P/G Engine
Experimental Platform
P/G Engine is an experimental implementation of specified P/G core mechanisms for controlled studies.
Its frozen computational core includes state updates, relational dynamics, affect, valuation, execution constraints, and decision/search.
Current scope
A relational architecture for controlled experiments
The frozen P/G core has been implemented end to end, including property and relational updates, affect, valuation, execution constraints, and the specified decision/search path. Closed-loop synthetic verification has been completed under benign and adversarially perturbed conditions.
These results establish implementation coherence under the tested synthetic conditions; they do not constitute human behavioral validation.
Implemented components
Specified state and updates
The experimental core represents declared property and relationship dimensions and computes updates from represented events. Versioned schemas and parameters define the scope of each experiment.
Implemented components
Traceable affect contributions
The affect computation retains signed contributions with their object and structural origin. An aggregate readout can be compared with judgments under a specified observation model.
Trust is a relational dimension; reputation is a property representation. Derived emotions such as gratitude and fear belong to a separate interpretive layer.
Further development
Authoring and response layers
The current experimental implementation uses default personality parameter tables. Big Five authoring, personalization layers, the witness-related extension (E4), and deployment hardening and interfaces remain future work.
Verification and evidence
Implementation checks and empirical evaluation
Specification-based checks ask whether the code realizes the intended dynamics. The completed registered CTT cycle tested computational predictions and model discrimination; R2-2 remains a documented limitation of the frozen mechanism set. This work is separate from the prepared Stage B human protocol, which has not started and would compare specified model predictions with human affect judgments.
Research applications
Where the framework may be tested
Game agents, human–robot interaction, and economics pose different research questions. Each direction requires its own models and evaluation.
Experimental platform
Demo coming soon
A real demo link will be added here when available.