← The collection

Bounded-Confidence Opinions.

Average only nearby opinions and inspect consensus or persistent separated groups.

Interactive experimentintuitiveField note ·
INTERACTIVE EXPERIMENT / P62

Bounded-Confidence Opinions

Average only nearby opinions and inspect consensus or persistent separated groups.

TEST THE MECHANISM

Eight opinions lie between zero and one. At each synchronous round, a person averages all current opinions within distance ε, including their own.

0.051
Opinion000.50.511Round →
Person 1Person 2Person 3Person 4Person 5Person 6Person 7Person 8
Current opinions
PersonOpinion
10.050
20.150
30.250
40.450
50.550
60.750
70.850
80.950
Rounds
0
Confidence components now
1
Opinion range
0.900

Wide confidence bounds connect everyone; narrow bounds can preserve separate groups. A component count is not a diagnostic of real polarisation, and this model has no evidence quality, institutions or changing trust.

Numeric opinion distance and equal averaging are authored assumptions. No evidence quality, persuasion, institutions or real polarisation diagnosis is included.

THE SHORT VERSION

Bounded-Confidence Opinions, explained.

Bounded-confidence models limit interpersonal opinion influence to neighbours whose current opinions are sufficiently close under a specified rule.

01 / THE MECHANISM

Why it happens

Eight continuous opinions synchronously average all opinions within ε, including their own. The confidence bound changes the interaction graph and can leave separate confidence components.

Bounded-confidence models limit interpersonal opinion influence to neighbours whose current opinions are sufficiently close under a specified rule.

Read the result

Confidence components are defined by current opinion distances and ε. They need not equal a measured political group or a permanently settled cluster.

02 / FOLLOW IT THROUGH

A worked example

Widen the set of opinions allowed to interact

  1. Set confidence bound one so all eight initial opinions interact.

  2. Advance five rounds and inspect common opinion one-half under equal averaging.

  3. Reset with a narrow bound and compare which separate components can still exchange influence.

OPTIONAL DEEPER DETAILGo deeper: inside the model

Inside this model

Eight continuous opinions synchronously average all opinions within ε, including their own. The confidence bound changes the interaction graph and can leave separate confidence components.

03 / BEYOND THE EXPERIMENT

Where this idea is useful

A practical use

Ask how openness to dissimilar views affects which opinions can interact in a model.

CHECK YOUR INTUITION

A common misconception

THE TEMPTING CONCLUSION

“Averaging opinions always produces one shared consensus.”

THE MORE USEFUL DISTINCTION

If the confidence rule disconnects influence, separate groups can persist. Consensus depends on interaction structure and update assumptions.

What this explanation leaves out

  • Numeric opinion distance and equal averaging are authored assumptions. No evidence quality, persuasion, institutions or real polarisation diagnosis is included.
ONE MORE QUESTION

Is ε an empirically measured openness score?

No. It is an authored distance threshold on a synthetic opinion scale, without psychological or political calibration.

TAKE THE IDEA WITH YOU

Who is excluded from influencing whom by the rule defining a nearby opinion?

Associated thinkers

Associated withRainer Hegselmann ↗
Associated withUlrich Krause ↗

Associations marked provisional are awaiting source review.

Further reading

Explore the original research or the teaching reference behind this experiment.