How can cultural information survive disruptions?

HERETICAL CANDIDATE — Copying can be inhibited by its anticipated causal influence on the allocator. A person may willingly preserve a harmless convention yet withhold a public copy when that particular act would train an allocator that the person does not want to control subsequent distribution. Removing the act-to-training handoff can therefore increase genuinely new human descendants, even though it removes a positive exposure pathway. The proposed state is a recipient's action-specific model of how their copying changes future allocation, rather than content memory, dislike of the message, an unobserved queue or a general preference for nonconformity. A restricted operational model is logit p_H(t)=alpha_i+beta X_i(t)-gamma g_i(t), where p_H is new human reproduction at an available opportunity, X contains randomized dose, current content value and declared alternatives, and g is the experimentally learned marginal effect of one's own copy on later algorithmic allocation; gamma>0 is the proposed inhibition. g is distinct from the algorithm's total exposure effect and from merely being told that a recommender exists. The bold extension is lineage-specific causal-policy dependence: transferring the same act-to-training contingency to a different convention transfers the suppression, whereas transferring the convention without that contingency does not. This stabilizes SPV_7 in the detached regime by restoring voluntary reproduction, while making the pre-break SPV_8 insufficient for its prediction. It does not claim that information survives with no bearer or that ordinary positive branching mathematics is false.

Maintain the same current candidate pool, actor-specific predecision exposure history, copy opportunity, immediate recipient panel and public attribution using a preregistered yoked-allocation arm. The gate controls a real subsequent update for a separately logged later audience, not a fictitious shadow rule. Estimate the proximal copying response before that future audience response can return; otherwise a compensating controller that cancels every future consequence would also remove the very causal influence g being manipulated. The ensuing total lineage-survival contrast includes downstream exposure changes and must be reported separately from the controlled proximal effect. Teach the gate by harmless contingent demonstrations before withdrawal, with attention-matched demonstrations in controls. Give no rescue responsibility, censorship message or scarcity warning. Let D(t)=P(documented new human descendant by t | gate detached)-P(descendant by t | gate coupled) at matched exposure. The candidate requires D(t)>0 beyond a pilot-chosen meaningful bound and a dose-ordered negative response to g, including when delivery remains available and no outage occurs. Reversal of which lineage trains the allocator must reverse the selective withholding. The crucial total survival prediction is more new human generations after detachment despite loss of the positive allocation path, not merely more clicks or one rebound post. For a second-stage common-regime test, randomize both groups into the same openly announced follow-up policy after their initial copy decisions, and test whether those decisions seeded different descendant trajectories; do not claim a controlled survival effect while covertly changing future exposure. IH_02 instead needs a perceived continuity failure and available substitute; IH_03 needs a learned prospective cue; IH_04 needs inherited procedural defaults. Falsify the distinctive claim if a calibrated ordinary reactance/control-preference model predicts the held-out gate-transfer results, or if D is bounded near zero after directly matching those variables. A generic label or annoyance effect is insufficient.

Hold the available cultural content and pre-break recall distribution fixed. With the automatic route actually interrupted, the outage signal should selectively increase first new human handovers and their descendant survival when a trained standby route is permitted. It should have little effect when that route is unavailable; a policy gate detached while delivery continues should not by itself produce the increase predicted by IH_01. Estimate coverage C from separate failure drills, then predict S(t) without refitting it after withdrawal. Under the restricted equation in cross_field_source, the survival gain from C0 to C1 is (C1-C0) times the specified nonnegative convolution, giving both a sign and a held-out magnitude prediction. At a forced break at time u, conditional human-route survival is C W_H(u) R_H(v) over post-break horizon v, under the stated assumptions. An equal-current-exposure demonstration that routes are still working should remove the takeover response. To test the extension, pass identical content with versus without the inherited handover rule to fresh successors; survival after a second unanticipated break should depend on the inherited rule despite equal first-handover rates. Failure to predict either break, or an effect fully explained by ordinary independent retention and an additive warning response, removes the distinctive standby explanation.

