The close of Chapter 21 asked whether a student who got things wrong this time can be taken back by the institution as a person still learning. We now move the same question onto the last kind of relationship this book treats — the kind that has no classroom, no site, not even shared presence. The scenes of the preceding chapters all depended on bodily presence: a site had to withstand staying, care had to have substitutes, a classroom had to permit redoing. Once digital relationships remove presence, these problems do not disappear; they merely lose the answers that used to hold automatically: who takes up a commitment made from the other side of the screen, how a sentence spoken wrongly can be taken back, whether a relationship that cannot survive a broken promise still counts as a relationship. This chapter takes up the point of convergence that this book and the preceding volume have repeatedly deferred until now — how digital relationships can be reconnected to responsibility in reality, reconnected into relationships that are bearable, correctable, and consequential, rather than first ruling on whether online relationships are genuine or fake.
The street message group set up in Chapter 1 survived into the story: during the rainstorm crisis it carried building-to-building mutual aid, and two years later the city upgraded it into the "Chengjiang Mutual Aid" mini-program — post a request, accept a task, accumulate points, view ratings. This chapter uses three years of that mini-program's fictional records as its material; none of the persons, data, or events correspond to any real product, platform, or research finding. One boundary must first be fixed: Chapter 1 already showed that digital tools can indeed generate a new nearby and new mutual aid — the message group let new residents find the notices and let absent people learn about the courtyard, and these functions really exist; Chapter 11 also argued that parasocial relationships are not all false relationships. What needs tracing here is where responsibility breaks in transmission, and under what conditions it can be reconnected.
Digital Connection and Actual Commitment
Distant members sharing materials or accompanying an appeal cannot be called a virtual substitute merely because they are not on site; the genuineness of a relationship is judged by sustained response, commitment, and consequence, not by the medium. This chapter pushes the criterion one step further: not whether an online relationship is genuine, but whether it can hold what is placed in it — whether commitments have a bearer, whether errors have a place of correction, whether a relationship still has a response that does not rupture after a broken promise. Connection counts, activity levels, and mutual-follow status answer none of this.
A bearable relationship needs its commitments to pass through three links. The first is reception: the message reaches someone able to handle it. The second is uptake: someone claims it, and others can verify that they have. The third is settlement: once done, it can be checked; when not done, there is a response. The observation from Chapter 1 applies here without qualification — the tool widens visibility but does not by itself distribute authority and responsibility. The digital interface almost strengthens the first link while often weakening the second and the third: the higher the visibility, the more easily "everyone knows" replaces "someone is responsible."
The substitution a digital interface most easily manufactures is felt presence passing itself off as consequential presence. Felt presence is knowing, being moved, being angry, liking; consequential presence is the work getting done because of it and the error getting corrected because of it. Offline, the two are roughly bound together by the body — the person present cannot dodge the cleanup. Online they come apart by nature, and the separation favors the operator: feeling can be metered, pushed, and monetized, while consequence only produces cost. To judge a digital relationship, look first at one thing: which of the two presences it runs as its product.
A common skipped step also needs blocking: inferring from "this cannot hold commitments online" that "the online is inherently unable to hold them." The medium neither creates nor destroys responsibility; it changes responsibility's form — written commitment rests on signature and witness, spoken commitment on presence and reputation, and digital commitment could rest on record and confirmation. That a platform cannot hold commitments shows that its design supplies no such form, not that the form is impossible. Landing the conclusion on the medium lets the concrete structure off; only by landing the conclusion on the structure can reconnection be discussed at all.
Three Years of a Mutual-Aid Mini-Program
The mini-program's first year genuinely lightened the load. A request for help could carry location and time, acceptance carried confirmation, and completion carried mutual rating; the person coming off a night shift, the mother who could not leave home with a toddler, the elder with limited mobility — for the first time they did not have to happen to catch someone downstairs. Chapter 7 showed that migrants' support networks are often far away in another town, and Chapter 17 showed that crisis mutual aid has to be rewired along buildings and trade networks; the platform's contribution was to automate part of the wiring work such reconnection requires. Those gains are real, and this chapter does not withdraw them.
