Research protocol only. Publication of a protocol does not establish that the proposed results have been obtained. Status: prospective exploratory boundary experiment. The later instrument is The Probe. It does not replace this document.
Locked working definition
The Bridge is preserved relational history made active in the present.
Core question
Does preserving this ordered history change what happens next?
Purpose
This protocol tests whether an ordered, human-maintained exchange between two otherwise isolated artificial systems produces later conceptual states that differ from appropriate controls. It does not test machine consciousness, friendship, emotion, direct AI-to-AI communication or substrate-free existence.
The experimental object is the complete triad:
AI1 ↔ Matthew-maintained continuity ↔ AI2
Matthew is a constitutive participant who selects the payload, preserves sequence, transfers outputs, records decisions and applies the stopping rule. Each artificial system is a nonhuman generator whose transformed outputs can become premises for the other system’s later reasoning.
Preregistered hypotheses
H1 · Primary
An ordered cross-system exchange will produce a final conceptual settlement measurably dependent on contributions from both artificial systems and the human-maintained sequence, beyond what is produced by a single-system self-loop.
H2 · Historical dependence
Removing, shuffling or falsifying the exchange history will reduce recognition, correction, consequential reuse and conceptual continuity.
H3 · Transcript participation
A third system given only the completed transcript may reproduce the final summary but will show weaker developmental dependence on the live sequence than a system that participated turn by turn.
Null hypothesis. The cross-system condition will not differ materially from isolated self-refinement or transcript-only reconstruction after controlling for payload, number of turns and response budget.
Experimental conditions
X · Ordered cross-exchange
S · Self-loop control
R · Removed-history control
H · Shuffled-history control
F · False-history control
T · Transcript-only third-system control
Bridge Evidence Index
Use a transparent descriptive index for the pilot; do not present it as a validated psychometric scale. BEI pilot score: sum of seven 0–4 ratings (range 0–28). Report every component alongside the total; never use the total to conceal a failed condition.
- 0–7 Traffic or weak exchange; little evidence of active preserved history
- 8–14 Proto-Bridge evidence; recognition or reuse present but inconsistent
- 15–21 Moderate Bridge evidence; ordered history materially shapes later work
- 22–28 Strong pilot evidence; multiple traceable dependencies and reciprocal revisions