{"q": "For each recovered high-autonomy fault episode, construct and cost the next-best lower-autonomy substitute: name a specific flown lower-autonomy analogue and cost its ground-loop recovery behavior (operator-hours, DSN/tracking-pass time, mission-downtime days) from TechPort plus design documentation, rather than inferring it from episodes that merely score lower on the ordinal index.", "facet": "identification", "raised_by": "fogel", "priority": "high", "query_terms": ["spacecraft fault recovery operator hours cost", "Deep Space Network tracking pass anomaly recovery duration", "ground-in-the-loop fault recovery cost lower-autonomy spacecraft", "social saving counterfactual costed substitute"], "status": "open"}
{"q": "Compute the upper and lower bounds on the autonomy hazard ratio the assembled NTRS/GAO/JPL population can support given its realized terminal-event count, and the event count at which the estimable bound is too wide to discriminate H1 from H0.", "facet": "empirics", "raised_by": "fogel", "priority": "high", "query_terms": ["Cox model minimum events per variable rare outcome", "events-per-variable survival analysis power", "Firth penalized Cox rare events bound", "minimum detectable hazard ratio event count"], "status": "open"}
{"q": "Measure the shadow price of recovery time and run a mediation/time-channel test showing the autonomy effect runs through reaction-time-saved (faster onboard isolation beating light-time-plus-ground-cycle) rather than through unobserved program quality, using fault-entry-to-resolution timestamps.", "facet": "mechanism", "raised_by": "fogel", "priority": "high", "query_terms": ["mediation analysis survival time-to-resolve channel", "light-time ground command cycle recovery latency deep space", "shadow price of time social savings methodology", "program quality confounder autonomy survival mediation"], "status": "open"}
{"q": "Construct the missing cost leg of the social-saving claim: do NTRS/GAO/JPL records carry episode-level recovery-cost fields (operator-hours, DSN time, dollars, downtime), or is the hazard-ratio-as-social-saving equivalence a rhetorical analogy unbacked by any cost covariate?", "facet": "economics", "raised_by": "fogel", "priority": "high", "query_terms": ["spacecraft anomaly recovery cost operator hours dataset", "mission operations cost fault recovery deep space", "DSN cost per tracking pass anomaly", "social saving cost counterfactual cliometrics"], "status": "open"}
{"q": "Partition the autonomy benefit into a survival channel (fewer losses) and a time-to-resolve channel (faster return to nominal among survivors), and report the distribution of fault-entry-to-recovery durations by autonomy level, since recovery is currently modeled only as censoring and the time-saved margin is discarded.", "facet": "empirics", "raised_by": "fogel", "priority": "high", "query_terms": ["time to recovery distribution spacecraft fault autonomy level", "competing risks recovery as event survival analysis", "Leunig time is money passenger social savings", "accelerated failure time recovery duration"], "status": "open"}
{"q": "Stratify losses by off-vehicle externality (localized single-vehicle deep-space loss vs reallocation-bearing loss of a relay/constellation node/shared asset or a debris-generating event) and test whether the corpus supports a cause-specific competing-risks treatment, since a uniform loss-hazard mis-aggregates the general-equilibrium reallocation channel.", "facet": "rival", "raised_by": "fogel", "priority": "normal", "query_terms": ["competing risks cause-specific hazard heterogeneous terminal events", "spacecraft loss externality relay constellation downstream mission", "debris-generating fragmentation event systemic cost", "general equilibrium market access social savings critique"], "status": "open"}
{"q": "From a feasibility computation on the actual assembled episode counts, report the minimum number of tail events the hardest-episode stratum will hold and demonstrate (dfbeta leave-one-out) that the tail-benefit conclusion cannot be flipped by adding, removing, or recoding a single tail event.", "facet": "empirics", "raised_by": "taleb", "priority": "high", "query_terms": ["dfbeta leave-one-out influence Cox single event reversal", "thin tail subgroup survival analysis instability", "heavy-tailed rare events undersampling exposure bound", "leverage diagnostics proportional hazards few events"], "status": "open"}
{"q": "Partition coded episodes by terminal-event externality (localizable single-vehicle loss vs systemic debris-generating/congested-shell loss) and establish whether NTRS anomaly narratives and GAO assessments record enough detail to support the partition, or whether the dependent variable is structurally blind to the systemic tail the precautionary frame invokes.", "facet": "measurement", "raised_by": "taleb", "priority": "high", "query_terms": ["orbital debris fragmentation loss congested shell coding", "spacecraft terminal event externality classification", "non-naive precautionary principle systemic ruin fat tails", "NTRS GAO anomaly narrative debris detail"], "status": "open"}