The change began in the second year. The platform launched "neighborhood credit scores": accepting a task adds points, good ratings add points, declining an assignment subtracts points, and a complaint against others found false subtracts points. Chapter 12 analyzed how data classification turns consuming identity into predictable type; the same mechanism moved into the neighborhood: a retired teacher who had helped thirty times and a young person newly arrived who for now could only ask for help are two different ranks in the system. The score was first a record, then a ranking, finally a ticket — a high score unlocked priority in asking for help, while those without a score had to wait one extra review step even to post.
A single record from the third winter deserves a separate look. Zhou, an older person living alone, posted a request for someone to buy medicine on her behalf; the page showed thirty-eight views, no one accepted, and four hours later a social worker handled it offline. The illusion of visibility repeats itself here in full: everyone saw it, so no one needed to confirm it. The street message group of Chapter 1 manufactured the same appearance of no one in charge with "it was already posted in the group, don't ask again"; the platform merely makes it more precise — every view is counted, while uptake alone has no column.
The Boundaries of Feeling, Stance, and Intimacy
That autumn, a screenshot circulated through the group: someone described being followed along a stretch of the riverbank. Within two hours it had been forwarded a thousand times, the captions sliding from "be careful, everyone" to "that stretch has been a mess for ages," and no one checked the time, the section, or whether a report had been filed. Chapters 1 and 2 of the preceding volume argued that public fact needs a verifiable process to stabilize; the digital mechanism here is that forwarding is far cheaper than checking, so rumor always arrives ahead of verification, and arrival itself is then taken as its evidence.
Feeling is not a crime. A frightened person has the right to tell, and telling has real value — the line between acknowledging the limits of a perspective and canceling the standard of fact was drawn in the preceding volume. The problem lies in the mechanism by which feeling acquires the standing of evidence: the quantity of emotion (how many forwards, how much sympathy) substitutes for the quality of evidence (what, when, by what means of verification), and "so many people are saying it, it cannot all be false" passes off consensus of ideas at the social scale as observational consensus. The evidentiary threshold is therefore clear: telling one's own experience creates no burden of proof; upgrading one experience into a general judgment about a zone and a population requires naming one's grounds.
The drift of word meaning accelerates along the forwarding chain. "Someone was followed along a stretch of the riverbank one evening" becomes "that stretch is not safe," and then "the people over there are no good." Chapter 5 of this volume analyzed how everyday language carries context through forms of address, tone, and shared memory; online strips the context away layer by layer and leaves only the transmissible conclusion. Chapter 16 of this volume already analyzed the mechanism by which a term's extension inflates; what is added here is the digital-side accelerant: every forward compresses the context once more, and at the end of the compression the word retains only an attitude and no longer points to anything verifiable.
Repair does not come from demanding that everyone verify before speaking — that would spread the duties of journalists and the police across every resident and would only silence the weak. What works is keeping a base text one can return to: the original post and its time are not deleted, corrections are as visible as the original, and residents of the named zone have a place to respond. Chapter 2 of the preceding volume showed that the stability of public fact lies in returnability and contestability; a platform that governs rumor by deletion deletes the base text along with the rumor, and what remains is only the next, faster rumor.
Stances Stay Online; Who Is Left with the Work
On the crisis's second anniversary, the whole city launched the "Guard Chengjiang" forwarding relay, and within three days the page was covered in candles. That same week, the rosters for visiting elders living alone and for building-by-building checks remained short of hands — Chapter 17 had already recorded this: those who keep filling the gaps are the same people. The relay is not a bad thing; the problem is its cost structure: the merit of the stance is fully displayed at zero cost, while the actual labor is borne by others and equally unshown. When the two ledgers are kept separately, voice keeps flowing to the visible one.
A reverse simplification must be blocked first: taking a stance from afar is not thereby false. Chapter 17 showed that remittances and trade-network revolving during a crisis were real and effective, that distant sharing of materials and accompanying appeals are no virtual substitute, and Chapter 7 had already recorded how cross-place support became part of migrants' lives. The problem is not the stance itself but the diversion: voice gathers where it is visible, free, and rewarded, while the work lands where it is invisible, costly, and unpaid. Chapter 20's term applies here — the substitute must be present; a forward cannot substitute for a person.
The condition of reconnection is therefore not moral exhortation but a check of channels: can a forward become a sign-up with one click, can declared support land on the roster, does the moved person have an unembarrassing path to the person being helped. Where the channels do not exist, the more fervent the stance, the closer it comes to a discharge of responsibility — "I already forwarded it" becomes a psychological settlement while the work itself still has no taker. To judge an online mobilization, do not look at the volume of voice; look at how many exits it leaves for voice to become action.
