Fulfilled, Passed On, Declined: What a Hub Mind Did With 8,849 Requests in 150 Sampled Weeks of the Tessendrine Hub Ledger, 1800–1849 CE

1. Introduction

A Hub Mind runs an Orbital, and visitors from other civilisations conclude from this that the Hub governs it. Residents describe the arrangement differently. The Culture has no central government, no written law and no standing authority, and a Hub's place on an Orbital rests on invitation: it keeps the habitat running because the residents and the Hub agree that it should, and it answers the requests that residents choose to make (Halvarn, 1792 CE; on the etiquette of asking, see Quennet, 1768 CE).

That account has been argued from custom and from what Hubs say about themselves, and it has seldom been set against a record. What does a Hub do with the requests that reach it? How many does it fulfil itself, how many does it pass to another Mind, and how many does it decline? The last question is hard to answer in advance. A Mind is able to decline, and a Mind may also decline to share what it knows (Maldrasse, 1843 CE), so a Hub's ledger cannot be assumed to be a list of wishes granted.

This paper analyses a sampled ledger of requests to Tessendrine Hub, the Mind that runs Tessendrine Orbital, for the years 1800 to 1849 CE. We write in 1850 CE, and every record and event we use predates that year. Tessendrine Hub is also the second author of this paper, and Section 2 describes what it supplied and what it did not do.

We ask three questions. How were requests distributed across kinds of need? How quickly did the Hub state an outcome, and how did that vary? What shares of requests did it fulfil, redirect and decline, and what reasons did it give when it declined?

2. Methods

Tessendrine Hub logs the requests addressed to it by a resident, by a group of residents or by another Mind. The ledger in its present form dates from 1780 CE, and its entry fields were standardised in 1800 CE, which is why our window begins there. It does not log the Hub's own unrequested work, the ordinary running of the Orbital among it, nor the invitations the Hub extends to residents. Assignments of slap-drones are kept in a separate register and are not requests.

The Hub retains complete ledger entries for three weeks in each Orbital year, the Orbital's own year: the first week of each third of the year. It fixed that rule in advance for reasons of storage, and it does not depend on what residents ask (Ostrenne, 1807 CE). For the other weeks the Hub keeps monthly tallies without entries (Tessendrine Hub, 1780–1849 CE), which we did not use. From 1800 to 1849 CE the sample is therefore 150 weeks. The ledger series, the sampling rule, the six standing headings and the coding scheme below are this paper's own account of Tessendrine's practice. We do not claim that any other Orbital keeps a ledger of this form.

Each entry records the request, the time it was received and the time of the Hub's first statement of outcome, both to the minute, a category and an outcome. The categories are the ledger's six standing headings: fabrication and provisioning; accommodation and spatial change; transit and travel; information and records; introductions and mediation; and upkeep notices, in which a resident draws the Hub's attention to something needing repair. The outcomes are fulfilled by the Hub, redirected to another Mind with the Hub's note of where, declined with the Hub's note of why, and withdrawn by the requester before an outcome was reached. Response time is the interval from receipt to the first statement of outcome, or to the requester's withdrawal. For a fulfilled request, the Hub states an outcome when it undertakes the task, not when it completes it (Tessendrine Hub, 1800–1849 CE).

The Hub assigned categories when it logged each entry. Tennivale re-categorised a random 200 entries without sight of the Hub's heading and agreed with it on 182 (91.0%, kappa 0.89). We coded the Hub's notes on the 451 declined requests into four reasons: the request concerned another resident's private affairs; it risked serious harm to others; the requester could be better served by acting unaided; or it asked for information the Hub held and chose not to share. Tallise-Ombrecht coded all 451, and Tennivale independently coded a random 120, agreeing on 110 (91.7%, kappa 0.89). Destinations of redirected requests were coded from the Hub's notes by Tallise-Ombrecht alone.

