Add data fabrication anomaly assistant

data-fabrication-anomaly-assistant/README.md

@@ -0,0 +1,23 @@
+# Data Fabrication Anomaly Assistant
+
+This is a focused AI-Powered Research Assistant Suite slice for issue #16. It reviews synthetic manuscript data packets before AI peer-review output is trusted and emits release, review, or hold decisions for data-forensics red flags.
+
+The assistant checks:
+
+- repeated measurement rows that need raw-data verification
+- invalid collection timestamps that cannot be audited
+- terminal-digit preference that suggests rounding or synthetic entry
+- unusually smooth measurement series
+- perfect group separation that needs preregistered exclusion and raw-row provenance
+
+All fixtures are synthetic. The module does not call external AI services, publisher APIs, private datasets, credential stores, payment systems, or live manuscript systems.
+
+## Run
+
+```sh
+npm run check
+npm test
+npm run demo
+```
+
+The demo writes reviewer artifacts under `reports/`.

data-fabrication-anomaly-assistant/demo.js

@@ -0,0 +1,31 @@
+import fs from "node:fs";
+import path from "node:path";
+import {
+  evaluateFabricationAnomalyPacket,
+  renderSummarySvg,
+  summarizeReview,
+} from "./src/assistant.js";
+import { cleanPacket, riskyPacket } from "./src/samplePackets.js";
+
+const reportsDir = path.join(process.cwd(), "reports");
+fs.mkdirSync(reportsDir, { recursive: true });
+
+const clean = evaluateFabricationAnomalyPacket(cleanPacket);
+const risky = evaluateFabricationAnomalyPacket(riskyPacket);
+
+fs.writeFileSync(
+  path.join(reportsDir, "clean-review.json"),
+  `${JSON.stringify(clean, null, 2)}\n`,
+);
+fs.writeFileSync(
+  path.join(reportsDir, "risky-review.json"),
+  `${JSON.stringify(risky, null, 2)}\n`,
+);
+fs.writeFileSync(path.join(reportsDir, "risky-review.md"), summarizeReview(risky));
+fs.writeFileSync(path.join(reportsDir, "summary.svg"), renderSummarySvg(risky));
+
+console.log("Wrote data fabrication anomaly assistant reports:");
+console.log("- reports/clean-review.json");
+console.log("- reports/risky-review.json");
+console.log("- reports/risky-review.md");
+console.log("- reports/summary.svg");

data-fabrication-anomaly-assistant/make-demo-video.js

@@ -0,0 +1,50 @@
+import { spawnSync } from "node:child_process";
+import path from "node:path";
+import { riskyPacket } from "./src/samplePackets.js";
+import { evaluateFabricationAnomalyPacket } from "./src/assistant.js";
+
+const reportsDir = path.join(process.cwd(), "reports");
+const demoMp4 = path.join(reportsDir, "demo.mp4");
+const resultPacket = evaluateFabricationAnomalyPacket(riskyPacket);
+
+function escapeDrawtext(text) {
+  return text.replaceAll("\\", "\\\\").replaceAll(":", "\\:").replaceAll("'", "\\'");
+}
+
+const font = "C\\:/Windows/Fonts/arial.ttf";
+const lines = [
+  "Data Fabrication Anomaly Assistant",
+  `Decision ${resultPacket.decision} | Findings ${resultPacket.findingCount}`,
+  "Flags duplicate rows, invalid timestamps, digit preference, smooth series, and perfect separation.",
+  `Audit digest ${resultPacket.auditDigest}`,
+];
+const drawText = lines
+  .map(
+    (line, index) =>
+      `drawtext=fontfile='${font}':text='${escapeDrawtext(line)}':x=48:y=${64 + index * 72}:fontsize=${index === 0 ? 34 : 24}:fontcolor=${index === 1 ? "0xffdddd" : "white"}`,
+  )
+  .join(",");
+
+const result = spawnSync(
+  "ffmpeg",
+  [
+    "-y",
+    "-f",
+    "lavfi",
+    "-i",
+    "color=c=0x111827:s=960x540:r=12",
+    "-t",
+    "4",
+    "-vf",
+    `${drawText},format=yuv420p`,
+    "-an",
+    demoMp4,
+  ],
+  { encoding: "utf8" },
+);
+
+if (result.status !== 0) {
+  throw new Error(result.stderr || "ffmpeg failed to render demo.mp4");
+}
+
+console.log(`Wrote ${path.relative(process.cwd(), demoMp4)}`);

