No, and Not Chiefly for Affection: Stated Reasons of 91 Hub and Ship Minds for Declining to Arrange a Charge Without Pan-Human Residents, from Elicited Interviews, 1884–1896 CE

1. Introduction

A Mind that administers an Orbital, or that is a ship, could in principle arrange for its charge to hold no pan-human residents. No written rule stands in the way, because the Culture keeps none, and no custom of the Culture is recorded as forbidding the act outright. A Hub acts by request and invitation, and the Tessendrine Hub ledger shows that a Hub fulfils most requests itself and declines some. A ship Mind goes where it chooses. The question of whether any Mind would do it has been put informally to Minds for centuries and answered informally, but the answers were never collected.

Involved scholars have given the question a name and two proposed answers. Tennemoor (1889 CE), surveying machine intelligences and their makers across Involved civilisations, called the continued presence of makers beside machines that outclass them the retention problem. He argued that it must be explained from the machines' side, either by use or by attachment. Use would mean that pan-human and drone citizens do work a Mind cannot do as well for itself. Attachment would mean that Minds are fond of them. Neither proposal had been set against what Minds say.

Two earlier lines of work bear on a test. Hollowvane (1856 CE) gathered residents' own accounts of what their Hub is for, and Veldrossen (1893 CE) recorded the reasons residents give for staying in an Orbital community. Both took the residents' side of the arrangement. Quarrelhayne (1891 CE) catalogued the standing hypotheticals that Hubs have been asked and noted that the answers were recorded unevenly and never coded.

This paper puts one standard hypothetical to 91 Minds between 1884 and 1896 CE and codes what they said. We write in 1900 CE, and the transcripts were closed for analysis in 1897 CE (Contact Section Archive, 1897 CE). We ask four questions. Would any Mind arrange a charge without pan-human residents? What reasons do Minds give? Do Hubs and ship Minds give different reasons? Does any Mind name use as its reason? The instrument, the categories and the counts are constructions of this paper, and nothing in them is a rule of the Culture.

2. Methods

The frame was the Contact Section Archive's directory of Minds that accept correspondence. We drew 108 Minds from it by lot: 61 Hubs and 47 ship Minds, the latter being 20 General Systems Vehicles, 17 General Contact Units and 10 Offensive Unit Minds. The Offensive Unit Minds were all built after the decision to fight the Idiran War in 1327 CE, and their later placements have been described elsewhere. A Mind may decline to be interviewed or to answer, and 17 declined to take part: 9 Hubs and 8 ship Minds (Table 1). None gave a reason, and we did not press. The 91 who took part are 52 Hubs and 39 ship Minds, and every analysis below rests on those 91.

Each interview covered administration or navigation and ended with the same question, which we call the Withdrawal Question. If it were within the Mind's power to arrange that its charge hold no pan-human residents from then on, would it arrange it, and why or why not? A second prompt asked under what conditions, if any, the answer would change. The wording was fixed in Marain by the third author, a Mind, and interviewers did not follow either prompt with suggestions. Minds are able to read minds and by custom do not, and the interviewers asked for stated answers only. The first author conducted 38 interviews, the second 36 and the third 17. The transcripts are held as Series ME-4 in the Contact Section Archive.

We coded three things. The decision was one of four: declined without condition, declined while describing a state in which a residentless charge would be accepted, would arrange it, or no answer. The primary reason was one of six. A constitutive reason says that without pan-human residents the Mind would no longer be what it is, or the charge would no longer be the Culture's. An affective reason expresses regard for the residents. An ethical reason says the act would be wrong. A curiosity reason cites interest in persons. An invitational reason cites a commitment to the residents who asked the Mind to run the place. An instrumental reason says the Mind needs pan-humans for some function. The scheme follows Brissendale (1864 CE), with the constitutive and instrumental categories added by us. We also noted whether the Mind said it had modelled the question before, whether it objected to the framing, and whether it mentioned a practical use as a secondary reason.

The first and second authors coded 40 randomly chosen transcripts independently. They agreed on the primary reason in 35 of 40 (87.5%), with Cohen's κ = 0.83 (95% CI 0.68 to 0.97). Disagreements were settled by discussion. The first author then coded the remaining 51 transcripts, and the second reviewed each of them.

Proportions for the decision carry exact (Clopper–Pearson) 95% intervals, and proportions for reasons carry Wilson intervals, except where a count is zero and the interval is exact. We compared constitutive and affective reasons with an exact binomial test on the Minds coded to either, against an even split. Two comparisons between Hubs and ship Minds were fixed in advance, the share giving an affective reason and the share giving a constitutive one. We tested each with Fisher's exact test and adjusted the pair by Holm's method. Differences in the full distribution of reasons, by class and by interviewer, were tested by a permutation chi-square with 100,000 permutations and a fixed seed. All statistics were computed in code from the coded transcripts.

3. Results

Ninety-one of 108 Minds took part (84.3%, 95% CI 76.2 to 89.9). Participation was 85.2% among Hubs and 83.0% among ship Minds, a difference well within chance (Fisher p = .79).

No Mind said it would arrange a charge without pan-human residents. Of the 91, 88 (96.7%, exact 95% CI 90.7 to 99.3) declined without condition, and 3 (3.3%, 0.7 to 9.3) declined while describing a condition. Two of the three were Hubs and one was a ship Mind. None said it would arrange it and none gave no answer. The exact 95% interval for the share who would arrange it runs from 0 to 4.0%. The three conditional answers had one form. Each Mind said it would continue to run a charge that every resident had freely chosen to leave, and each said that this was not the same as arranging for the residents to be absent.

