / SE / CI95 / PI95<br/>━━━━━━━━━━<br/>{chosen measure + formula}"]
    end

    subgraph Overlay ["OVERLAY TYPE"]
        O1["{error_bar / violin / box}<br/>━━━━━━━━━━<br/>{distribution-aware choice}"]
    end

    subgraph Verdict ["VERDICT"]
        V1["{OK / WARNING / CRITICAL}<br/>━━━━━━━━━━<br/>{reason}"]
    end

    N1 --> M1
    M1 --> O1
    O1 --> V1

    class N1 stateNode;
    class M1 cli;
    class O1 output;
    class V1 detector;

Color Legend:

---

## Pre-Diagram Checklist

Before creating the diagram, verify:

- [ ] LOADED `/autoskillit:mermaid` skill using the Skill tool
- [ ] Using ONLY classDef styles from the mermaid skill (no invented colors)
- [ ] Diagram will include a color legend table
- [ ] Every CRITICAL (n_seeds == 1) figure is flagged
- [ ] Every stat_overlay has both `measure` and `n_seeds` filled in