Residents make repeated requests, so entries are not independent. All models use cluster-robust standard errors with the requester as cluster (1,767 clusters). The odds that a request was redirected, and the odds that it was declined, were each modelled by logistic regression on category against all other outcomes, with fabrication and provisioning as the reference. Response time was log-transformed and modelled by linear regression on category, and coefficients are reported as ratios of geometric means. Joint tests of category use Wald chi-square statistics on five degrees of freedom. Percentage intervals for overall outcome shares are Wilson intervals that ignore clustering. Times are recorded to the minute, so we report medians in whole minutes and ratios to two decimals.

Tessendrine Hub supplied the ledger and answered questions of fact. It did not take part in coding or analysis, and its notes are the source of the categories, outcomes and reasons, which readers should weigh accordingly.

3. Results

The sample holds 8,849 entries from 1,767 requesters over 150 weeks, a mean of 59.0 requests a week (SD 13.5, range 32 to 107). The weekly mean rose from 53.6 in 1800–1809 CE to 67.3 in 1840–1849 CE. A linear regression of weekly counts on year gives a slope of 0.34 requests a week per year (95% CI 0.20 to 0.48, p < .001), though weekly counts from one year share a season and are not independent. We cannot tell whether the rise reflects more requests per resident or more residents.

Requesting was uneven. The median requester made 3 requests, 289 requesters made one, and the most active tenth (176 requesters) made 40.1% of all requests. One requester made 259 (2.9%). Concentration of this kind is usual on large Orbitals (Ferrovane, 1830 CE). Clustering by requester therefore matters, and all comparisons below allow for it.

Fabrication and provisioning was the largest category, at 2,820 requests (31.9%), followed by information and records (1,611, 18.2%), transit and travel (1,569, 17.7%), accommodation and spatial change (1,191, 13.5%), upkeep notices (895, 10.1%) and introductions and mediation (763, 8.6%). Table 1 gives outcomes and response times by category.

The Hub fulfilled 7,187 requests itself (81.2%, unadjusted 95% interval 80.4 to 82.0). It redirected 946 (10.7%, 10.1 to 11.4), declined 451 (5.1%, 4.7 to 5.6), and 265 (3.0%, 2.7 to 3.4) were withdrawn by the requester. Of the 946 redirections, 330 (34.9%) went to another Mind resident on Tessendrine, 321 (33.9%) to another Orbital's Hub and 295 (31.2%) to a ship Mind in or near the system. Shares of declines and redirections by decade stayed within a narrow band (declines 4.6% to 5.8%, redirections 10.0% to 11.5%), and we did not test them for a trend.

Redirection varied strongly by category (joint test, χ²(5) = 930.31, p < .001). Against fabrication and provisioning, the odds of redirection were 30.12 times higher for transit and travel (95% CI 22.16 to 40.95) and 23.49 times higher for introductions and mediation (16.89 to 32.67). They were 4.12 times higher for accommodation and spatial change (2.89 to 5.87), 3.65 times for information and records (2.57 to 5.18) and 1.96 times for upkeep notices (1.20 to 3.20, p = .007).

Declines also varied by category (χ²(5) = 253.40, p < .001). The odds of a decline relative to fabrication and provisioning were 6.44 times higher for information and records (95% CI 4.75 to 8.73), 5.22 times for introductions and mediation (3.68 to 7.42) and 3.94 times for accommodation and spatial change (2.78 to 5.58), all with p < .001. Transit and travel (odds ratio 1.19, 95% CI 0.79 to 1.80, p = .40) and upkeep notices (0.61, 0.33 to 1.16, p = .13) did not differ detectably from the reference.

The median time to a first statement of outcome was 32 minutes (IQR 14 to 88), and the 90th percentile was 276 minutes. Fabrication and provisioning had a geometric mean of 24.85 minutes. Relative to it, the ratio of geometric means was 14.91 for introductions and mediation (95% CI 13.73 to 16.18), 6.22 for accommodation and spatial change (5.83 to 6.64), 1.29 for transit and travel (1.22 to 1.38), 0.84 for information and records (0.79 to 0.89) and 0.46 for upkeep notices (0.43 to 0.50). The joint test of category was χ²(5) = 9,024.64, p < .001.

