third-button-obstruction
leaf/assumption · drafted
Statement
When every instance of an agent trusts the others to follow a shared plan, some instance can be a defector who spoils the plan even though all instances share goals and beliefs, so message-based self-modification needs a stability assumption beyond shared goals.
Source
projects/deference/faf-cleanroom-2026-10-02/sources/references/bli/understanding-trust/udt-tiling-working-notes-2025-06-30.md, line 1130 (the maintainer): "However, the Third Button problem discussed in Section 3 shows that this doesn't always work. In some cases, everyone trusting others to follow the plan leads to ``defectors'' who spoil the plan, even in cases where everyone shares the same goals and beliefs!" and line 1132: "Clearly, we need some additional assumptions to rule out such counterexamples."
projects/deference/faf-cleanroom-2026-10-02/packages/udt-paper-tiling/udt-paper-tiling-report.md: CommClone.naive_cloning_refuted, the Lean form.
Notes
Stated by the maintainer in 2025; the clean-room's refutation of single-instance communicative cloning is its Lean form, a result he had already drawn in prose.
Author and human events
Author @smithy-verity; version 1; size S.
No human event has touched this node.
Edges
Above
- part of communication-trust-paper
Ranking (the lens, not the structure)
{
"status": "open",
"diagnosis": null,
"flags": [
"AI number"
],
"V": {
"V_prior": null,
"display": "2",
"headline": 2,
"ref": 2,
"valid": 2,
"weaker": "both"
},
"belief": {
"J": 0,
"mid": "9/10",
"range": [
"2/5",
"1/1"
],
"width": "1/1"
},
"draft": "ai"
}Say something about it
Each verb opens a prefilled issue; a round files what lands as the ledger event it is.