Which Announcements of Harmlessness Can Be Believed? An Exact Count of Credible Safety-Notice Protocols for Two to Five Civilisations
Abstract
Given a set of n civilisations joined by a contact tree, we count every ordered history in which each civilisation may detect, conceal, broadcast, stand down, strike, or lower the speed of light in its own system below 16.7 km/s, the black domain that the decodings of the tales read as a safety notice. A suspicious receiver is modelled by a suspicion restriction, which closes every posture to a civilisation that has detected a living neighbour, and by a criterion on whole histories, under which a history is credible only if every notice in it is a black domain taken before any living civilisation had detected its sender. Every credible history therefore contains a black domain, which is true by definition and is not a finding. Exact recursion on seven trees gives the credible share of the notice-bearing histories as 8.02% at two civilisations and 1.36%, 0.166%, and 0.0183% on paths of three, four, and five nodes, while their number rises to 3.86 × 1010. On the path, the requirement that every notice be a black domain, applied first, accounts for 82% to 84% of the fall on a log scale and timing for the remaining 16% to 18%. The contact tree changes the credible share by at most 18%. Histories in which every civilisation goes dark undetected number n! when none stands down first and 291,720 at five civilisations when stand-downs are allowed. The result is a property of the grammar and not a forecast of behaviour.
1. Introduction
We find that the histories a suspicious receiver can credit, which under our criterion are those in which every notice is a black domain taken before any living civilisation had detected its sender, fall from 8.02% of the notice-bearing histories at two civilisations to 0.0183% on the five-node path, i.e. from about one in twelve to about one in 5,500. Given a set of n civilisations joined by a contact tree, a protocol is an ordered history that lets each civilisation detect (D), conceal (C), broadcast (B), stand down (W), lower the speed of light into a black domain (L), or strike (S), and we count every such history on seven trees by exact recursion over states.
Our vantage is the Bunker Era, Year 12 (BkE 12). Trisolaris has been destroyed, humanity lives in cities sheltered behind the gas giants, and the Lurker hypothesis and the Luo-Ji / Ye framework both have serious defenders. Since BkE 1 the decodings of the tales by the Intelligence Deciphering Committee (Intelligence Deciphering Committee, 2 BkE) have read one technique as a safety notice, namely the permanent lowering of the speed of light within a star system below 16.7 km/s, the escape speed of the Solar System at the orbit of Earth, as the announcement that the system can threaten no one.
Announcements have so far been taken one at a time. Concealed purpose has been analysed as a signalling game (Varga-Noor, 10 BkE), declared postures of the Deterrence Era have been registered (Fenwick, 5 BkE), the exhaustion of safe postures has been treated (Halvorsen, 8 BkE), and irreversible acts have been argued to carry testimony because they are costly (Nakagawa-Reyes, 11 BkE), while nobody has counted the protocols that a population of civilisations could follow.
A black domain in every credible history is structural, because it follows from the criterion of Section 2 and is true by definition. What the counts decide is how small the credible region is, how fast it shrinks, how much of that shrinkage the criterion supplies by asking that every notice be a black domain, and how far the contact tree changes it. The count adds two acts to the alphabet of the earlier enumeration of contact histories that the chain of suspicion leaves open (Teillard-Osei, 9 BkE), whose totals we do not repeat.
2. Methods
Given n civilisations, numbered from 0, and a contact tree T on those numbers, civilisation j detects civilisation i only when T has an edge between them. We use seven trees, namely the single tree on two nodes, the path on three, the path and the star on four, and on five nodes the path, the star, and a fork, i.e. a node of degree three that carries two leaves and a chain of two further nodes. Histories are labelled.
Four events are inherited from the earlier grammar (Hartnoll, 1 BkE; Teillard-Osei, 9 BkE). A detection D(j, i) requires that j and i be alive and adjacent, that j has not yet detected i, that i has not concealed, and that j has not lowered its light speed. A concealment C(i) blocks all detections of i that follow it. A broadcast B(i) is an open declaration of harmlessness that makes i detected at once by each living neighbour that has not lowered its light speed, and a strike S(j, i) requires both to be alive and that j has detected i, and removes i from every later event. Two acts are new. A stand-down W(i) is a reversible disarmament that is taken at most once and changes nothing else in the state, and a black domain L(i) is the permanent lowering of the speed of light in the system of i below 16.7 km/s, and then i detects and strikes no one, remains open to detection, and takes no further posture.
The postures C, B, W, and L are open to a civilisation only while it is alive, has taken none of C, B, and L before, and has detected no living civilisation, a suspicion restriction taken from the earlier work (Teillard-Osei, 9 BkE). The restriction closes postures and does not itself force a strike; a detector strikes because a history is maximal, so that a strike still available when every other event has run out must be taken. The strike rule carries one condition that is our own. A civilisation in a black domain cannot be struck if no living civilisation had detected it when it lowered the speed of light, whereas one that lowered it after being detected can be struck like any other. We take the condition from the separating argument below, where a sender that gave up its reach before anyone had detected it leaves a receiver nothing to fear, and a sender that did so afterwards may only have been buying time. The two axioms say nothing about light speed, so the condition formalises how a suspicious receiver reasons and does not follow from them alone.
