add ADARD logic engine selection study #16

Merged

lost-rob0t merged 1 commit from agent/add-probabilistic-logic-engine-study into main

2026-07-20 02:16:29 +00:00

lost-rob0t commented

2026-07-20 02:16:10 +00:00

(Migrated from github.com)

What changed

adds a bounded ADARD research track for selecting logic engines by situation
requires investigation of ProbLog, N-Prolog, SWI-Prolog, XSB/PITA, cplint, DeepProbLog, s(CASP), and Soufflé
defines shared comparison criteria, benchmark fixtures, routing outputs, and acceptance criteria
keeps the existing ADARD loop active and does not create a second loop or select production dependencies prematurely

Why

ADARD will need different reasoning semantics for deterministic rules, probabilistic evidence, recursive tabled inference, defaults and exceptions, neural-symbolic predicates, and bulk Datalog closure. The repository needs an explicit selection study rather than forcing one Prolog implementation into every role.

Validation

branch is one commit ahead of main and contains one new Org research file
the file includes a stable Org-roam ID, related-node links, status, glossary, source list, situation matrix, experiments, and acceptance criteria
no implementation design or production dependency was activated

Closes #15 only when the broader ADARD controller is complete; this PR does not close it.

## What changed - adds a bounded ADARD research track for selecting logic engines by situation - requires investigation of ProbLog, N-Prolog, SWI-Prolog, XSB/PITA, cplint, DeepProbLog, s(CASP), and Soufflé - defines shared comparison criteria, benchmark fixtures, routing outputs, and acceptance criteria - keeps the existing ADARD loop active and does not create a second loop or select production dependencies prematurely ## Why ADARD will need different reasoning semantics for deterministic rules, probabilistic evidence, recursive tabled inference, defaults and exceptions, neural-symbolic predicates, and bulk Datalog closure. The repository needs an explicit selection study rather than forcing one Prolog implementation into every role. ## Validation - branch is one commit ahead of `main` and contains one new Org research file - the file includes a stable Org-roam ID, related-node links, status, glossary, source list, situation matrix, experiments, and acceptance criteria - no implementation design or production dependency was activated Closes #15 only when the broader ADARD controller is complete; this PR does not close it.