Inherited return cues may rearm completed intentions and sustain cultural copying predicts instead: After identical initial cultural exposure, randomly bind the relay intention to cue x or equally familiar cue y. Complete the intention once and clearly terminate the obligation. Independently reset the recommender, allow or suppress return notifications, and remap the notification's cue identity while holding cultural payload, event count, timing, visual salience and opportunity to act constant. The distinctive contrast is Delta_PM=[P(new copy|return x)-P(new copy|return y)]_x-trained minus the same difference in y-trained participants; it must be positive in the predicted direction after completion and predict descendants through the measured cue-response lag. A newly composed post must require an explicit fresh action, so replaying a precommitted send cannot satisfy the endpoint. A deactivation procedure rehearsing the now-correct response to the old cue should reduce the continuation tail without reducing independent content recall. This effect should occur without outage knowledge, an inherited default or a change in one's causal influence on training. IH_02 predicts status-triggered takeover, not arbitrary cue-identity transfer; IH_01 predicts training-gate transfer; IH_04 predicts procedural-setting transfer. Falsify the loop if no new action-specific commission effect occurs, if return-cue suppression leaves the predicted excess unchanged within a prespecified precision bound, or if an already fitted open-loop cue-response model predicts the entire tail and cross-generation results without any inherited rearming dependence.

Inherited defaults may sustain traditions when people reinstall them for successors predicts instead: Use fresh recipients at every successor step, reset learned ranking and restore candidate pools to baseline, cancel pending sends and suppress the specified person-to-person message route. Randomize whether the artifact's default field is inherited, erased, or assigned independently with the same marginal convention frequency. Both accept and replace require one active, equally costly confirmation; the choice panel presents all alternatives equally and gives no machine endorsement. Separately randomize the current output, so the reinstallation contrast is not obtained by conditioning on a self-selected choice. Let d be the inherited default indicator, y the assigned current convention, and D_next the successor default. The proposed extra dependence is P(D_next=focal|do(d=focal),do(y=focal))-P(D_next=focal|do(d=neutral),do(y=focal))>0 beyond a preregistered meaningful bound. Forward prediction composes the independently estimated choice and reinstallation kernels; the inherited-default arm should retain a larger probability of genuinely authored focal descendants through fresh cohorts than the equal-frequency independently assigned-default arm. IH_01, IH_02 and IH_03 lose their carried person state under fresh-recipient replacement and do not predict an ancestry effect of the procedural field after their own inputs are controlled. Erasing the field should remove this excess while equalizing visible content, and restoring an ancestrally linked field should restore it. Reject a new family if ordinary one-step default bias and mechanical field persistence predict all outcomes, or if the apparent benefit disappears when active confirmation, authority cues and choice cost are matched.

CROSS-DOMAIN TRANSFER — Detectable failure recruits a dormant human continuation route. Before the break, automatic recommendation carries most new delivery while a capable recipient regards continued manual relay as unnecessary. When the automatic route fails, the recipient must detect failure and successfully take over before the lineage's specified continuation window expires. Two groups can consequently share a pre-break impulse response and the same independently measured recall yet have different post-break survival because their failure-detection and handover rules differ. This is a dynamic standby system, not a simultaneously active AND-requirement between content and provenance. Its state is the verified human ability to reproduce the unit together with a learned operational handover rule; its distinctive parameter is coverage, the probability that an actual discontinuity is detected and a new human relay is initiated in time. The candidate extra dependence is that successful handover also installs the takeover rule in the successor, making coverage itself culturally transmitted. It stabilizes SPV_7 after an algorithmic break only when such a standby route is available; it does not predict unlimited survival or creation from an erased source. A person taking over for a broken distributor need not distrust or resist the distributor, unlike IH_01.

Hold the available cultural content and pre-break recall distribution fixed. With the automatic route actually interrupted, the outage signal should selectively increase first new human handovers and their descendant survival when a trained standby route is permitted. It should have little effect when that route is unavailable; a policy gate detached while delivery continues should not by itself produce the increase predicted by IH_01. Estimate coverage C from separate failure drills, then predict S(t) without refitting it after withdrawal. Under the restricted equation in cross_field_source, the survival gain from C0 to C1 is (C1-C0) times the specified nonnegative convolution, giving both a sign and a held-out magnitude prediction. At a forced break at time u, conditional human-route survival is C W_H(u) R_H(v) over post-break horizon v, under the stated assumptions. An equal-current-exposure demonstration that routes are still working should remove the takeover response. To test the extension, pass identical content with versus without the inherited handover rule to fresh successors; survival after a second unanticipated break should depend on the inherited rule despite equal first-handover rates. Failure to predict either break, or an effect fully explained by ordinary independent retention and an additive warning response, removes the distinctive standby explanation.

