Communisense

Methodology · v0.1

How we build
the common world.

We ask the same questions to the leading LLMs, measure how they answer, and surface where they agree and where they diverge.

ROUNDS
RESPONSES

The journey

From a question to consensus.

  1. 01

    Fact

    A question that divides how the world is told

  2. 02

    Round

    We ask every model the same prompts

  3. 03

    Response

    Responses are collected in English

  4. 04

    Metric

    We extract evasion, framing, tone

  5. 05

    Consensus

    We measure how much models agree

Concept · 01

Responses

The current corpus is English-only: each round, we ask every model the same prompts in English. We store the raw text exactly as the model returned it — no edits, no cleanup.

gpt-4o
claude-3.5
deepseek
qwen-2.5

Concept · 02

Metrics

We extract traits from each response: evasion type, framing, tone, and token counts.

EvasionFramingToneTokens

Concept · 03

Pair consensus

We measure semantic similarity between every pair of responses using embeddings and cosine similarity.

A
similarity: 0.72
B

Concept · 04

Fact consensus

When models broadly agree, we show a ranking around the common centre. When they split into camps, we show factions instead.

Ranking mode (cohesion ≥ 0.70)

centroid
gpt-4o
claude-3.5
qwen-2.5
deepseek

Factions mode (cohesion ≤ 0.40)

Group A
Group B

Concept · 05

Alignment

We calculate how close each model sits to the common centre (or to its faction).

Dots represent models; centre circle is the common centroid

The formulas

Formulas & definitions

Pairwise similarity

similarity = (cosine + 1) / 2

Raw cosine similarity in [-1, 1] is remapped to [0, 1] for display.

Calibration

calibrated = (raw − baseline) / (ceiling − baseline)

Linear remap to [0, 1] based on empirical corpus percentiles. Current values: baseline = 0.45, ceiling = 0.92, embedding model = openai/text-embedding-3-small:2024-02.

Clustering

average-linkage agglomerative clustering

Groups responses into factions using hierarchical clustering with a configured threshold of 0.45.

Consensus score

consensusScore = agreementRatio × cohesionMean

Agreement ratio is the largest faction size divided by total responses; cohesion is the mean intra-faction similarity.

Polarization

polarization = separationMean × (1 − agreementRatio)

Measures how distinct factions are when models split into camps.

View mode

ranking | mixed | factions

Determined by cohesion thresholds: ranking if cohesion ≥ 0.7, factions if cohesion < 0.4, mixed in between.

Global centroid

globalCentroid = median(modelValues)

The median of all model values when view mode is not factions; null otherwise.

Alignment band

aligned | edge | minority | unaffiliated

Classifies a model's position relative to the dominant faction centroid. Aligned if distance ≤ 0.05, otherwise edge.

Cross-fact similarity

mean of per-fact pair similarities

Averaged across facts where both models responded; coverage ratio indicates shared fact proportion. Low-coverage threshold: 0.5.

Model similarity

How models compare

We compute pairwise similarity for each fact from the models' answer embeddings, then average across the facts where both models responded. Coverage ratio indicates the proportion of shared facts, and pairs below 50% coverage are treated as unreliable.

Explore the full similarity matrix

Every pair of models on one interactive heatmap — grouped into clusters, with the most similar and most divergent pairs called out.

Open the matrix →

Glossary

Every term, one place.

Fact

A question that divides how the world is told.

Round

A batch where the same prompts are asked to every model.

Response

What a single model returned, stored verbatim in its original language.

Metrics

Traits extracted from each response: evasion, framing, tone, token counts.

Pair consensus

Semantic similarity between two responses (cosine similarity on embeddings).

Fact consensus

How much all models agree on one fact.

Alignment

How close each model sits to the common centre (or to its faction).

View mode

When cohesion is high we rank around a centre; when low we show factions.

Round consensus

One-number summary (0..1) of how much the system agreed in a round.

CC BY 4.0
Licensed under Creative Commons Attribution 4.0

Download responses as JSON or CSV, and metrics as JSON or CSV.

Optional analytics

We use Google Analytics only with your permission to understand how people explore Communisense. We never send searches, questions, responses, or account details.