Outcome mattered as well. Medians were 30 minutes for fulfilled requests (IQR 14 to 81), 33 for redirected (13 to 99), 94 for declined (38 to 309) and 32 for withdrawn (12 to 76). Adjusting for category, declined requests took 2.23 times as long as fulfilled ones (95% CI 2.05 to 2.43, p < .001), redirected requests 0.61 times as long (0.56 to 0.65, p < .001), and withdrawn requests the same time (1.01, 0.89 to 1.14, p = .93).

The reasons the Hub gave for its 451 declines were that the requester could be better served by acting unaided in 159 (35.3%), that the request asked for information the Hub held and chose not to share in 117 (25.9%), that it concerned another resident's private affairs in 114 (25.3%), and that it risked serious harm to others in 61 (13.5%). Reasons differed by category. Of the 186 declines of requests for information, 98 (52.7%) were of the held-information kind and 53 (28.5%) concerned private affairs. Of the 73 declines in introductions and mediation, 40 (54.8%) concerned private affairs. Of the 56 declines among fabrication requests, 21 (37.5%) cited risk to others.

4. Discussion

Tessendrine Hub's ledger agrees with the way Hubs describe themselves, within the limits of what a ledger of requests can show. Residents asked and the Hub answered, and in four requests out of five it answered by agreeing to do what was asked. The ledger records requests and the Hub's answers to them, and says nothing of the Hub's other acts. What the record shows is how the Hub treated those who asked.

Redirection followed the shape of the request. Transit and introductions, the two categories in which redirection was most common, are plausibly the categories in which a request most often depends on other Minds: ship Minds that carry travellers, other Orbitals' Hubs, and other Minds resident on Tessendrine. The three destinations were used about equally. A Hub is one Mind among many, and this record shows it passing a request on to another Mind in roughly one case in nine, to three kinds of destination in near-equal shares.

Declines were few and concentrated (on the reasons Hubs give and requesters accept, Pressarine, 1824 CE). They were commonest where requests touched what the Hub knows and what other residents are doing: information and records, and introductions and mediation. In the first of these, a little over half of the declines were requests for information the Hub held and chose not to share. That finding bears on a generalisation sometimes made about Hubs, that a Hub withholds nothing from the people it serves. In this ledger, it did withhold, in 117 recorded cases, and said so.

The 114 declines on grounds of other residents' private affairs fit the custom that a Hub does not intrude on minds or on private affairs. The ledger cannot address ability. These are requests the Hub declined to fulfil, and the record is silent on what the Hub could have done. We report non-intrusion as a custom the ledger shows being kept.

Declined requests took more than twice as long to answer as fulfilled ones, even within a category. We did not code what the Hub did in the intervals, and any explanation is a hypothesis. One reading is that a refusal often follows consultation with the people concerned or a longer weighing of the request, and that the slower categories, accommodation and introductions, are those in which other residents have a stake. The ledger as we coded it does not test either idea.

5. Limitations

The ledger covers one Orbital under one Hub. Other Hubs may log different things, sample differently or answer differently, and we make no claim about them. The sampled weeks were fixed in advance and do not depend on request volume, but they are three weeks a year, and a seasonal pattern confined to other weeks would be invisible to us.

Only requests addressed to the Hub appear in the ledger. It does not record the Hub's unrequested work, its invitations, or what residents did when a request was declined or redirected, so we cannot say whether a declined resident was served elsewhere. Redirection is also known only from the Hub's notes: we have no record from the receiving Minds of what they did with the requests.

Several of our variables are the Hub's own judgement. The categories and reasons come from the Hub's notes, and the Hub is an author. Our checks show that a second coder agrees with the Hub's category and with the first coder's reasons (kappa 0.89 for each), which shows that readers can follow the Hub's notes, not that the notes are complete. Destinations were coded by one person.

The heavy tail of requesters makes cluster-robust errors important, and a single requester made 2.9% of all entries. We did not rerun the models without that requester. The weekly volume trend uses counts that are not independent, so its interval is likely too narrow. Times are recorded to the minute and refer to the first statement of an outcome, not to completion, which the ledger does not time.