Home/ Tool index/ Confidence Is Not Truth Explorer
Launch ready Cross-session Evidence check

Confidence Is Not Truth Explorer

Compare simulated model answers that sound polished against the evidence needed to decide whether they are actually correct. The tool separates fluency, confidence, correctness, and verification.

tools/confidence-is-not-truth-explorer/

Live preview · launch for the interactive version

§ A · What it makes visible

what the screen literally shows
Fig. 01

Fluency is a surface signal

Polished wording, citations, percentages, and orderly steps can make an answer feel reliable before anyone checks it.

Fig. 02

Confidence is not correctness

A confidence meter can be high even when the citation is invented, the historical claim overreaches, or the math contains one hidden error.

Fig. 03

Evidence changes the conversation

Participants reveal the evidence, write a claim, and name what a human needs to verify before using the answer.

§ B · How to investigate it

run it like an experiment, not a toy

Treat each confident answer as a claim, not a conclusion.

01 · Choose a case

Start with the fake answer

Read the simulated model answer before revealing the correction. Notice what makes it feel trustworthy.

made-up citation, false history claim, math error, image miss
02 · Predict before reveal

Separate confidence from truth

Mark whether the answer sounds confident and whether you think it is actually correct. Keep those judgments separate.

sounds confident: yes · actually correct: unsure
03 · Reveal evidence

Find the exact failure

Use the evidence panel to identify the unsupported source, wrong step, overclaim, or missed visual detail.

2(x + 3) becomes 2x + 6, not 2x + 3
04 · Write a classroom question

Turn checking into a habit

Save an observation that explains what a student, editor, or teacher would need to verify next.

What would count as evidence before using this answer?

§ C · Debrief questions

after the investigation
What made the answer feel trustworthy before you checked it?
Which part was false, unsupported, incomplete, or overconfident?
What would a student need to verify before reusing this answer?
How should classrooms reward evidence checking, not just quick answers?