{"q": "Show whether the missions whose documentation produced the autonomy scores were authored by parties exposed to the downside of loss or by the same engineering organizations whose autonomy investment is being evaluated, and whether that authorship asymmetry makes the favorable upper-bound estimate the cheap assurance the skin-in-the-game filter discards.", "facet": "governance", "raised_by": "taleb", "priority": "high", "query_terms": ["self-report bias program documentation treatment scoring", "independent oversight GAO versus implementing program framing", "skin in the game assurance authorship exposure", "TechPort TRL self-classification autonomy maturity"], "status": "partial"}
{"q": "Replace the monotone hazard ratio with a second-order response: estimate via an autonomy-by-stressor-dose interaction or spline whether the marginal protective effect of one autonomy level rises (convex/antifragile) or falls (concave/fair-weather) with dose, and show the event count can resolve the curvature, not just the average.", "facet": "mechanism", "raised_by": "taleb", "priority": "high", "query_terms": ["antifragility convex response dose stressor", "autonomy by severity interaction spline hazard", "second order convexity fragility precautionary", "power to detect interaction curvature survival rare events"], "status": "open"}
{"q": "On missions with telemetry-reconstructable safe-mode entries (independent of whether a narrative report was filed), estimate the fraction of true autonomous fault-handling events that generated no codable record, broken out by autonomy level, and test whether recording probability itself correlates with autonomy (the silently-handled autonomous save is the H1 cell).", "facet": "measurement", "raised_by": "taleb", "priority": "high", "query_terms": ["safe mode entry telemetry reconstruction unreported", "differential censoring recording probability treatment", "silent evidence survivorship autonomous fault handling", "spacecraft anomaly underreporting completeness by capability"], "status": "open"}
{"q": "Partition loss episodes into (a) loss despite autonomy's correct-but-insufficient action, (b) loss where ground intervention was foreclosed/delayed by the autonomous response, (c) loss where an autonomous action is the documented proximate aggravating cause, and report whether the autonomy-loss cell concentrates in the iatrogenic (b)/(c) categories.", "facet": "rival", "raised_by": "taleb", "priority": "high", "query_terms": ["iatrogenic autonomous recovery worse configuration loss", "autonomy reconfiguration masked fault proximate cause", "via negativa transfer of fragility automation", "commission error autonomous fault management space"], "status": "open"}
{"q": "Specify the episode-level manipulation defining Y_i(1)/Y_i(0); if the only honest manipulation is the program built one autonomy level higher, the unit is the program and conditioning on fault entry conditions on a post-treatment variable. Exhibit two episodes matched on every coded covariate differing only in the autonomy score, or show every autonomy contrast moves complexity, era, and program richness in lockstep.", "facet": "identification", "raised_by": "rubin", "priority": "high", "query_terms": ["potential outcomes well-defined manipulation design property", "matched pair common support autonomy covariate", "no causation without manipulation Holland SUTVA", "program-level versus episode-level treatment confound"], "status": "open"}
{"q": "Estimate the propensity (ordinal autonomy on complexity, distance, age), report covariate overlap/common support across autonomy levels, and the R-squared of autonomy on the controls; determine whether the within-stratum Cox contrast is adjustment or extrapolation past support.", "facet": "empirics", "raised_by": "rubin", "priority": "high", "query_terms": ["propensity score overlap common support diagnostic", "positivity violation extrapolation observational", "R-squared treatment on covariates collinearity", "ordinal treatment propensity overlap survival"], "status": "open"}
{"q": "Run the blinding experiment: have a coder blinded to each episode's end state score autonomy from documents stripped of post-fault narrative, and report the measured agreement between outcome-blind and outcome-aware autonomy codings and whether disagreement correlates with end state.", "facet": "measurement", "raised_by": "rubin", "priority": "high", "query_terms": ["outcome-blind coding inter-rater agreement bias", "design blind to outcomes objective causal inference", "weighted kappa autonomy score reliability blinded", "reverse coding contamination end-state correlation"], "status": "open"}