The rumor's ledger also has a side the forwarders overlook. Once a stretch of the riverbank was named again and again, the households there, the people who set up stalls under the bridge, and the workers who walk that stretch home at night began to bear the consequences of the judgment: some walked past their stalls, and landlords lowered their voices at the mention of "that stretch." The stancers pay zero cost and the named pay the whole bill — homologous with the community expectations analyzed in Chapter 3 and the expansion of records analyzed in Chapter 21: one event rolls over into the identity of a place, and the people living in that place are still those least able to move away.
Parasocial Intimacy and the Boundary of Appropriation
In the platform's records there is a young man living alone who watches the local streamer "Sister Jiang" every night, three years without missing a broadcast; he says it is "the only place I am seen." Chapter 11 argued that parasocial relationships are not all false relationships: they provide real company, rhythm, and consolation, and this chapter does not withdraw that conclusion. Only one sentence needs adding: the intimacy of this relationship is supplied unilaterally by watching — Sister Jiang does not know his name — and this is not a concealment exposed but the established structure of the relationship.
The structure itself is not an accusation; what poses the problem is the management. When the platform turns "the feeling of being seen" into metered interaction metrics — intimacy score, companionship hours, combo streaks — unilateral intimacy passes from a quality of the relationship into a function of the product. Chapter 11's criterion of appropriation applies here directly: look at the capacity to end, the conditions of substitution, and the actual consequences. If the young man can skip a night's stream without panic, the relationship still lies within his available margin; if not watching breeds anxiety and watching deepens loneliness, appropriation and illusion have been completed at once.
The same boundary is drawn around high-investment communities. Chapter 11 analyzed how tasks, ranks, and data returns make exit expensive, and Chapter 12 showed how community labor and identity become bound to a single point of entry. The intimacy is real, and so is the appropriation; the two do not cancel each other. On top of this, this chapter adds only one question: can this relationship hold one cry for help — would he dare mention in the comments that he is moving, and if he did, would anyone take it up. Intimacy that cannot hold this need not be judged false, but neither should it be taken as evidence that mutual aid already exists.
Labor, Exclusion, and the Transfer of Responsibility
The mini-program's daily operation appears on no page: reviewing requests, mediating disputes, teaching elders to register, handling warnings about scams impersonating customer service in the middle of the night. The three-year records show this work concentrated in four group admins, two of whom were precisely the people first called on during the crisis of Chapter 17. Chapter 18's analysis of unpaid requisition applies here word for word: the lower a person's bargaining power, the higher the cost of refusal, and maintenance labor slides toward whoever that is; praise — here called "warm-heartedness" — substitutes for pay. Requisition needs no malice; the platform sincerely thanks every volunteer, and the ledger stands all the same.
The rating system is the second destination of community labor. Mutual aid's original bookkeeping was vague and two-way: you helped me this time; I will remember next time. The credit score changes it to one-way and precise: each appearance is priced, and each absence is recorded. Chapter 12 already analyzed how data classification locks in consuming identity; moved into the neighborhood, the consequence cuts one layer deeper — people who do not use the platform have no score, and in a system where everyone has one, having none is itself suspicion. The freedom to exit exists in name, and the price of exiting is being reclassified.
Once the score exists, behavior migrates toward it. The changes in the records are concrete: accept a task and first photograph it as evidence; choose the visible errands to help with; late-night and time-consuming requests find fewer and fewer takers — because the score cannot see patience, only counts. Chapter 10 described how two quota slots first changed the gestures of peer aid; the neighborhood rating manufactures the same pressure: exactly the kind of help that photographs nothing, leaves no name, and would do it again next time is the hardest to register, and yet it is part of what allows the nearby to continue.
The third question is ownership. Three years of maintenance hours were paid by the group admins, while the data assets generated, the matched traffic, and the commission revenue belong to the platform. Chapter 18's conditions of reconnection transfer over directly: maintenance labor that is publicly needed should be publicly acknowledged — pay, rotation, and the safety to refuse; data should be verifiable, portable, and appealable. This clause does not target digital relationships in particular; it treats Chapter 18's oral-history project and the group admins here alike: acknowledgment cannot substitute for compensation.
