Cognitive Bias in the ISE
Every belief carries reasons, evidence, values, and interests. It also carries predictable cognitive biases, the mental shortcuts that bend reasoning even when everyone is acting in good faith. The ISE scores those too.
A bias here is a first-class object, like an argument or an obstacle to resolution: it gets its own page, its own scores, and its own links to the beliefs it distorts, which feed the Default Sorting Score (DSS). Build a new bias page from the Bias Template.
🔎 How the ISE Flags a Likely Bias
Bias labels are not assigned by intuition. Five structured signal sources raise or lower the confidence that a bias is operating, and each is itself scored.
| Signal Source | What It Measures | Feeds |
|---|
| ReasonRank of the arguments |
Weak evidence, emotional appeals, and unfalsifiable claims on one side raise bias likelihood for that side. |
Argument scores |
| Bias patterns inside arguments |
Specific tells: absolutist language, us-versus-them framing, anecdote over base rate, exaggerated small risks. |
Detection signals |
| Centrality in movement media |
When core books and leaders lean on moral panic or identity threat, supporter bias likelihood rises. |
Content analysis |
| Everyday explanations |
Do ordinary supporters cite evidence, or fall back on slogans, identity, and anecdotes? |
Prevalence estimate |
| Historical base rates |
Some categories reliably trigger known biases across cultures (e.g. threat bias in crime policy). |
Confidence score |
Each bias gets its own page built from the Bias Template: detection signals, the beliefs it affects, evidence that it is real, and what reduces it. Pages are being built; unlinked entries are next in the queue.
| Bias | Family | The Tell |
|---|
| Confirmation bias |
Motivated reasoning |
Searches for, and remembers, only what confirms the prior. |
| Availability bias |
Heuristic |
Vivid anecdotes outweigh statistics. |
| Motivated reasoning |
Motivated reasoning |
Reasons toward a wanted conclusion, not toward the truth. |
| Outgroup / tribal bias |
Social |
Us-versus-them framing replaces the merits. |
| Loss aversion |
Heuristic |
Small risks get exaggerated; gains get discounted. |
| Overconfidence |
Heuristic |
Absolutist language: "never," "always," "obviously." |
| Status-quo bias |
Heuristic |
The current arrangement feels normal, so change feels radical. |
| In-group favoritism |
Social |
One's own side's norms are treated as the default reasonable view. |
🎯 Scoring a Bias: the DSS
A worked example on the belief "Immigrants increase crime." Each side carries its own biases, scored separately.
| Biases Affecting Supporters | Biases Affecting Opponents |
|---|
| Bias | DSS | Bias | DSS |
|---|
| Outgroup threat bias |
80 |
Motivated reasoning |
60 |
| Availability bias |
75 |
In-group favoritism (cosmopolitan norms) |
55 |
| Confirmation bias |
70 |
Status-quo bias |
45 |
DSS = (confidence the bias applies to the typical holder of that side) × (confidence the behavior is a bias in this scenario), on a 0-100 scale, weighted by ReasonRank signal strength. It maps predictable distortions; it does not judge morality, and it never feeds back into the belief's own truth score. DSS supports belief sorting, not truth scoring.
⚖ A Bias Accusation Is Itself an Argument
This is the firewall against "bias" becoming a rhetorical weapon. Claiming the other side is biased is not a trump card. It is a claim that earns its weight the same way every other claim does, by being scored.
| What a bias score does | What it never does |
|---|
| Adds a scored, challengeable claim that a specific distortion is operating, with its own evidence and linkage. |
Automatically slash the other side's credit. There is no auto-penalty for an alleged bias. |
| Gets weighed against the argument it targets, so a strong argument from a biased source still stands on its merits. |
Depend on who is speaking. Identity does not change what an argument is worth. |
📊 What Bias Scores Are Good For
Mapped biases make disagreement legible. They show why two people read the same evidence differently, separate differences in strength from differences in logic, flag beliefs that need better evidence, and point toward fairer conflict resolution. They integrate with truth scoring, linkage scores, assumptions, interests, and obstacles. Bias mapping does not shame anyone; every brain takes shortcuts. It surfaces the shortcuts so we stop talking past each other.
🔢 Quality and Confidence
A bias score is only as good as its inputs. Every score is held to the same standard, and low-confidence scores are marked and down-weighted in belief sorting.
| Requirement | What It Means |
|---|
| Statistical inputs |
Representative sampling of both sides; content analysis of movement media; validated pattern-matching. |
| Transparency |
Public confidence intervals, documented criteria, open community challenge, evidence-tier requirements. |
| Continuous learning |
Scores update as new data arrives; user feedback and cross-cultural testing refine detection. |
🔗 Related Concepts
Bias-adjacent topics with their own pages: hate, bigotry, prejudice, intolerance, gays, lesbians.
📬 Contribute
Maintained by ~Myclob. Contact me to help develop bias-detection methods or refine the DSS.
See the framework on GitHub for the scoring algorithms and technical detail.
Comments (0)
You don't have permission to comment on this page.