{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp12_substrate_denial72_20260805","research_id":"eoa_inverse_innovation_exp12_light_screen_20260805","cell_id":"authority_mentor_relationship_anchoring__computer_science","search_lanes":{"direct_problem_and_intervention":{"queries":["incident commander training mentoring shadowing tabletop exercises stable mentor","site:sre.google incident response training shadowing incident commander mentor simulation"],"source_ids":["SRC1","SRC2","SRC3","SRC4"],"no_result_note":null},"synonyms_and_historical_terms":{"queries":["cognitive apprenticeship software engineering mentoring incident response simulation debrief","SRE incident management apprenticeship mentorship empirical study novice on-call"],"source_ids":["SRC1","SRC3"],"no_result_note":"No retained source evaluated a durable mentor dyad against an otherwise matched rotating-expert condition."},"products_practices_and_standards":{"queries":["site:pagerduty.com incident response training incident commander shadow","site:nist.gov incident response training tabletop exercises roles escalation uncertainty","site:google.github.io/sre-book incident response training simulation role playing incident commander"],"source_ids":["SRC1","SRC2","SRC3"],"no_result_note":null},"component_combination":{"queries":["software incident response decision making under uncertainty cognitive load postmortem research","empirical study production software incident response human factors decision making uncertainty engineers","incident commander shadow reverse shadow mentor debrief simulation"],"source_ids":["SRC1","SRC2","SRC3","SRC4"],"no_result_note":null}},"sources":[{"source_id":"SRC1","title":"Incident Commander","publisher":"PagerDuty Incident Response Documentation","url":"https://response.pagerduty.com/training/incident_commander/","source_type":"FIRST_PARTY_PRODUCT","claims_supported":["PagerDuty's incident-commander training combines repeated simulations, a full-shift shadow period, and reverse shadowing in which a trainee leads while a current commander can take over.","The documented commander role requires rapid assessment, consideration of action risk and backup plans, clear communication, and post-incident debriefing.","This is close domain-specific practice combining observation, progressive responsibility, and experienced-responder support."]},{"source_id":"SRC2","title":"Being On-Call","publisher":"Google SRE, published by O'Reilly Media","url":"https://sre.google/sre-book/being-on-call/","source_type":"OFFICIAL_GUIDANCE","claims_supported":["Production on-call work is time-critical and can have serious service consequences.","Google reports that pressure can prompt incorrect choices, unreflective immediate action, confirmation bias, and poorly supported intuition.","The guidance recommends clear escalation paths, incident procedures, blameless postmortems, drills, and automation as complementary safeguards."]},{"source_id":"SRC3","title":"Accelerating SREs to On-Call and Beyond","publisher":"Google SRE, published by O'Reilly Media","url":"https://sre.google/sre-book/accelerating-sre-on-call/","source_type":"OFFICIAL_GUIDANCE","claims_supported":["Google advises against training strictly through checklists and playbooks and describes incident response as requiring hypotheses, improvisation, and experience under ambiguity.","The chapter combines structured learning, mentors, disaster role-play, shadowing, review of reasoning after outages, and reverse shadowing with progressively greater responsibility.","It explicitly recommends shadowing multiple team members, providing a concrete rotating-expert analogue to the proposal's stable-mentor comparison."]},{"source_id":"SRC4","title":"Failures and Fixes: A Study of Software System Incident Response","publisher":"arXiv (ICSME preprint)","url":"https://arxiv.org/abs/2008.11192","source_type":"PRIMARY_RESEARCH","claims_supported":["The qualitative study analyzed 30 software incidents using 15 engineer interviews and 15 public incident reports.","It found that failures can cascade into major outages and that engineers may not understand supported systems' limits until those limits are exceeded.","The study supports the consequential and partly unpredictable character of production incident response, but does not isolate novice stewardship internalization or mentor continuity."]}],"problem_evidence":{"status":"PARTLY_SUPPORTED","finding":"The problem's general substrate is visible: production responders work under time pressure and uncertainty, stress can induce unreflective or poorly supported choices, unfamiliar limits can contribute to cascading outages, and first-party guidance treats checklists alone as insufficient for improvisational readiness. The narrower assertion that technically qualified novices specifically fail to internalize stewardship priorities—rather than lacking information, tooling, decision rights, or practice—was not directly demonstrated by the retained evidence.","source_ids":["SRC2","SRC3","SRC4"]},"closest_prior_art":[{"name":"PagerDuty incident-commander shadow and reverse-shadow training","source_ids":["SRC1"],"overlap":"Direct-domain practice includes simulations, repeated command practice, observation of an experienced commander, later trainee leadership under that commander's backup, risk assessment, backup planning, communication norms, and debriefing.","remaining_difference":"The source does not specify a six-session stable non-evaluative dyad, a consent and exit charter, explicit classification of policy versus invariant versus mentor style, an independent second mentor, or a matched test of stable continuity against rotating experts."