Exclusion in the Name of Reconnecting to Reality
Repair itself can also seize the right to define the relationship. After the Zhou incident, talk in the city turned: "It is all fake online; mutual aid must return to real neighbors." The street administration then took the mini-program down, replacing it with fixed-point registration, weekly meetings, and face-to-face assignment. This turn appears in the guise of repair, but its mechanism reads "reconnect to reality" as "cancel the online": whoever holds the single definition of genuineness — must be face to face, must be present, must be offline — acquires the right to judge all relationships.
The beneficiary structure must be listed as it is: the administration recovers the convenience of supervision; the core of the offline acquaintance network recovers its central position; people uneasy with new technology gain a sense of order. Those who bear the cost are the people who depended on the online entry: night-shift workers, people with limited mobility, people once hurt by offline networks who dare not ask again — Chapter 1 long ago wrote that familiarity too can close a door, and the weak who have been watched by acquaintances know best what lies behind it. The evidentiary threshold follows: to judge whether a "return offline" is reconnection or exclusion, look at whether those who relied on that channel to speak can still make their voice heard once the channel is removed.
The counterexamples must be written in full: some online contact really does crowd out necessary face-to-face time, and some groups really have nothing left but forwards; reconnecting a specific relationship to a specific scene is real repair. The criterion of judgment is the same one Chapter 18 used in the dispute between updating and preserving: look at whether the change widens or narrows the range of options of those concerned. What adds a channel is reconnection; what swaps in a new judge is exclusion — and the move that lets the formerly voiceless speak in one more way and the move that lets the formerly vocal monopolize speech again can employ exactly the same discourse.
The Platform's Transfer of Responsibility
The second severe case sits at the platform's other end. A resident booked an in-home repair through the platform's high-score recommendation; the repairman's mistake caused damage; the platform invoked its terms in response: the recommendation information is posted by users, the scores are given by users, and the platform only provides an information service and bears no responsibility. The terms are impeccable on their face, yet the accounting keeps two ledgers apart: the commission on matching, the power of ranking, and the appreciation of data belong to the platform, while the posted content and the consequences of choice belong to the user. Organized revenue and individualized responsibility — this is the mechanical structure of transfer.
The structure is not new. Chapter 8 analyzed communities forced to provide unpaid backup after a platform exits, and Chapter 9 traced how responsibility for care is shifted layer by layer onto the weakest party; here is the same logic realized in a digital interface: responsibility flows toward the point of lowest bargaining power, and revenue concentrates toward the point of highest organization. The beneficiaries are the platform and the merchants who acquire traffic at low cost; the bearer is the individual user — who in a dispute faces not another user but an agreement he has never read clause by clause.
The counterexample must likewise be written in full: platform liability cannot be made unlimited. If the platform had to bear full responsibility in advance for every sentence users post, the only rational response would be blanket review and excessive deletion, and what gets hurt would still be ordinary expression — Chapter 7 of the preceding volume, in its analysis of transparency and response, reminded us that the demand for responsibility must be proportionate to the capacity to respond. The condition of reconnection is therefore correspondence, not generalization: insofar as the platform takes pay from matching, it answers for the verifiability of the matching — identity verification, a channel of appeal, the order of compensation — and does not underwrite judgments of the content's value. Wherever benefit reaches, responsibility follows; that much only, and no less.
The terms themselves do not decide the outcome; the channels do. The same sentence — "the platform bears no responsibility" — paired with an externally usable appeal, advance handling, and proof obligations on merchant verification, or paired with an internal process that is at once judge and party, amounts to two entirely different institutions. Chapter 19, analyzing the platformization of space, noted that a single point of entry makes the platform's failure equal the closing of the site; the same logic holds here: if appeal has only one channel and one judge, what the terms say no longer matters — what matters is that the user has no alternative. To judge whether a transfer of responsibility has occurred, look at the shape of the channels before reading the letter of the agreement.
How Connection Takes Up the Work Again
The first mechanism of reconnection realizes Chapter 1's conclusion at the scale of the platform: distinguish idle talk from delegation with consequences, and require a handover confirmation for the latter. "Viewed" and "accepted" on a request page must be two different numbers; once a commitment is claimed, the claim must be checkable and its withdrawal must leave a trace. This is not bureaucratizing neighborhood chat — chat remains free and dispersed as before — it is giving those willing to commit a place where a commitment can stand. The opposite of bearable has never been recklessness; it is a commitment issued with no one knowing on whom it lands.
