----------------------
Screening Trust Worlds
----------------------

This directory contains a single Notation3 file that encodes a case 
study about the reliability of positive diagnostic test results in four 
different “worlds” of reasoning. Each screening scenario is 
described by a disease prevalence, a test sensitivity and a test 
specificity. The reference world W0 applies Bayes’ rule to compute 
the positive predictive value (PPV) and only “trusts” a positive 
result when PPV is high enough. W1 behaves like a naive rule-of-thumb 
that looks only at sensitivity and ignores prevalence and false 
positives. W2 uses a simple heuristic that demands minimum prevalence 
and high specificity before trusting positives. W3 is a cautious 
version of W0 that adopts a slightly higher PPV threshold, so it treats 
some borderline positives as unreliable.

From these inputs the core rules derive, for each scenario and world, 
whether a positive result is trusted or rejected. An ARC layer then 
derives an Answer node that summarises how the worlds agree and 
disagree on four illustrative scenarios (from a very rare disease in 
mass screening to a moderately common condition tested with different 
specificities), a Reason node that explains how the different rules 
relate to Bayes’ formula, and five Check nodes that confirm 
relationships such as “W1 is the only world that trusts positives in 
very rare mass screening” and “W3 is at least as cautious as W0 on 
these examples”. All ARC triples are derived from the same numeric 
facts and thresholds as the core rules, so if you change the scenarios 
or the worlds’ rules, the Answer, Reason and Checks will evolve 
accordingly.