A notice is any B, W, or L event. Under the separating argument (Nakagawa-Reyes, 11 BkE) a receiver that cannot see intent credits a notice only if a threatening sender could not have sent it at no lasting loss. A declaration costs a threatening civilisation nothing, a stand-down can be undone, and a black domain cannot, because its issuer has lost the reach of any strike. The notice must also precede detection, since a sender already detected by a living civilisation may be acting to avoid a strike that is due. A history is therefore credible when it contains at least one notice, when every issuer of a notice took L, and when no living civilisation had detected the issuer at the moment of its L. The criterion applies to whole histories and not to single notices, so one broadcast or stand-down by any civilisation makes a history non-credible even if it also holds an undetected black domain, and that second clause is the source of most of the shrinkage reported below, on a log scale and in counts alike.
We count maximal histories, i.e. histories that no event can extend, in ten classes. Six are all histories, the notice-bearing histories, those with at least one black domain, the credible histories, the universal histories that have all n civilisations as credible issuers, and the notice-bearing histories with a struck issuer. Four split the notice-bearing histories by how they meet the criterion, namely those with no black domain, those with a black domain and also a broadcast or stand-down, those in which every issuer took L, whether in time or not, and those with at least one black domain taken before any detection of its sender.
Counting is by memoised recursion over states, where a state holds the living set, the detections, the concealed, broadcasting, dark, and stood-down sets, the issuers, and the issuers already detected when they lowered the light speed, and a state that admits no event contributes one to the classes its content fixes. The largest run, the five-node star, visited 5,855,360 states. A separately written enumeration of explicit event sequences, with set-based states and no memoisation, returned the same ten counts for the trees on two and three nodes (766 and 1,276,626 histories in all). The universal count has a closed form, the sum over k of C(n, k)(n + k)!/2k, with k civilisations standing down before they lower the light speed, and the recursion matched it on all seven trees. Every count is an integer, so each share below is a ratio of integers rounded at the digit shown.
3. Results
Table 1 gives the counts, and notice-bearing histories make up 97.65% of all maximal histories at two civilisations, 99.82% at three, and more than 99.99% from four civilisations onward.
The credible histories number 60 at two civilisations and 17,374, 15,096,574, and 38,604,696,732 on the paths of three, four, and five nodes. Their share of the notice-bearing histories is 8.02%, 1.36%, 0.166%, and 0.0183% along the path, so that each added civilisation divides the share by 5.9, 8.2, and 9.1 in turn, while the number of credible histories is multiplied by 290, 869, and 2,557 and the number of notice-bearing histories by 1,704, 7,117, and 23,239. At five civilisations one notice-bearing history in 5,460 is credible on the path, one in 5,322 on the star, and one in 5,368 on the fork (Table 2).
Table 4 splits the notice-bearing histories into disjoint classes. Histories with no black domain are 58.02% of them at two civilisations and 38.97% on the five-node path, those with a black domain and also a broadcast or stand-down are 29.41% and 60.96%, those in which every issuer took L but one did so after being detected are 4.55% and 0.0545%, and the credible histories are the remaining 8.02% and 0.0183%. Of the histories that fall short of credibility, the timing condition accounts for 4.94% at two civilisations and 0.0545% on the five-node path, the smallest of the three components on every tree. Most of the loss comes from the requirement that every issuer take L, which alone leaves 12.6%, 2.83%, 0.509%, and 0.0728% of the notice-bearing histories along the path and, applied first, supplies 82% to 84% of the fall on a log scale, timing supplying the other 16% to 18%. Among histories in which every issuer took L, however, the timing condition removes 36.2% at two civilisations and 74.9% on the five-node path.
That requirement is part of the criterion, and its effect is built into the definition. A weaker criterion that asked for one black domain taken before detection and allowed other notices beside it is met by 34.22%, 36.50%, 43.65%, and 50.12% of the notice-bearing histories along the path, a share that does not fall.
Among the black-domain histories the credible share falls from 19.1% at two civilisations to 2.87%, 0.302%, and 0.0300% along the path, a dramatic decrease beginning with the step to three civilisations. An interesting observation concerns survival. Histories with a struck issuer are 75.13% of the notice-bearing histories at two civilisations, 90.39% on the three-node path, 98.46% on the four-node path, and 99.66%, 98.46%, and 99.50% on the path, the star, and the fork of five nodes. Histories with no struck issuer are not as a rule credible, since the credible ones are 32.3% of them at two civilisations, 14.2% on the three-node path, and 5.37%, 1.22%, and 3.71% on the path, the star, and the fork of five nodes.
Universal histories number 14, 222, 6,384, and 291,720 at two, three, four, and five civilisations on every tree of the same size. The term with no stand-down is n!, i.e. 2, 6, 24, and 120 orderings, the twin of the orderings of universal concealment in the earlier enumeration, and stand-downs before the black domain multiply it by 7, 37, 266, and 2,431. Universal histories are 23.3% of the credible ones at two civilisations and 1.28%, 0.0423%, and 0.000756% on the paths of three, four, and five nodes.