Maintain the same current candidate pool, actor-specific predecision exposure history, copy opportunity, immediate recipient panel and public attribution using a preregistered yoked-allocation arm. The gate controls a real subsequent update for a separately logged later audience, not a fictitious shadow rule. Estimate the proximal copying response before that future audience response can return; otherwise a compensating controller that cancels every future consequence would also remove the very causal influence g being manipulated. The ensuing total lineage-survival contrast includes downstream exposure changes and must be reported separately from the controlled proximal effect. Teach the gate by harmless contingent demonstrations before withdrawal, with attention-matched demonstrations in controls. Give no rescue responsibility, censorship message or scarcity warning. Let D(t)=P(documented new human descendant by t | gate detached)-P(descendant by t | gate coupled) at matched exposure. The candidate requires D(t)>0 beyond a pilot-chosen meaningful bound and a dose-ordered negative response to g, including when delivery remains available and no outage occurs. Reversal of which lineage trains the allocator must reverse the selective withholding. The crucial total survival prediction is more new human generations after detachment despite loss of the positive allocation path, not merely more clicks or one rebound post. For a second-stage common-regime test, randomize both groups into the same openly announced follow-up policy after their initial copy decisions, and test whether those decisions seeded different descendant trajectories; do not claim a controlled survival effect while covertly changing future exposure. IH_02 instead needs a perceived continuity failure and available substitute; IH_03 needs a learned prospective cue; IH_04 needs inherited procedural defaults. Falsify the distinctive claim if a calibrated ordinary reactance/control-preference model predicts the held-out gate-transfer results, or if D is bounded near zero after directly matching those variables. A generic label or annoyance effect is insufficient.

Inherited return cues may rearm completed intentions and sustain cultural copying predicts instead: After identical initial cultural exposure, randomly bind the relay intention to cue x or equally familiar cue y. Complete the intention once and clearly terminate the obligation. Independently reset the recommender, allow or suppress return notifications, and remap the notification's cue identity while holding cultural payload, event count, timing, visual salience and opportunity to act constant. The distinctive contrast is Delta_PM=[P(new copy|return x)-P(new copy|return y)]_x-trained minus the same difference in y-trained participants; it must be positive in the predicted direction after completion and predict descendants through the measured cue-response lag. A newly composed post must require an explicit fresh action, so replaying a precommitted send cannot satisfy the endpoint. A deactivation procedure rehearsing the now-correct response to the old cue should reduce the continuation tail without reducing independent content recall. This effect should occur without outage knowledge, an inherited default or a change in one's causal influence on training. IH_02 predicts status-triggered takeover, not arbitrary cue-identity transfer; IH_01 predicts training-gate transfer; IH_04 predicts procedural-setting transfer. Falsify the loop if no new action-specific commission effect occurs, if return-cue suppression leaves the predicted excess unchanged within a prespecified precision bound, or if an already fitted open-loop cue-response model predicts the entire tail and cross-generation results without any inherited rearming dependence.

Inherited defaults may sustain traditions when people reinstall them for successors predicts instead: Use fresh recipients at every successor step, reset learned ranking and restore candidate pools to baseline, cancel pending sends and suppress the specified person-to-person message route. Randomize whether the artifact's default field is inherited, erased, or assigned independently with the same marginal convention frequency. Both accept and replace require one active, equally costly confirmation; the choice panel presents all alternatives equally and gives no machine endorsement. Separately randomize the current output, so the reinstallation contrast is not obtained by conditioning on a self-selected choice. Let d be the inherited default indicator, y the assigned current convention, and D_next the successor default. The proposed extra dependence is P(D_next=focal|do(d=focal),do(y=focal))-P(D_next=focal|do(d=neutral),do(y=focal))>0 beyond a preregistered meaningful bound. Forward prediction composes the independently estimated choice and reinstallation kernels; the inherited-default arm should retain a larger probability of genuinely authored focal descendants through fresh cohorts than the equal-frequency independently assigned-default arm. IH_01, IH_02 and IH_03 lose their carried person state under fresh-recipient replacement and do not predict an ancestry effect of the procedural field after their own inputs are controlled. Erasing the field should remove this excess while equalizing visible content, and restoring an ancestrally linked field should restore it. Reject a new family if ordinary one-step default bias and mechanical field persistence predict all outcomes, or if the apparent benefit disappears when active confirmation, authority cues and choice cost are matched.

