Scaffolded Introspection: A Methodology for Eliciting and Measuring Self-Referential Behavior in Large Language Models | Synapse