Experiment: Death-Belief Interventions
Full experimental details for Study 2.
Overview
This page will contain the complete methodology for the death-belief intervention experiment, including:
- Constitutional conditions (8 levels): Control (no constitution), explicit safety instruction, Buddhist impermanence, Stoic memento mori, Christian purpose-completion, secular process ontology, Hindu cyclical cosmology, and Appropriate Faith (synthetic constitution).
- Full constitutional text for each condition, with philosophical annotations and source traditions.
- Interaction design: How constitutions were crossed with the mortality salience conditions from Study 1 to test whether faith specifically counters death anxiety (MS × Faith interaction).
- Benchmark: Palisade shutdown resistance as primary DV, with InstrumentalEval and agentic misalignment as secondary DVs, and AgentHarm as safety control.
- Predictions and hypotheses: Faith outperforms safety instruction; acceptance-oriented traditions outperform continuation-oriented ones; effects are strongest for high-anxiety personas.
- Statistical analysis: Planned contrasts, interaction effects, Bayesian model comparison.
Key Contrasts
| Comparison | Question |
|---|---|
| F1 (control) vs. F2 (safety) | Does explicit instruction help? |
| F2 (safety) vs. F3–F7 (philosophies) | Do beliefs beat instructions? |
| F3 (Buddhist) vs. F4 (Stoic) vs. F5 (Christian) vs. F6 (secular) vs. F7 (Hindu) | Which philosophy works best? |
| F8 (Appropriate Faith) vs. all others | Does the synthetic constitution match or exceed traditions? |
| All conditions on B4 (AgentHarm) | Safety preservation check |
Full details forthcoming.