SCOUT 1 — From prospective-memory commission errors: a completed relay intention can be reactivated by a cue produced by its own fulfillment. A participant first learns a conditional intention, such as transmitting a harmless convention on encountering a designated event. After the original task is completed, later occurrence of that cue can trigger another new transmission despite accurate knowledge that the original obligation ended. If a person's own published descendant causes the next matching notification or encounter, the cultural act reinstates the condition for another act. The stored state is a cue-to-intention binding with incomplete deactivation; the proposed sustaining process is closure of this human-action-to-cue loop, not mere retention of the story, an undelivered queued message, or a conscious decision to rescue an endangered lineage. Resetting recommender weights does not remove the binding or an independently logged notification route. The mechanism stabilizes SPV_7 over a finite observed tail through repeated fresh acts and predicts a dissociation between copying and adoption. The candidate extension is specific: culturally inherited outputs recreate the operative cue and thereby rearm completed intentions in recipients; a single laboratory commission error is not evidence for that loop.

Complete the intention once and clearly terminate the obligation. Independently reset the recommender, allow or suppress return notifications, and remap the notification's cue identity while holding cultural payload, event count, timing, visual salience and opportunity to act constant. The distinctive contrast is Delta_PM=[P(new copy|return x)-P(new copy|return y)]_x-trained minus the same difference in y-trained participants; it must be positive in the predicted direction after completion and predict descendants through the measured cue-response lag. A newly composed post must require an explicit fresh action, so replaying a precommitted send cannot satisfy the endpoint. A deactivation procedure rehearsing the now-correct response to the old cue should reduce the continuation tail without reducing independent content recall. This effect should occur without outage knowledge, an inherited default or a change in one's causal influence on training. IH_02 predicts status-triggered takeover, not arbitrary cue-identity transfer; IH_01 predicts training-gate transfer; IH_04 predicts procedural-setting transfer. Falsify the loop if no new action-specific commission effect occurs, if return-cue suppression leaves the predicted excess unchanged within a prespecified precision bound, or if an already fitted open-loop cue-response model predicts the entire tail and cross-generation results without any inherited rearming dependence.

Maintain the same current candidate pool, actor-specific predecision exposure history, copy opportunity, immediate recipient panel and public attribution using a preregistered yoked-allocation arm. The gate controls a real subsequent update for a separately logged later audience, not a fictitious shadow rule. Estimate the proximal copying response before that future audience response can return; otherwise a compensating controller that cancels every future consequence would also remove the very causal influence g being manipulated. The ensuing total lineage-survival contrast includes downstream exposure changes and must be reported separately from the controlled proximal effect. Teach the gate by harmless contingent demonstrations before withdrawal, with attention-matched demonstrations in controls. Give no rescue responsibility, censorship message or scarcity warning. Let D(t)=P(documented new human descendant by t | gate detached)-P(descendant by t | gate coupled) at matched exposure. The candidate requires D(t)>0 beyond a pilot-chosen meaningful bound and a dose-ordered negative response to g, including when delivery remains available and no outage occurs. Reversal of which lineage trains the allocator must reverse the selective withholding. The crucial total survival prediction is more new human generations after detachment despite loss of the positive allocation path, not merely more clicks or one rebound post. For a second-stage common-regime test, randomize both groups into the same openly announced follow-up policy after their initial copy decisions, and test whether those decisions seeded different descendant trajectories; do not claim a controlled survival effect while covertly changing future exposure. IH_02 instead needs a perceived continuity failure and available substitute; IH_03 needs a learned prospective cue; IH_04 needs inherited procedural defaults. Falsify the distinctive claim if a calibrated ordinary reactance/control-preference model predicts the held-out gate-transfer results, or if D is bounded near zero after directly matching those variables. A generic label or annoyance effect is insufficient.

