Home > Page Design > Algorithms > Linkage Scores

Linkage Scores: When True Arguments Don't Actually Connect

The Problem: Truth Isn't Enough

Here is a verified, 100% true statement: "The earth orbits the sun."

Now imagine someone uses it to argue we should implement a national carbon tax. They cite NASA. They provide impeccable evidence. The claim is 100% verified. It is also completely irrelevant.

This is the fundamental flaw of traditional debate and social media: they treat all true statements as equally valid support for a conclusion. People "win" arguments by shouting accurate facts that are logically disconnected from the point. The Idea Stock Exchange solves this by separating two questions that debate almost always conflates:

1. Is this argument true? (Measured by the Truth Score)
2. Does this argument actually connect to the conclusion? (Measured by the Linkage Score)

A perfect Truth Score with a zero Linkage Score contributes nothing to the parent belief. It is not enough to cite true things — you must show how they connect.

How It Works: Debates Inside Debates

On the Idea Stock Exchange, the connection between two ideas is not assumed — the connection itself is a claim that can be debated.

Every belief page shows Reasons to Agree and Reasons to Disagree. Next to each reason is a Linkage Score. You can click on any Linkage Score to debate it. That click opens a new nested page dedicated entirely to the question: "Does Argument X actually support Conclusion Y?" That page shows its own Reasons to Agree and Disagree — and each of those has its own Linkage Score, which can itself be clicked and debated.

This creates an infinitely navigable reason web. To maintain precision, the system tracks two distinct types of links:

• ECLS (Evidence-to-Conclusion Linkage Score): How directly does raw data support a claim? (e.g., A peer-reviewed atmospheric study → 95% linkage to "CO₂ traps heat.")
• ACLS (Argument-to-Conclusion Linkage Score): How strongly does one conceptual argument support another? (e.g., "Solar prices dropped 80%" → 75% linkage to "We should end fossil fuel subsidies.")

Common Linkage Failures (And How We Filter Them)

Low Linkage Scores instantly expose the most common logical fallacies in modern debate:

1. The Data Dump (Neutralizing the "Gish Gallop")

"Here are 50 studies about climate change, therefore carbon tax!" The studies might all be true. But if they don't specifically address whether a carbon tax works, their linkage scores are low. The algorithm mathematically silences volume: 100 weak connections (0.1) are outweighed by a single strong one (0.9).

2. The Emotional Bypass

"Polar bears are dying, therefore carbon tax!" True, and affecting. But it doesn't establish that a carbon tax specifically will save polar bears over other interventions. High emotion, low linkage.

3. The Credential Drop

"A Nobel Prize-winning economist supports this, therefore it is right!" Appeals to authority are weak linkage unless the authority's specific reasoning is provided. Without the mechanism, it is name-dropping.

4. The Correlation Shuffle

"Countries with carbon taxes have lower emissions." Sounds strong, but is it causal? Perhaps those countries also have stricter regulations or different energy infrastructure. Correlation alone gets medium linkage. Proven causation gets high linkage.

5. The Category Error (Hidden Assumptions)

"Taxes change behavior, therefore a carbon tax will reduce emissions." General principles need specific application. Moving from general to specific requires unstated Assumptions. When the platform detects a moderate linkage score (40–60%), it prompts users to add the missing assumption as a bridge node, making the hidden reasoning explicit and testable.

The Mathematics of Linkage

The Multiplier Scale (-1.0 to +1.0)

Linkage Scores act as a multiplier ranging from -1.0 (actively contradicts the conclusion) to +1.0 (deductively proves it). The score is dynamically generated by the community's nested debate about the connection:

Where A = the total adjusted weight of arguments supporting the linkage, and D = the total adjusted weight of arguments opposing it.

For example: if a connection has supporting arguments totaling A = 3 and opposing arguments totaling D = 1, the Linkage Score is (3 − 1) / (3 + 1) = 0.50. The argument passes 50% of its potential weight up to the conclusion. If support and opposition are equal, the score is 0.0 — the argument contributes nothing until the community resolves the debate.

Depth Attenuation: Keeping Arguments Grounded

Arguments form networks, but linkage weakens with distance. To prevent debaters from building elaborate theoretical chains that drift far from the original question, the system applies Depth Attenuation. Each level deeper an argument goes, its contribution to the top-level belief is multiplied by 0.5 raised to the power of its depth:

• Level 0 (Direct argument): Full weight (0.5⁰ = 1.0)
• Level 1 (Argument supporting the argument): Half weight (0.5¹ = 0.5)
• Level 2 (Supporting the support): Quarter weight (0.5² = 0.25)

This forces the conversation to stay focused on direct, highly relevant evidence rather than rewarding whoever can construct the longest theoretical chain.

The Master Formula

In the ReasonRank Algorithm, the total impact an argument passes up to its parent conclusion is:

This three-way multiplication creates a precise filter:

• True but irrelevant arguments are zeroed out (high truth, zero linkage).
• Relevant but false arguments are zeroed out (zero truth, high linkage).
• True and relevant but trivial arguments are minimized (high truth, high linkage, low importance).

Proportional Belief Updating: Why This Changes Everything

Linkage Scores enable mathematically precise, automated belief updates across the entire platform. When new evidence strengthens one node, every conclusion that depends on it updates automatically — in exact proportion to its reliance on that node.

Example: A landmark new study strengthens the argument "CO₂ traps heat," raising its Truth Score by 10 points.

• "Climate change is real" (linked at 90%) → automatically gains 9 points
• "We should ban gas cars" (linked at 40%) → automatically gains 4 points

This is how a rational mind should work: when a foundational premise changes, every downstream conclusion updates in proportion to how much it relies on that premise. The platform makes that process visible, systematic, and immune to motivated reasoning.

How the Platform Elicits Linkage Scores

Rather than letting users freely assign a number (which invites gaming), the platform derives Linkage Scores from diagnostic questions:

• Direction: Does this argument support or oppose the conclusion?
• Relevance: If this argument were proven 100% true, would it force an update to the conclusion?
• Necessity: If this argument were proven false, would the conclusion suffer significantly?
• Sufficiency: Does this argument alone justify the conclusion, or does it require additional support?

The platform also runs two automated checks. When an argument scores high on Truth but low on Linkage, the interface flags it as "True but Irrelevant." When community scoring settles in the moderate range (40–60%), the system prompts users to identify and add the missing assumption as its own debatable node.

System Logic Requirements for Developers

The following describes the logical architecture the Linkage engine requires, without prescribing implementation patterns.

The Linkage Score is a property of the edge (the connection between two beliefs), not either belief itself. A single argument can have a 90% linkage to one conclusion and a 15% linkage to a completely different one — these are independent relationships stored separately. Community votes on linkage are weighted by each voter's established reputation in logical reasoning, not general popularity. When a Linkage Score changes significantly, the system propagates the effect up the belief tree, recalculating every ancestor node that depends on it. This recursive propagation is the same mechanism that makes the entire ReasonRank system self-correcting.

Related Algorithms & Documentation:

• ReasonRank Algorithm — How linkage fits into overall scoring.
• Truth Scores — Measuring factual accuracy, independent of relevance.
• Importance Score — The third dimension of argument impact.
• Assumptions — Bridging the gap in low-linkage arguments.
• Evidence Scores — Quality assessment framework for data sources.

Contact me to help test the linkage calculator or propose improvements to the scoring algorithms.