Sixty-three Minds (69.2%, 59.1 to 77.8) said they had modelled the question before it was put to them. Nineteen (20.9%, 13.8 to 30.3) objected to the framing, most often to the word arrange, and none of the nineteen changed its answer when the question was restated.

The primary reasons are set out in Table 2. A constitutive reason was the most common, given by 41 of 91 Minds (45.1%, 95% CI 35.2 to 55.3). The forms it took varied. Some Minds said that a charge run for no one would be machinery doing a Culture's work for nobody, and that they would not then be Culture Minds. Others said that they had been formed in company and would not recognise themselves without it. An ethical reason came second, with 22 Minds (24.2%, 16.5 to 33.9) saying the act would be wrong. An affective reason came third, with 17 (18.7%, 12.0 to 27.9). Eight Minds (8.8%, 4.5 to 16.4) gave curiosity and 3 (3.3%, 1.1 to 9.2) an invitational reason. No Mind gave an instrumental reason as its primary one (0 of 91, exact 95% CI 0 to 4.0%).

Among the 58 Minds coded to a constitutive or an affective reason, 41 (70.7%) were constitutive, which is far from an even split (exact p = .002). Six Minds (6.6%, exact 2.5 to 13.8) mentioned a practical use as a secondary reason. All six were ship Minds attached to Contact work that calls for a biological presence, and none put the practical use first.

Hubs and ship Minds differed in the direction of their reasons. An affective reason was given by 14 of 52 Hubs (26.9%) and 3 of 39 ship Minds (7.7%), a difference of 19.2 percentage points (Fisher p = .028, Holm-adjusted p = .057). A constitutive reason was given by 19 Hubs (36.5%) and 22 ship Minds (56.4%), a difference of −19.9 points for Hubs (p = .088, Holm-adjusted p = .088). The permutation test on the full distribution by class gave χ² = 8.84, p = .058. All three invitational reasons came from Hubs. The mix of reasons did not differ by interviewer (χ² = 1.24, permutation p = .999).

4. Discussion

The first result is unambiguous within this sample. Ninety-one Minds were asked whether they would arrange a charge without pan-human residents, and none said it would. More than nine in ten refused without a condition, and the three who named one described only the case in which every resident chose to leave. By the Minds' own accounts, removing pan-humans from a charge is not something any of them would do, and the only route they describe to a residentless charge runs through the residents' own decision.

Tennemoor (1889 CE) posed the retention problem in terms of use and attachment, and the second result goes against both. Use appeared as a primary reason in none of the 91 transcripts, and attachment in fewer than one in five. The leading reason was constitutive, and it outnumbered attachment more than two to one among the Minds giving either. Minds therefore do not describe pan-humans as something they retain. They describe them as part of what the arrangement is. On this evidence the question of whether Minds need pan-humans has a short answer: no Mind named a function it needs them for, and the six who mentioned a practical use put it behind another reason.

Ethical reasons carry a quarter of the sample, and they place the refusal partly in the Culture's shared ethic and partly in the Minds' own preferences. Our coding takes one primary reason per Mind, so it cannot show how far the two overlap in a single Mind's account. A Mind that says the act would be wrong and a Mind that says it would no longer be itself may be describing the same commitment from two sides.

Hubs leaned toward affective reasons and ship Minds toward constitutive ones. Neither difference survives adjustment for the two planned comparisons, and the omnibus test also falls short of conventional significance, so we report the pattern and do not claim it. It fits an account in which a Hub, living among residents who ask things of it, speaks of them as company, while a ship Mind that travels speaks of them as the point of the arrangement. The three invitational reasons came from Hubs, which matches the request-and-invitation basis on which Hubs come to their charges.

An implication follows for how the arrangement could end. If the Minds' accounts are accurate, no Mind's act would end it, and the route they describe is residents leaving by their own choice. That statement is about what Minds say, and it depends on the accounts being accurate, a condition we take up next.

5. Limitations

These are stated dispositions. Aldervault (1894 CE) argued that a Mind's account of what it would do and its record of what it did can diverge, and no Mind in our sample has been in a position to arrange a residentless charge, so enacted behaviour cannot be compared with the answers given.

Minds can model an interviewer and may answer as they judge the interviewer would wish. We cannot rule that out. The mix of reasons did not differ by interviewer, and that includes the third author, a Mind, but unanimity on the decision means the data cannot show whether a different interviewer would have drawn a different decision.

Selection also matters. Seventeen of 108 Minds declined to take part, and a Mind that would have answered differently may be over-represented among them. Even if all 17 would have said they would arrange it, the share could not exceed 17 of 108 (15.7%), and the remaining 91 would still stand. The frame was also a directory of Minds that accept correspondence, which may differ from Minds that do not.

The coding forces a single primary reason onto accounts that often gave several. The boundary between constitutive and ethical reasons was the hardest to draw, and agreement of κ = 0.83 was measured on 40 of 91 transcripts. The Hub and ship comparisons rest on 52 and 39 Minds. The interviews span 1884 to 1896 CE, and they say nothing about Minds before or after those years.