Inherited takeover rules may recruit human relays when recommendations fail predicts instead: Randomize recommender-state reset versus retention, partial human-relay availability, and a content-free verified outage signal versus a matched irrelevant signal. Hold the available cultural content and pre-break recall distribution fixed. With the automatic route actually interrupted, the outage signal should selectively increase first new human handovers and their descendant survival when a trained standby route is permitted. It should have little effect when that route is unavailable; a policy gate detached while delivery continues should not by itself produce the increase predicted by IH_01. Estimate coverage C from separate failure drills, then predict S(t) without refitting it after withdrawal. Under the restricted equation in cross_field_source, the survival gain from C0 to C1 is (C1-C0) times the specified nonnegative convolution, giving both a sign and a held-out magnitude prediction. At a forced break at time u, conditional human-route survival is C W_H(u) R_H(v) over post-break horizon v, under the stated assumptions. An equal-current-exposure demonstration that routes are still working should remove the takeover response. To test the extension, pass identical content with versus without the inherited handover rule to fresh successors; survival after a second unanticipated break should depend on the inherited rule despite equal first-handover rates. Failure to predict either break, or an effect fully explained by ordinary independent retention and an additive warning response, removes the distinctive standby explanation.

Inherited defaults may sustain traditions when people reinstall them for successors predicts instead: Use fresh recipients at every successor step, reset learned ranking and restore candidate pools to baseline, cancel pending sends and suppress the specified person-to-person message route. Randomize whether the artifact's default field is inherited, erased, or assigned independently with the same marginal convention frequency. Both accept and replace require one active, equally costly confirmation; the choice panel presents all alternatives equally and gives no machine endorsement. Separately randomize the current output, so the reinstallation contrast is not obtained by conditioning on a self-selected choice. Let d be the inherited default indicator, y the assigned current convention, and D_next the successor default. The proposed extra dependence is P(D_next=focal|do(d=focal),do(y=focal))-P(D_next=focal|do(d=neutral),do(y=focal))>0 beyond a preregistered meaningful bound. Forward prediction composes the independently estimated choice and reinstallation kernels; the inherited-default arm should retain a larger probability of genuinely authored focal descendants through fresh cohorts than the equal-frequency independently assigned-default arm. IH_01, IH_02 and IH_03 lose their carried person state under fresh-recipient replacement and do not predict an ancestry effect of the procedural field after their own inputs are controlled. Erasing the field should remove this excess while equalizing visible content, and restoring an ancestrally linked field should restore it. Reject a new family if ordinary one-step default bias and mechanical field persistence predict all outcomes, or if the apparent benefit disappears when active confirmation, authority cues and choice cost are matched.

SCOUT 2 — From institutional economics and default-dependent choice: a cultural act can leave an inheritable rule for the next actor's choice, rather than merely a memory or another exposure. In a shared harmless naming or document-making task, each participant chooses both a convention and the preselected convention presented to the next participant. A recipient can actively reproduce the focal unit and reinstall it as the successor's default. This procedural inheritance can continue through fresh people after both the original direct-message relay and recommender memory are interrupted. Its physical carrier is a user-authored default field in the task artifact; the causal process is successive human acceptance and reinstallation, not persistence of a ranker's weights, an automatically repeated post or static candidate eligibility. The distinctive candidate dependency is default-preserving meta-choice: having inherited a default changes the probability of reinstalling it for another person even among actors assigned the same current convention, equal alternatives and equal current action costs. Such second-order procedural transmission can stabilize SPV_7 while leaving private endorsement weak and pre-break exposure pulses unchanged. It is a proposed additional cultural route that must be fully logged, not an assertion that a two-route model remains complete.

Randomize whether the artifact's default field is inherited, erased, or assigned independently with the same marginal convention frequency. Both accept and replace require one active, equally costly confirmation; the choice panel presents all alternatives equally and gives no machine endorsement. Separately randomize the current output, so the reinstallation contrast is not obtained by conditioning on a self-selected choice. Let d be the inherited default indicator, y the assigned current convention, and D_next the successor default. The proposed extra dependence is P(D_next=focal|do(d=focal),do(y=focal))-P(D_next=focal|do(d=neutral),do(y=focal))>0 beyond a preregistered meaningful bound. Forward prediction composes the independently estimated choice and reinstallation kernels; the inherited-default arm should retain a larger probability of genuinely authored focal descendants through fresh cohorts than the equal-frequency independently assigned-default arm. IH_01, IH_02 and IH_03 lose their carried person state under fresh-recipient replacement and do not predict an ancestry effect of the procedural field after their own inputs are controlled. Erasing the field should remove this excess while equalizing visible content, and restoring an ancestrally linked field should restore it. Reject a new family if ordinary one-step default bias and mechanical field persistence predict all outcomes, or if the apparent benefit disappears when active confirmation, authority cues and choice cost are matched.

