{"schema_version":1,"experiment_id":"eoa_inverse_innovation_exp09_archetype_breadth150_20260804","cell_id":"authority_mentor_relationship_anchoring__computer_science","arm":"BREADTH_PROBE_ONE_SHOT","candidate_id":"authority_mentor_relationship_anchoring__computer_science__P1","proposal_index":1,"version":0,"title":"Bounded Incident-Judgment Mentorship for New On-Call Engineers","problem":"New software engineers joining a production on-call rotation may know the monitoring tools and written runbooks yet lack the tacit professional judgment needed when an incident falls between documented cases. They must interpret incomplete signals, choose between waiting and intervening, limit blast radius, disclose uncertainty, and escalate without treating escalation as failure. Those norms are embedded in experienced responders' situated decisions and are not fully represented by procedural documentation.","actors":["Newly admitted on-call software engineer as mentee","Experienced incident commander or site-reliability engineer as mentor","Service owner accountable for production changes","Independent second-opinion mentor or incident-review facilitator","Engineering manager or training coordinator administering safeguards","Users and downstream services exposed to incident-response decisions"],"observable_state":"During retrospective incident exercises or supervised incidents, a new responder can recite the applicable runbook but either freezes when evidence conflicts, proposes an irreversible or broad intervention without stating its blast radius, or conceals uncertainty until prompted. Review artifacts show that the responder cannot explain when to stop following the runbook, seek another interpretation, or prefer a reversible diagnostic step.","consequence":"Ambiguous incidents can produce delayed escalation, unnecessarily risky mitigation, wider service disruption, or incomplete disclosure of mistakes and uncertainty, preventing the responder from becoming safely independent.","affected_objective":"Safe, transparent, and increasingly independent production-incident judgment under incomplete information.","intervention":"Create a time-bounded Incident-Judgment Mentorship for each consenting new on-call engineer. Select a mentor for demonstrated incident competence, calm disclosure of uncertainty, ethical standing, and lack of direct performance-rating authority. Define the enculturation payload as reversibility, evidence-proportional action, blast-radius containment, early escalation, explicit uncertainty, and blameless correction. The mentee observes the mentor interpret sanitized incident replays and, where operationally permitted, real incidents; each observation is followed by a structured dialogue separating shared response norms from the mentor's personal style. The pair then conducts joint simulated responses with corrective feedback before the mentee handles transfer cases independently. A second-opinion mentor, confidential reassignment route, and explicit graduation criteria prevent single-mentor capture and dependency.","structural_mapping":[{"archetype_element":"Legitimate Mentor Anchor","domain_realization":"A screened incident commander or senior reliability engineer whose legitimacy rests on demonstrated response competence, transparent uncertainty management, care for affected people, and peer-reviewed conduct rather than rank alone."},{"archetype_element":"Mentee Readiness and Consent Boundary","domain_realization":"Participation, observation scope, confidentiality, recording rules, withdrawal, and reassignment are agreed before pairing; mentorship is separated from performance evaluation."},{"archetype_element":"Relational Safety Container","domain_realization":"Scheduled private debriefs permit the mentee to reveal confusion, challenge a decision, and discuss mistakes without the mentor controlling ratings or on-call eligibility."},{"archetype_element":"Cultural Norm and Value Payload","domain_realization":"The named payload is reversible action, evidence proportionality, bounded blast radius, prompt escalation, explicit uncertainty, blameless disclosure, and protection of users over personal heroics."},{"archetype_element":"Modeled Practice and Judgment Window","domain_realization":"The mentee observes the mentor narrate tradeoffs while diagnosing sanitized incident replays and may shadow authentic response only under the production incident commander's rules."},{"archetype_element":"Dialogic Interpretation Loop","domain_realization":"A post-case protocol asks what evidence changed the mentor's mind, which actions were reversible, which norms were general, and which choices reflected personal style or contested practice."},{"archetype_element":"Secondary Reference Anchor","domain_realization":"A different qualified responder reviews selected cases and remains available for confidential second opinions, allowing guidance to be compared with documented service standards and another professional interpretation."},{"archetype_element":"Progressive Autonomy Release","domain_realization":"The sequence moves from observation, to jointly controlled simulations, to independent transfer cases, with production authority granted only through the organization's existing qualification process."},{"archetype_element":"Autonomy and Exit Safeguard","domain_realization":"The mentee may decline live shadowing, request reassignment, consult the secondary mentor, or leave the program without the primary mentor imposing employment or credential consequences."}],"mechanism_mapping":[{"mechanism_slug":"guided_shadowing_with_debrief","role":"Expose the mentee to situated incident interpretation and immediately convert tacit cues and tradeoffs into discussable reasons.","counterfactual_removal":"Without debrief, observation may produce surface imitation without revealing why escalation, waiting, or a narrow intervention was appropriate."},{"mechanism_slug":"joint_practice_with_corrective_feedback","role":"Let mentor and mentee respond to incident simulations together while corrections can be explained before unsafe habits become independent practice.","counterfactual_removal":"Without joint practice, the mentor cannot observe how the mentee applies the norms or correct mistaken transfer before autonomy increases."