{"q": "Enumerate the distinct fault-management architectures placed at each ordinal autonomy level, with per-level count and within-level architectural variance, to establish the levels are well-defined treatments (no hidden versions) rather than bags of materially different interventions whose post-fault potential outcomes differ.", "facet": "measurement", "raised_by": "rubin", "priority": "high", "query_terms": ["SUTVA no hidden versions of treatment", "FDIR architecture heterogeneity within autonomy level", "within-level treatment variation attenuation bias", "fault detection isolation recovery architecture taxonomy"], "status": "open"}
{"q": "Audit no-interference: quantify how often between-episode fault-management changes (patch, reconfiguration, operational learning) occurred for multi-episode missions, and restate the estimand to respect that earlier episodes' treatment alters later episodes' potential outcomes; a clustered/frailty variance estimator does not repair an interference-violated estimand.", "facet": "identification", "raised_by": "rubin", "priority": "high", "query_terms": ["SUTVA no interference recurrent events within unit", "between-episode flight software patch reconfiguration", "carryover effect recurrent survival estimand", "frailty clustered variance does not fix interference"], "status": "open"}
{"q": "Specify and report an outcome-side blinded coding protocol: who adjudicated each episode's end state and event-time/censoring, whether that adjudication was blind to the autonomy score, with distinct coders, sealed treatment labels, and inter-rater agreement on end-state and event-time specifically.", "facet": "empirics", "raised_by": "rubin", "priority": "high", "query_terms": ["outcome adjudication blinded to treatment coding protocol", "inter-rater agreement event time censoring classification", "design trumps analysis outcome determination blind", "differential outcome measurement bias survival"], "status": "open"}
{"q": "Draw the DAG for selection-on-fault-entry and name an observed adjustment set that d-separates autonomy from loss given conditioning on entry; if recorded-entry is a collider (common effect of autonomy and unobserved severity, also caused by autonomy) and no observed set closes the path through unobserved severity, declare the effect non-identified rather than report a hazard ratio.", "facet": "identification", "raised_by": "pearl", "priority": "high", "query_terms": ["collider selection bias conditioning treatment-caused node", "back-door criterion d-separation unobserved severity", "endogenous selection bias survival risk set", "principal stratum selection model fault entry"], "status": "open"}
{"q": "Give the graphical test separating a back-door confounder path from a mediated path for distance and complexity (each may be a regime autonomy exploits), split each control into its static pre-treatment component (adjustable) and its realized exploitation-under-distance component (a mediator that must be left unadjusted to avoid overadjustment bias toward the null).", "facet": "identification", "raised_by": "pearl", "priority": "normal", "query_terms": ["overadjustment bias mediator versus confounder", "distance complexity mediator autonomy effect deep space", "temporal precedence confounder pre-treatment", "DAG adjustment set arrow direction credible"], "status": "open"}
{"q": "Run the Grambsch-Therneau scaled Schoenfeld residual test specifically on the autonomy treatment coefficient to confirm time-homogeneity; report the early-versus-late stratified hazard ratios that would falsify the single time-invariant coefficient and force a time-varying-coefficient or time-stratified estimand (the tail-concentration frame predicts a negative residual-vs-time slope).", "facet": "empirics", "raised_by": "pearl", "priority": "high", "query_terms": ["scaled Schoenfeld residual test treatment coefficient cox.zph", "time-varying coefficient Cox proportional hazards violation", "early late hazard divergence tail concentration", "treatment by time interaction step function post-entry clock"], "status": "open"}
{"q": "Add a risk-classification/requirements-stringency regime node R as a common cause (R drives autonomy investment and, via margin/redundancy/abort-disposal policy, survival), show the R->loss path stays open inside a fixed-autonomy stratum (not closed by complexity/distance/age), and name the disposal-policy bypass path that makes R distinct from the already-signed program-quality node.", "facet": "identification", "raised_by": "pearl", "priority": "high", "query_terms": ["NASA payload risk classification Class A-D assurance regime", "omitted common cause confounding margin disposal policy", "back-door path open fixed treatment stratum", "risk-based safety mission assurance design margin"], "status": "open"}