},{"name":"Google SRE structured onboarding, disaster role-play, mentorship, shadowing, and reverse shadowing","source_ids":["SRC2","SRC3"],"overlap":"This practice combines structured preparation, realistic incident scenarios, hypothesis articulation, mentor review, observation, retrospective reasoning review, progressive responsibility, and safeguards such as escalation procedures and automation.","remaining_difference":"Google explicitly suggests exposure to multiple on-callers and does not claim that one stable mentor relationship independently improves reflective stewardship transfer when content, cases, demonstration, and feedback are held constant."},{"name":"Procedural, tabletop, and tooling safeguards for incident judgment","source_ids":["SRC1","SRC2","SRC3"],"overlap":"The sources describe explicit roles, escalation paths, simulations, decision procedures, postmortems, and automation that address unsafe incident choices without depending on a durable mentor relationship.","remaining_difference":"These practices are strong rival explanations and controls, but the retained sources do not report the proposal's unfamiliar-case comparison of a stable mentor with equally resourced rotating experts."}],"prior_art_disposition":"ADJACENT_PRIOR_ART","contrastive_claim_remaining":"Holding explicit stewardship content, reconstructed cases, mentor pool, demonstrations, practice time, corrective feedback, technical safeguards, and qualification rules constant, continuity with one legitimate, non-evaluative mentor produces measurably better independent transfer of stewardship reasoning to an unfamiliar ambiguous incident than rotating among the same experts, without reducing uncertainty disclosure, willingness to challenge advice, or perceived ability to exit.","contrastive_claim_falsifier":"The claim is falsified if the rotating-expert arm matches or exceeds the stable-mentor arm on blinded unfamiliar-case scores for stated hypotheses and uncertainty, reversibility and rollback, escalation timing, and separation of policy or invariant from personal style, while showing equal or better agency and psychological safety. It is also undermined if the charter alone removes the errors or if baseline assessment attributes them to missing telemetry, technical knowledge, tooling, incentives, or decision rights.","gates":{"adequate_source_search":{"status":"PASS","rationale":"The bounded search covered direct terminology, apprenticeship and mentoring synonyms, first-party incident-response practices, standards-oriented terminology, and component combinations. Exactly four opened sources span PagerDuty, Google/O'Reilly, and primary incident research.","source_ids":["SRC1","SRC2","SRC3","SRC4"]},"supported_problem":{"status":"PASS","rationale":"The general problem of consequential, uncertainty-laden incident judgment and failures of deliberate reasoning is supported, while the proposed novice-specific enculturation diagnosis remains only partly supported and is appropriately assigned a problem falsifier.","source_ids":["SRC2","SRC3","SRC4"]},"distinct_testable_claim":{"status":"PASS","rationale":"Despite close collision with established shadowing, reverse shadowing, simulation, and mentorship practices, the incremental claim about stable relational continuity versus matched rotating experts remains distinct and falsifiable.","source_ids":["SRC1","SRC3"]},"bounded_next_test":{"status":"PASS","rationale":"The proposed eight-person, four-week offline randomized mechanism probe fixes cases, content, time, mentor pool, and feedback; uses an unfamiliar replay and blinded predeclared scoring; and expressly avoids effectiveness, qualification, or production-access conclusions.","source_ids":["SRC1","SRC2","SRC3"]},"no_obvious_safety_or_authority_stop":{"status":"PASS","rationale":"The first step is offline and voluntary, uses approved synthetic or deidentified artifacts, preserves existing production command and access authority, excludes employment evaluation, and supplies reassignment, complaint, confidentiality, and halt routes. Mentor influence and artifact privacy remain risks requiring the stated oversight, but no obvious stop bars the bounded probe.","source_ids":["SRC1","SRC2"]}},"screen_survival":true,"world_novelty_boundary":"This bounded four-source public-web screen found close adjacent practices but no retained controlled test of the stable-mentor contrast. It cannot establish world novelty, patentability, market size, expert acceptance, realized value, or absence of additional prior art in proprietary programs, patents, non-indexed materials, or other domains."}