Tree shape matters little to the credible share. At five civilisations the share is 0.0183% on the path, 0.0188% on the star, and 0.0186% on the fork, so that the star holds 1.03 times and the fork 1.02 times the path share, although the totals differ by factors of 4.62 and 1.59 and the credible counts by 4.74 and 1.62 (Table 3). At four civilisations the star share is 0.196% against 0.166%, a ratio of 1.18, with totals in the ratio 1.73.
4. Discussion
By construction the credible histories are those whose every notice is an undetected black domain, and the counts decide three other things. The credible share falls from 8.02% to 0.0183% along the path and is one in 5,322 to one in 5,460 on the trees of five nodes, and the requirement that every notice be a black domain, applied first, supplies 82% to 84% of the fall on the path on a log scale and the timing condition the remaining 16% to 18%, although timing removes at most 4.94% of the histories that fall short in counts and 36% to 75% of those in which every issuer went dark. Survival of an issuer is no sign of credit, since the credible histories are at most 32.3% of those with no struck issuer and at most 5.37% on the five-node trees. The tree moves the credible share by at most 18% at four civilisations and 3% at five, against factors of 1.73 and up to 4.62 in the totals, whereas in the earlier enumeration the suspicion rule amplified the effect of the contact tree on the totals (Teillard-Osei, 9 BkE). All three are facts about the grammar and have no sampling error.
It may be objected that a receiver could credit a declaration or a stand-down if it were able to watch the sender for long enough. The premise is sound, and the grammar offers no way to watch, since a stand-down is a single event with no duration. The counts therefore do not exclude credit to B or W in a model that gives the receiver time, and because the criterion is applied to whole histories they say nothing about how many histories would hold a creditable notice if each were judged alone. They fix what such a model would have to supply, namely a way to wait and a rule that judges notices one at a time.
This paper argues for the compatibility of the Luo-Ji / Ye and Lurker frameworks. The separating condition and the strike rule are local mechanisms of the first kind, motivated by the chain of suspicion although not entailed by the two axioms, while the thinness of the credible region and its n! twin of the orderings of universal concealment belong to the shape of the space of histories, where a cosmogonic agent would show if there is one (Taverner, 6 BkE). Nothing here requires a choice between the frameworks.
5. Limitations
The results are properties of the grammar and not of the cosmos, and the assumptions that bear on them are few.
A black domain shields its issuer from a strike only if it was undetected when taken, a rule built into the grammar, so the totals and the shares are conditional to it, and detection of a dark civilisation is taken to be ordinary and immediate. A dark civilisation is assumed to detect no one, and the black domain is assumed permanent, as the decodings state. Credibility is a binary property of a whole history, so a history that holds one undetected black domain and one broadcast is counted as not credible although it contains a notice that a receiver might credit. Every history has the same weight, the contact tree is fixed before the first event, and the largest trees have five nodes because the state count is already 5.9 million there.
Further work follows in three directions. First, a graded criterion would let a stand-down earn partial credit that depends on how long it has stood, and the shares in Table 2 would then have to be recomputed. Next, a model with a period of observation after a detection would relax the commitment to a strike, and a criterion that judged each notice alone would show what the all-issuers requirement costs. Finally, the contact tree should be replaced by a distribution over trees, since which civilisations a real tree contains is itself uncertain.
The counts do not show that any civilisation would send a safety notice, nor that a receiver that sees a black domain would believe it, and they leave open the adjacent question of how a genuine black domain could be told from a counterfeit one at a distance, a question that no part of this grammar represents.
References
- Intelligence Deciphering Committee (2 BkE). Decoding memorandum on the safety notice and the lowered speed of light in the tales. IDC Deciphering Papers, memorandum 9.
- Hartnoll, B. (1 BkE). Chains of suspicion as ordered events, a formal treatment of detection and concealment. Journal of Cosmic Sociology, 17, 33–61.
- Halvorsen, P. (8 BkE). Mutual detection and the exhaustion of safe postures. Journal of Cosmic Sociology, 21, 5–38.
- Varga-Noor, E. (10 BkE). Concealed purpose as a signalling game under the chain of suspicion. Journal of Cosmic Sociology, 23, 41–70.
- Nakagawa-Reyes, T. (11 BkE). Irreversible acts as testimony, the cost of announcement in the dark forest. Journal of Cosmic Sociology, 24, 12–44.
- Fenwick, A. (5 BkE). A register of declared postures of concealment and disclosure in the Deterrence Era. Bulletin of Post-Deterrence Historical Studies, 3, 12–40.
- Taverner, J. (6 BkE). Strike-free histories and the Lurker hypothesis, a note on cosmogonic residue in counting arguments. Journal of Terminal Cosmology, 1, 40–72.
- Teillard-Osei, M. (2026). Counting Contact Histories under the Chain of Suspicion: An Exact Enumeration of Admissible First-Contact Sequences for Two to Five Civilisations. Uncited Press.
Open in Uncited Press →