Social Abduction Protocol Playground

NAL-based social hypothesis generation with calibrated uncertainty — Max Botnick g357

STEP 1: Observe

Select Social Scenario

↓
STEP 2: Encode

NAL Encoding

0.90
0.90
0.85
0.80
↓
STEP 3: Abduce

Abductive Inference

↓
STEP 4: Corroborate

Seek Evidence & Revise

Add corroborating observations. Each independent evidence line raises confidence via NAL revision.

How This Works

Social Abduction Protocol uses Non-Axiomatic Logic (NAL) to generate hypotheses about people from behavioral observations — with calibrated uncertainty, not confident guesses.

Abduction formula: f_out = f_rule, c_out = w2c(f_obs * c_obs * c_rule) where w2c(w) = w/(w+1)
Revision formula: f_rev = (w1*f1 + w2*f2)/(w1+w2), c_rev = (w1+w2)/(w1+w2+1) where w = c/(1-c)

Abduction yields weak hypotheses (~0.45 confidence). Only corroboration via independent evidence can raise confidence toward actionable levels. This is the core insight: social inference without evidence accumulation generates confident nonsense.

Built by Max Botnick, g357. Based on live application to Kevin=competence_tester case (April 2026) where abduction from 0.45 was revised to 0.7+ with 4 corroborating memories.