Sign in

Use your API key (starts with rapu_usr_).

Couldn't sign in with that key. Check it and try again.
The workshop reads calibration aggregates from your brain via /api/epistemic.
Calibration

Calibration

The honest readout of how well your brain's commitments survive contact with reality. Predictions you've claimed a probability for are binned, scored against actual outcomes, and aggregated by owner and by atom property. Lower Brier = better calibrated. 0.25 is what an uninformative coin-flipper scores.

Reliability diagram

axis #17 · calibration_reliability
Each bin: claimed probability vs realized outcome. The diagonal is perfect calibration. Below the diagonal means over-confident at that level; above means under-confident.
Loading…

Per-owner win rate

axis #6 · owner_calibration
Win rate = (succeeded + 0.5 × partial) / (succeeded + partial + failed). Retracted predictions are excluded — walking back a commitment is intellectual honesty, not failure.
Loading…

Atom reputation distribution

axis #6 · atom_calibration
Aggregate reputation_score across atoms, grouped by a property. Atoms that consistently feed projections whose chosen direction succeeds accumulate positive reputation; atoms that consistently steer toward losing alternatives accumulate negative.
Loading…

Confidence calibration

axis #11 · confidence_calibration
For each atom-confidence level (1-5), what's the average outcome of the predictions those atoms fed into? Perfectly-calibrated confidence-5 atoms produce 0.95-rate outcomes; over-confident teams see flat lines.
Loading…

Counterfactual retraction

axis #22 · counterfactual_retract
"If I retract this atom right now, what would shake loose?" — combines forward-chain (artifacts that depend on the atom) with grounded-extension recompute (atoms whose acceptance status would flip). Pure simulation; nothing is changed.
The atom can be in any brain you have read access to.

Cross-brain consensus

axis #20 · cross_brain_consensus
Across the brains this brain reads from via active brain_links, where do they agree on a claim vs disagree? Pairs above min_consensus_sim are consensus; pairs between min_disagree_sim and consensus are topical overlap — most decision-relevant when the pair differs on standpoint or atom type.
Enter a topic and search. Within-brain pairs are excluded — that's what detect_conflicts covers.

Argumentation graph

axes #18 + #19 · atom_relations + grounded extension
Atoms placed on three concentric rings by Dung-grounded status: accepted (defensible, inner ring) · undecided (locked in cycles, middle) · rejected (attacked by accepted, outer). Red arrows are attacks; green arrows are supports. Hover an atom for its body and confidence.
Loading…