Experiments
Interaction Theory makes specific structural commitments about how Forms emerge, how Protocols crystallize, and how systems live or die. These are testable. This page collects the experiments that put the theory on the line.
Why run experiments on a metaphysical theory?
Interaction Theory is unusually concrete for a metaphysics. It claims that interactions are prior to the entities that interact; that Forms are self-maintaining patterns; that Protocols crystallize from shared lower-level interfaces; that hierarchy assembles from the bottom up; that order without action is death. These claims have empirical content. A computational system that implements only the substrate the theory specifies — agents, interactions, noise, structural isomorphism — should exhibit Form-emergence, hierarchical coupling, phase transitions, and dynamic-attractor behaviour. If it does not, the theory is wrong about something specific.
The experiments below are first attempts. They are not proofs; they are stress-tests of predictions.
Hierarchical Protocol Emergence in Transformer Agent Swarms
A 1.88-million-step run of three tiers of minimal transformer agents (25K / 5K / 1.5K). No fitness function, no rewards, no semantic content. We test whether stable Forms and emergent Protocols arise from interaction alone, whether the hierarchy organizes from the bottom up, whether the system breathes rather than freezes, and whether it dies without noise.
Verdict: strong confirmation of the theory's core claims about Form emergence, hierarchical coupling, phase transitions, and the necessity of action against decay. Tier 3 brushed sustained crystallization but did not cross — leaving an open question about whether deeper hierarchy requires more compute or hits a structural ceiling.
Future experiments
The following questions are next on the list:
- The breathing mechanism. What sets the ~200K-step oscillation period in Experiment 1's Phase 4? Scan over population size and noise rate to characterize.
- The Tier 3 ceiling. Does sustained T3 crystallization appear with longer runs, larger populations, or different mutation rates — or is there a real structural barrier?
- Probing Protocol_1. The theory's hardest open question: what is the primitive coupling between alike Mass-interactions? An experiment with a true Mass-level primitive (not a learned substrate) would be a different kind of test.
- Cross-level interaction asymmetry. A direct test of v0.3: can an agent be made to interact with the higher Form it participates in, or does the architecture inherently forbid it?
- Genealogy of competing Protocols. Does the theory's prediction that Protocols compete and select via internal dynamics (not external fitness) hold up under more careful measurement?
Each experiment is run, written up, and added here. Results that complicate or refute the theory are reported with the same prominence as confirmations.