{"q": "Identify a faithful proxy for R (risk-class/margin-requirements regime) codable from exactly NTRS/GAO/JPL/TechPort, distinct from a complexity or cost-class proxy; if risk class and margin-policy requirements are not separately recorded in any of the four sources, declare the autonomy hazard ratio non-identified on regime-confounding grounds.", "facet": "measurement", "raised_by": "pearl", "priority": "high", "query_terms": ["risk classification proxy spacecraft mission documentation", "margin policy redundancy depth coding source", "unmeasured confounder non-identification declaration", "E-value Rosenbaum sensitivity hazard ratio"], "status": "open"}
{"q": "Conditional on R unmeasured, name any instrument that moves autonomy investment yet stays excludable from R (e.g. a TRL-maturation timing shock orthogonal to risk class), or instantiate a front-door rescue via a fully mediating onboard detection-isolation-recovery mechanism unconfounded by R from JPL state logs; state which, if either, the four sources can support.", "facet": "identification", "raised_by": "pearl", "priority": "high", "query_terms": ["instrumental variable exclusion restriction autonomy", "front-door criterion mediator unconfounded outcome", "TRL infusion timing shock instrument risk class", "JPL state logs onboard recovery mediator corpus coverage"], "status": "open"}
{"q": "Report the catalog hygiene before any hazard ratio: how many candidate fault episodes survive an independent second-reader re-coding of the fault-entry event, and the inter-coder kappa on (a) the fault-entry timestamp and (b) the recovery-versus-mission-ending-loss end-state classification.", "facet": "measurement", "raised_by": "mcdowell", "priority": "high", "query_terms": ["inter-coder kappa fault event reproducibility catalog", "second reader re-coding survival count agreement", "event definition reconciliation reproducible spacecraft fault", "catalog hygiene census precondition McDowell"], "status": "open"}
{"q": "Quantify the recording floor: across missions where the true number of safe-mode entries can be independently reconstructed (releasable telemetry/ops logs), what fraction of actual fault episodes appear in NTRS, GAO, and JPL releasable records, and how does the recovery-versus-loss distribution of captured episodes differ from the reconstructed full set?", "facet": "empirics", "raised_by": "mcdowell", "priority": "high", "query_terms": ["spacecraft fault reporting completeness reconstructed ground truth", "survivorship bias on-orbit failure reporting fraction", "safe-mode entry telemetry completeness versus narrative", "observed versus reconstructed recovery loss distribution"], "status": "open"}
{"q": "Bound the selection bias: report the autonomy-level and distance-regime composition of missions whose anomaly records were never releasable (classified/small), and compute a worst-case Manski-style nonparametric bound or tipping-point sensitivity on the unreleasable stratum to test whether the SIGN of the autonomy hazard ratio is identified.", "facet": "identification", "raised_by": "mcdowell", "priority": "high", "query_terms": ["Manski nonparametric bounds selection unreleasable stratum", "tipping point sensitivity sign of effect missing data", "classified small mission anomaly record selection bias", "worst-case bound autonomy hazard ratio sign"], "status": "open"}
{"q": "On the subset of missions covered by two or more of the four registers (NTRS, GAO, JPL anomaly records, TechPort), report the fraction of fault episodes appearing in all overlapping sources versus only one, and the inter-source agreement on fault-entry timestamp, end-state classification, and autonomy level.", "facet": "measurement", "raised_by": "mcdowell", "priority": "high", "query_terms": ["cross-source reconciliation concordance multiple registers", "inter-source agreement fault episode fields", "NTRS GAO TechPort overlap fault catalog", "spacecraft anomaly database reconciliation"], "status": "open"}
{"q": "Show fault episodes coded per spacecraft-year as a time series by launch epoch and demonstrate the autonomy variable is not confounded with the catalog's improving completeness over time, since safe-mode logging granularity, ISA reporting, and TechPort all postdate the earliest missions and later missions are both higher-autonomy and better-recorded.", "facet": "identification", "raised_by": "mcdowell", "priority": "high", "query_terms": ["recording rate time series launch epoch completeness", "non-stationary detection probability confound autonomy era", "trend not snapshot catalog completeness over time", "epoch confounding observability dependent variable"], "status": "open"}