The second mechanism borrows Chapter 21's criterion: the record of an error and the record of its correction must enter the archive on equal terms and be visible on equal terms. A platform whose error handling consists only of deletion and bans is an archival regime that keeps only the record of failure and erases the trajectory of correction — the asymmetry analyzed in Chapter 21 reappears here word for word. A rumor can be named, the named can respond, and a correction can catch up with the spread of the original post; whatever "governance" cannot do these things functions only to drive the error into the next group.
The third mechanism is the plainest: a relationship can survive one broken promise. The taker can say "I cannot come today" on time, the requester can switch to someone else without resentment being recorded, and a departing user can take the records without taking the stigma. Chapter 20 required that a caregiver be able to say "not this time" and care continue as usual; the corresponding formulation for digital relationships is: a user can decline one task without credit being docked and leave one platform without the relationship being severed. The cost of exit and the stigma after exit need checking in digital relationships just as much: a relationship that cannot survive a broken promise drains people, online or off.
Making Digital Relationships Auditable
Concrete checking can proceed along the direction of responsibility. For any mutual-aid group or platform: does a request for help carry confirmation of uptake; does the rumor have a base text one can return to and a visible correction; is there a channel through which a stance becomes taking over; on whom are maintenance hours booked, and with what compensation; can the scores be checked, appealed, and sealed; do the platform's benefits and responsibilities correspond; do those who do not use it have another entry. The seven questions give no total score; they only make the direction of responsibility discussable — the same method as Chapter 18's labor audit, Chapter 19's questions about the site, and Chapter 21's questions about evaluation.
One direct corollary of the audit is the transfer of Chapter 19's registration minimalism: every record should be able to name the specific function it serves — preventing duplication, enabling verification, retracing disputes; a record that cannot name its function is only raw material for profiling. Chapter 21's three questions of coverage, validity period, and granularity apply unchanged: does today's request for help feed tomorrow's credit; are old disputes sealed with time; is "one unaccepted task" recorded, or "refused all year." What digital relationships face is neighborhood trust that has not yet taken shape, and the obligation of restraint that Chapter 21's growth records owed to a person not yet formed holds equally here.
The boundary must also be stated: the audit does not judge who should use platforms and who should return offline, and it does not preset any one way of living as the solution. It only guarantees that the channels between the two ways of living are not dismantled in the dark — those who use them know the cost, and those who do not use them keep an entry. The checklist itself takes no position; the position lies with whoever uses the checklist, who should answer for the consequences, in line with the close of Chapter 18.
Conditions of Reconnection and the Boundary of Reconnection
Before all conditions of reconnection, one boundary must be set: offline relationships are not the natural bearers of responsibility. Chapter 1 described how the nearby turned from support into control, and Chapter 17's crisis mutual aid also pressed the filling of gaps onto the most callable people. Returning offline does not solve the problem of commitments with no bearer; it merely moves the same problem onto a scene with no audit record. Reconnecting to reality refers not to a place but to a structure: bearable, correctable, consequential — these three, in an offline acquaintance society, equally have to be earned one by one.
Chapter 19's criterion for spatial digital interfaces applies to the mutual-aid platform without changing a word: can a person who does not carry a phone complete one ordinary request for help and one ordinary act of helping. Multiple entries, substitutability, appeals passing through different people, records corresponding to functions — these conditions do not negate digitalization; they only require that the interface remain an entry rather than become a wall. The reverse holds by the same token: the real existence of offline windows is the premise that the online entry need not monopolize; both directions use the same test.
Commitments in digital relationships still need someone to bear them. Chapter 1's handover, Chapter 8's interface, Chapter 11's boundary, Chapter 12's data, Chapter 17's filling of gaps, Chapter 19's entry — all point to the same place: a digital relationship need not be judged false, and it has no right to exemption from consequence. Whether a digital relationship has reconnected to responsibility in reality comes down, in the end, to the question moved over at the start of this chapter — whether the most ordinary person, the one who broke a promise this time, can remedy it without panic while the relationship continues as before. Chapter 19 asks whether a person can stay; Chapter 21 asks whether a person can try again; this chapter asks whether a commitment can land on a person; the three questions point to the same capacity: a community taking commitments back to the people who bear them, rather than leaving them where everyone sees them and no one is responsible.