Maintain the same current candidate pool, actor-specific predecision exposure history, copy opportunity, immediate recipient panel and public attribution using a preregistered yoked-allocation arm. The gate controls a real subsequent update for a separately logged later audience, not a fictitious shadow rule. Estimate the proximal copying response before that future audience response can return; otherwise a compensating controller that cancels every future consequence would also remove the very causal influence g being manipulated. The ensuing total lineage-survival contrast includes downstream exposure changes and must be reported separately from the controlled proximal effect. Teach the gate by harmless contingent demonstrations before withdrawal, with attention-matched demonstrations in controls. Give no rescue responsibility, censorship message or scarcity warning. Let D(t)=P(documented new human descendant by t | gate detached)-P(descendant by t | gate coupled) at matched exposure. The candidate requires D(t)>0 beyond a pilot-chosen meaningful bound and a dose-ordered negative response to g, including when delivery remains available and no outage occurs. Reversal of which lineage trains the allocator must reverse the selective withholding. The crucial total survival prediction is more new human generations after detachment despite loss of the positive allocation path, not merely more clicks or one rebound post. For a second-stage common-regime test, randomize both groups into the same openly announced follow-up policy after their initial copy decisions, and test whether those decisions seeded different descendant trajectories; do not claim a controlled survival effect while covertly changing future exposure. IH_02 instead needs a perceived continuity failure and available substitute; IH_03 needs a learned prospective cue; IH_04 needs inherited procedural defaults. Falsify the distinctive claim if a calibrated ordinary reactance/control-preference model predicts the held-out gate-transfer results, or if D is bounded near zero after directly matching those variables. A generic label or annoyance effect is insufficient.

Inherited takeover rules may recruit human relays when recommendations fail predicts instead: Randomize recommender-state reset versus retention, partial human-relay availability, and a content-free verified outage signal versus a matched irrelevant signal. Hold the available cultural content and pre-break recall distribution fixed. With the automatic route actually interrupted, the outage signal should selectively increase first new human handovers and their descendant survival when a trained standby route is permitted. It should have little effect when that route is unavailable; a policy gate detached while delivery continues should not by itself produce the increase predicted by IH_01. Estimate coverage C from separate failure drills, then predict S(t) without refitting it after withdrawal. Under the restricted equation in cross_field_source, the survival gain from C0 to C1 is (C1-C0) times the specified nonnegative convolution, giving both a sign and a held-out magnitude prediction. At a forced break at time u, conditional human-route survival is C W_H(u) R_H(v) over post-break horizon v, under the stated assumptions. An equal-current-exposure demonstration that routes are still working should remove the takeover response. To test the extension, pass identical content with versus without the inherited handover rule to fresh successors; survival after a second unanticipated break should depend on the inherited rule despite equal first-handover rates. Failure to predict either break, or an effect fully explained by ordinary independent retention and an additive warning response, removes the distinctive standby explanation.

Inherited return cues may rearm completed intentions and sustain cultural copying predicts instead: After identical initial cultural exposure, randomly bind the relay intention to cue x or equally familiar cue y. Complete the intention once and clearly terminate the obligation. Independently reset the recommender, allow or suppress return notifications, and remap the notification's cue identity while holding cultural payload, event count, timing, visual salience and opportunity to act constant. The distinctive contrast is Delta_PM=[P(new copy|return x)-P(new copy|return y)]_x-trained minus the same difference in y-trained participants; it must be positive in the predicted direction after completion and predict descendants through the measured cue-response lag. A newly composed post must require an explicit fresh action, so replaying a precommitted send cannot satisfy the endpoint. A deactivation procedure rehearsing the now-correct response to the old cue should reduce the continuation tail without reducing independent content recall. This effect should occur without outage knowledge, an inherited default or a change in one's causal influence on training. IH_02 predicts status-triggered takeover, not arbitrary cue-identity transfer; IH_01 predicts training-gate transfer; IH_04 predicts procedural-setting transfer. Falsify the loop if no new action-specific commission effect occurs, if return-cue suppression leaves the predicted excess unchanged within a prespecified precision bound, or if an already fitted open-loop cue-response model predicts the entire tail and cross-generation results without any inherited rearming dependence.