coulomb-research/seed/ResearchCharts.md

# Research Charts

## Purpose

Research Charts make scientific activity culturally legible and discoverable.

They borrow from music charts, box-office rankings, and certification systems while refusing to equate popularity with truth.

The purpose is relevance marketing, discovery, recognition, and system sensing.

## 1. Core Principle

> Charts rank observable signals within a declared scope. They do not rank truth.

Every chart must make clear:

- what is measured,
- over what time period,
- within what comparison cohort,
- with what normalization,
- using which data sources.

## 2. Chartable Objects

Primary chart targets:

- papers,
- propositions,
- datasets,
- software,
- replications,
- criticisms,
- open problems,
- research programs,
- research collectives.

Person-centric rankings should be secondary.

The system should reward useful research objects and contributions before celebrity.

## 3. Initial Chart Families

### General

- Global Research Top 100
- Research Top 100 — Year
- Research Top 100 — YTD
- Research Top 100 — Month

### Field-specific

- AI Top 50
- Physics Top 50
- Biology Top 50
- Systems Top 50
- field-defined charts

### Dynamic

- Fastest Rising
- Breakthrough Movers
- Long-Tail Classics
- Comeback Papers

### Scientific-function charts

- Most Reused
- Most Replicated
- Most Challenged
- Most Generative
- Most Discussed
- Most Influential Methods
- Most Productive Open Problems
- Most Valuable Criticisms
- Replication of the Year
- Most Useful Negative Result

These categories deliberately broaden prestige beyond publication.

## 4. Time Windows

Each chart should support explicit windows such as:

- day,
- week,
- month,
- quarter,
- year-to-date,
- calendar year,
- rolling 12 months,
- all-time.

Velocity charts should be distinct from accumulated-impact charts.

## 5. Normalization

Research fields differ dramatically in publication volume, citation behavior, and attention patterns.

Possible normalization dimensions:

- field,
- year,
- age,
- artifact type,
- size of active research community.

Preferred representation:

```text
percentile within field + cohort + period
```

rather than arbitrary universal raw thresholds.

## 6. Certifications

Coulomb may recognize durable observable influence through certification tiers.

Candidate labels:

- Bronze
- Silver
- Gold
- Platinum
- Diamond

Certification does not mean:

- scientifically true,
- endorsed by Coulomb,
- free of criticism,
- superior in all dimensions.

It means only that a declared observable threshold has been crossed.

## 7. Multi-Dimensional Certification

Avoid one universal certification.

Possible forms:

```text
Gold — Reach
Gold — Reuse
Gold — Replication
Gold — Generativity
Gold — Influence
```

Example artifact profile:

```text
PLATINUM — Reuse
GOLD — Generativity
SILVER — Reach
Support — Contested
```

This communicates more than one compressed prestige score.

## 8. Threshold Models

Possible threshold approaches:

### Absolute

Example:

```text
1,000 verified downstream reuses
```

Easy to understand, but poor across fields.

### Percentile

Example:

```text
Gold     = top 1% within field/year cohort
Platinum = top 0.1%
Diamond  = top 0.01%
```

Numbers above are illustrative, not normative.

Percentile-based systems are likely better for heterogeneous research domains.

### Hybrid

Require:
- minimum absolute signal,
- minimum age,
- and percentile threshold.

This may reduce distortion in very small fields.

## 9. Research Collectives

A research collective may contain:

- humans,
- agents,
- organizations,
- temporary collaborators.

Collectives may accumulate portfolio recognition through their artifacts.

Example:

```text
Coulomb Intelligence Research

2× Platinum
5× Gold
11× Silver

Top 100 appearances: 27
```

This supports identity and community formation without making individual academic rank the primary prestige unit.

## 10. Agent Contributions

Charts should preserve mixed human-agent contribution provenance.

Recognition may be attributed to:

- artifact,
- collective,
- contributing actors,
- agent lineage,
- research program.

The system should avoid pretending that only humans contributed where agents performed material work.

## 11. Relevance Marketing

Research Charts provide a public interface that answers questions such as:

- What research is moving right now?
- Which old work is unexpectedly becoming relevant again?
- Which paper is being replicated most?
- Which research artifact is generating the most follow-up work?
- Which criticism changed an active research program?
- Which open problems are attracting new resources?

This makes research easier to encounter outside conventional academic discovery channels.

## 12. Cybernetic Role

Charts are sensors on the research system.

Possible automated responses:

```text
unusual attention spike
  → deploy Critic daemon

rapid reuse
  → schedule replication review

high dependency centrality
  → perform provenance audit

high controversy
  → generate competing-model synthesis

high generativity
  → study successful research pattern
```

Public relevance signals can therefore feed directly back into research quality.

## 13. Goodhart Resistance

Anti-gaming measures should include:

- transparent metric definitions,
- multiple dimensions,
- rotating or evolving chart functions,
- field normalization,
- verified provenance,
- separation of attention from support,
- independent recomputability,
- anomaly detection,
- no hidden master score.

Chart algorithms themselves should be versioned research objects.

## 14. Design Maxim

> Celebrate what moves research forward, not merely what accumulates attention.