{"q": "Name the independent register of record (an as-flown fault-protection capability description authored before and independent of the anomaly record) against which the reader-coded autonomy score can be externally validated on a held-out subset, and report the agreement rate; absent it, treatment and outcome are drawn from one entangled archive.", "facet": "measurement", "raised_by": "mcdowell", "priority": "high", "query_terms": ["external validation autonomy score independent register", "as-flown fault protection capability description", "held-out subset agreement rate treatment coding", "independent anchor treatment measurement spacecraft"], "status": "open"}
{"q": "Establish whether NTRS, GAO, JPL anomaly records, and TechPort can reconstruct the within-episode state path (time-ordered sequence of autonomous actions and state responses) for enough episodes to fit a coarse stock-and-flow recovery model as a rival to the Cox reduction, or concede on the record that the data cannot distinguish the hazard-ratio story from a structural-recovery story.", "facet": "measurement", "raised_by": "forrester", "priority": "high", "query_terms": ["intra-episode event log spacecraft fault state sequence", "stock and flow recovery model within episode", "anomaly record granularity sub-episode timeline schema", "NTRS GAO TechPort data dictionary fault episode"], "status": "open"}
{"q": "Specify and estimate at least one model in which autonomy and the one-way-light-time delay enter as interacting feedback (autonomy-by-light-time interaction, effect allowed to grow as delay lengthens) and show whether the protective effect is endogenous to the loop structure; a stable hazard ratio across light-time regimes falsifies the deep-space autonomy rationale, an unstable one falsifies the single proportional-hazards coefficient.", "facet": "identification", "raised_by": "forrester", "priority": "normal", "query_terms": ["autonomy light-time interaction term Cox model", "delayed feedback loop ground intervention deep space", "effect modification distance regime hazard", "time-varying covariate light-time delay survival"], "status": "open"}
{"q": "Specify the simultaneous-equation / stock-and-flow structure (latent regime stock = program risk aversion; two measured outflows = autonomy level and abort/margin/test stringency) needed to separate the autonomy lever from the co-produced risk-posture regime, and show whether the four sources record a requirements-stringency or margin-policy variable at mission level at all; concede if the Cox specification cannot represent a co-produced-regime rival.", "facet": "identification", "raised_by": "forrester", "priority": "high", "query_terms": ["simultaneous equation latent regime two outflows autonomy stringency", "instrumental variable causal hazard ratio endogenous treatment", "co-produced regime confound balancing loop risk aversion", "margin policy abort threshold mission-level variable"], "status": "open"}
{"q": "Partition coded episodes by an independently coded stringency proxy, hold autonomy level fixed, and test whether survival still varies with stringency and whether stringency-matched high- versus low-autonomy episodes have indistinguishable hazard; state whether a stringency proxy is independently codable per episode or whether stringency and autonomy are so collinear in the sources that the regime rival is unfalsifiable.", "facet": "empirics", "raised_by": "forrester", "priority": "high", "query_terms": ["stringency-matched falsification regime versus lever", "independently coded margin abort test depth proxy", "collinearity treatment regime unfalsifiable", "within-stratum stringency coefficient survival"], "status": "open"}
{"q": "Distinguish the transferable-lever interpretation from the regime-symptom interpretation of the autonomy hazard ratio (adding autonomy software without the conservative margin/abort/test posture should yield little benefit); if the four sources cannot identify which holds, state in the contribution that the estimate is a within-regime association of unknown transferability rather than an architecture-trade input.", "facet": "rival", "raised_by": "forrester", "priority": "high", "query_terms": ["leverage point transferable lever versus symptom", "policy resistance single parameter intervention low leverage", "within-regime association transferability architecture trade", "endogenous treatment lever identification system dynamics"], "status": "open"}