data-fabrication-anomaly-assistant/package.json

@@ -0,0 +1,13 @@
+{
+  "name": "data-fabrication-anomaly-assistant",
+  "version": "1.0.0",
+  "private": true,
+  "description": "Synthetic data-forensics review assistant for SCIBASE AI research review packets.",
+  "type": "module",
+  "scripts": {
+    "check": "node --check src/assistant.js && node --check src/samplePackets.js && node --check test.js && node --check demo.js && node --check make-demo-video.js",
+    "test": "node test.js",
+    "demo": "node demo.js",
+    "demo:video": "node make-demo-video.js"
+  }
+}

data-fabrication-anomaly-assistant/reports/clean-review.json

@@ -0,0 +1,9 @@
+{
+  "packetId": "clean-growth-study",
+  "manuscriptTitle": "Enzyme response in replicated plant growth chambers",
+  "decision": "RELEASE",
+  "findingCount": 0,
+  "findings": [],
+  "auditDigest": "5db12124295e2085",
+  "assistantScope": "Synthetic data-fabrication anomaly red flags before AI peer-review output is trusted."
+}

data-fabrication-anomaly-assistant/reports/risky-review.json

@@ -0,0 +1,45 @@
+{
+  "packetId": "risky-cytokine-study",
+  "manuscriptTitle": "Cytokine response claims from a disputed assay batch",
+  "decision": "HOLD",
+  "findingCount": 5,
+  "findings": [
+    {
+      "id": "duplicate-measurement-rows",
+      "severity": "high",
+      "title": "Repeated measurement rows need raw-data verification",
+      "evidence": "1 duplicate row pair(s): #12/#13",
+      "action": "Hold AI review output until source instruments, audit trail, and import logs explain the repeated rows."
+    },
+    {
+      "id": "invalid-collection-timestamps",
+      "severity": "high",
+      "title": "Collection timestamps are not machine-auditable",
+      "evidence": "Invalid timestamp rows: #14",
+      "action": "Require normalized ISO-8601 collection timestamps before the assistant cites this dataset."
+    },
+    {
+      "id": "terminal-digit-preference",
+      "severity": "medium",
+      "title": "Terminal digit distribution is unusually concentrated",
+      "evidence": "50% of numeric measurements end in 0",
+      "action": "Ask for raw instrument exports or an explanation of rounding rules before trusting fine-grained effects."
+    },
+    {
+      "id": "over-smooth-measurement-series",
+      "severity": "medium",
+      "title": "Measurement series is smoother than expected for independent observations",
+      "evidence": "control delta sd/value sd=0",
+      "action": "Route to statistical reviewer for instrument drift, interpolation, or synthetic-row checks."
+    },
+    {
+      "id": "perfect-between-group-separation",
+      "severity": "medium",
+      "title": "Groups separate perfectly without overlap",
+      "evidence": "control: n=6, range=10-12.5; treated: n=8, range=20-23",
+      "action": "Require preregistered exclusion rules and raw row provenance before strong causal or classification claims are released."
+    }
+  ],
+  "auditDigest": "6466f6cc655b23a7",
+  "assistantScope": "Synthetic data-fabrication anomaly red flags before AI peer-review output is trusted."
+}

data-fabrication-anomaly-assistant/reports/risky-review.md

@@ -0,0 +1,22 @@
+# Data fabrication anomaly review: Cytokine response claims from a disputed assay batch
+
+Decision: **HOLD**
+Audit digest: `6466f6cc655b23a7`
+
+## Findings
+
+- **HIGH** Repeated measurement rows need raw-data verification (duplicate-measurement-rows)
+  Evidence: 1 duplicate row pair(s): #12/#13
+  Action: Hold AI review output until source instruments, audit trail, and import logs explain the repeated rows.
+- **HIGH** Collection timestamps are not machine-auditable (invalid-collection-timestamps)
+  Evidence: Invalid timestamp rows: #14
+  Action: Require normalized ISO-8601 collection timestamps before the assistant cites this dataset.
+- **MEDIUM** Terminal digit distribution is unusually concentrated (terminal-digit-preference)
+  Evidence: 50% of numeric measurements end in 0
+  Action: Ask for raw instrument exports or an explanation of rounding rules before trusting fine-grained effects.
+- **MEDIUM** Measurement series is smoother than expected for independent observations (over-smooth-measurement-series)
+  Evidence: control delta sd/value sd=0
+  Action: Route to statistical reviewer for instrument drift, interpolation, or synthetic-row checks.
+- **MEDIUM** Groups separate perfectly without overlap (perfect-between-group-separation)
+  Evidence: control: n=6, range=10-12.5; treated: n=8, range=20-23
+  Action: Require preregistered exclusion rules and raw row provenance before strong causal or classification claims are released.