{"q": "Name from NTRS+JPL records a set of high- and low-autonomy fault episodes genuinely comparable on complexity, distance regime, and age, count how many such matched pairs exist in the assembled population, and show whether that number can carry a pooled hazard ratio or collapses to a handful of incomparable cases.", "facet": "measurement", "raised_by": "yin", "priority": "high", "query_terms": ["matched pair count comparable fault episodes autonomy", "case comparability complexity distance age matching", "unit of analysis defensible case boundary", "handful of cases pooled hazard ratio adequacy"], "status": "open"}
{"q": "For the program-quality rival, point to specific fault episodes where a high-autonomy and a low-autonomy spacecraft of demonstrably comparable program quality faced comparable faults and show the survival difference tracked autonomy not quality, rather than asserting from the coefficient's sign that residual bias is bounded.", "facet": "rival", "raised_by": "yin", "priority": "high", "query_terms": ["program-quality-comparable matched episodes autonomy survival", "rival explanation pattern matching case by case", "sign-of-bias upper bound versus matched evidence", "within-case discrimination autonomy versus quality"], "status": "open"}
{"q": "State the generalization claim exactly (statistical estimation of a population parameter versus analytic test of a theoretical proposition in a narrow flagship stratum) and report from the coverage table how many spacecraft and distinct mission classes contribute events.", "facet": "identification", "raised_by": "yin", "priority": "normal", "query_terms": ["analytic versus statistical generalization case study", "coverage table contributing spacecraft mission classes", "flagship stratum theoretical proposition generalization", "replication logic narrow stratum survival"], "status": "open"}
{"q": "State the program-quality/requirements-stringency regime as a positive rival theory and write the one observable prediction it makes that H1 does not (within a single autonomy level survival still tracks independently-coded stringency, and high-autonomy losses cluster in weakest-stringency programs), ensuring the prediction is mutually exclusive from H1's rather than a relabeled nuisance term.", "facet": "rival", "raised_by": "yin", "priority": "high", "query_terms": ["positive rival theory mutually exclusive prediction", "stringency regime marker not lever falsifiable", "pattern matching competing theories different patterns", "test depth program quality decision variable"], "status": "open"}
{"q": "Name specific fault-episode pairs from NTRS+GAO+JPL+TechPort where a high-autonomy and a low-autonomy spacecraft of comparable complexity, distance regime, and age faced a comparable fault and the two programs differ in documented requirements-stringency, and show whether survival tracks autonomy or stringency; if no named stringency-discriminating pairs can be assembled, the rival is unmeasured rather than ruled out.", "facet": "identification", "raised_by": "yin", "priority": "high", "query_terms": ["stringency-discriminating matched anomaly pairs", "named fault episode dyads autonomy stringency", "lever versus marker population coefficient adjudication", "requirements stringency coding NTRS GAO JPL TechPort"], "status": "open"}
{"q": "State whether NTRS, GAO, JPL anomaly records, and TechPort can code a requirements-stringency-regime ordinal separately from the autonomy score via a blind double-coding pilot with a reported separability statistic; if not separable, concede autonomy is constitutively confounded with the regime that selected it and the unit of value-attribution is the regime, not the onboard system.", "facet": "measurement", "raised_by": "yin", "priority": "high", "query_terms": ["blind double-coding separability stringency autonomy", "constitutive confounding selection by quality regime", "ROBINS-I baseline confounding confounding by indication", "inter-coder separability pilot held-out sample"], "status": "open"}
{"q": "MISSING ANGLE (moderator divergence injection for Phase 1): no panelist supplied an external, executed empirical benchmark for the expected magnitude and direction of an autonomy-versus-survival or capability-versus-reliability hazard ratio from an adjacent flown-hardware literature (e.g. spacecraft subsystem reliability by mass class, redundancy, or autonomy proxy). The interrogation grounded the design's internal feasibility and the threats, but never anchored a prior expected effect size against the Castet-Saleh-style population reliability record or any flown autonomy-reliability comparison. Phase 1 must close this with an external effect-size prior so the eventual estimate is benchmarked, not free-floating.", "facet": "empirics", "raised_by": "moderator", "priority": "high", "query_terms": ["spacecraft subsystem reliability autonomy redundancy hazard ratio", "Castet Saleh on-orbit failure population reliability mass class", "fault management capability versus mission success empirical", "expected effect size autonomy spacecraft survival prior"], "status": "open"}