},{"mechanism_slug":"reflective_norm_dialogue","role":"Distinguish profession-level response norms from mentor preference and invite the mentee to contest or adapt interpretations.","counterfactual_removal":"Without reflective dialogue, mentor authority may encourage compliance while leaving the learner unable to justify or revise the inherited judgment."},{"mechanism_slug":"mentor_rotation_or_second_opinion_channel","role":"Provide an alternate interpretation and a confidential safety route when the primary relationship is biased, coercive, or mismatched.","counterfactual_removal":"Without a second anchor, the mentee may mistake one responder's habits for the field's norms or remain dependent on an unsafe relationship."}],"causal_chain":["Written runbooks leave residual ambiguity about tradeoffs, exceptions, emotional stance, and escalation norms.","A legitimate mentor repeatedly makes those judgments observable in bounded incident cases.","Guided debriefs connect visible actions to evidence, values, uncertainty, and limits rather than leaving the mentee to imitate behavior.","Joint simulations let the mentee practice the named norms and receive timely corrective explanations inside a psychologically safe relationship.","A secondary reference exposes mentor-specific preferences and contested interpretations before they are treated as universal standards.","Progressive autonomy shifts responsibility to the mentee while independent cases reveal whether judgment transfers beyond the mentor relationship.","If the mechanism works, the mentee should become better able to articulate and apply bounded incident-response norms without seeking the mentor's approval for each decision."],"baseline":"Existing onboarding consists of service documentation, runbooks, tool instruction, incident-review notes, and occasional ad hoc pairing. These resources communicate explicit procedures but do not establish a governed relationship for observing, interpreting, questioning, and independently transferring tacit response norms.","nearest_rivals":["Cognitive apprenticeship through narrated expert incident walkthroughs: it makes reasoning visible, but need not use trusted relational authority to transmit response identity, disclosure norms, and belonging over repeated interactions.","Tabletop incident exercises: they provide situated practice, but a facilitator-led exercise need not create a bounded mentor relationship or a progressive transfer of professional judgment.","Runbook expansion and decision trees: they reduce procedural ambiguity but cannot exhaustively encode when evidence is weak, values conflict, or an exception should be escalated.","Routine manager shadowing: it provides exposure to authority but lacks the separation from evaluation, explicit cultural payload, dialogic safeguards, and alternate anchor required here."],"remaining_contrastive_claim":"The candidate's distinctive causal claim is that repeated participation with a trusted, role-bounded incident mentor can make uncertainty disclosure, restraint, escalation, and user-protective judgment personally salient and reflectively internalized; equivalent documentation or one-off expert explanation would not instantiate that relational enculturation pathway. This is a testable structural distinction, not a novelty or effectiveness claim.","authority_safety":{"decision_authority":"The designated production incident commander and service owner retain all authority over live diagnosis, mitigation, access, and rollback. The mentor controls only agreed training activities and cannot independently change production, grant credentials, or determine the mentee's employment evaluation.","authorized_first_step":"The training coordinator may recruit consenting volunteers and run a retrospective, sanitized incident-replay session using existing access permissions; no live-system action is authorized.","excluded_actions":["Allowing a mentee to execute a production change solely because the mentor approves it","Granting the mentor unilateral control over performance ratings, promotion, or on-call eligibility","Sharing restricted incident, customer, or security data outside existing authorization","Recording confidential debriefs without explicit participant consent","Penalizing withdrawal, reassignment requests, disagreement, or consultation with the second-opinion mentor","Using the program to bypass existing access controls, change review, incident command, or rollback procedures"],"halt_rollback":"Pause the pairing if either participant reports fear, retaliation, confidentiality breach, dependency, discriminatory guidance, pressure to conceal uncertainty, or inability to question decisions. Cancel planned sessions, remove the mentor from training access, preserve only already-authorized audit records, offer a confidential second-opinion review, and permit reassignment or withdrawal without penalty. Any live operational action remains governed by the service's normal rollback procedure."},"negative_tests":{"strongest_counterevidence":"Mentees may perform no better on independent ambiguous cases than peers receiving the same cases and explanations without a continuing mentor relationship, or they may improve only on cases closely resembling the mentor's examples while becoming more deferential and less willing to disclose disagreement.","problem_falsifier":"The problem is falsified for the selected cohort if observed incident decisions are fully determined by accessible runbooks and missing factual or tool knowledge, with no recurring residual need for tacit interpretation, uncertainty disclosure, escalation judgment, or norm conflict.","intervention_falsifier":"The intervention is undermined if blinded reviewers find no improvement in the coherence of independent transfer-case rationales after mentor observation and debrief, or if mentees cannot distinguish shared response norms from mentor-specific style, show increased approval-seeking, or report reduced freedom to question and exit.","risks":["Mentor idiosyncrasies or outdated operational habits may be transmitted as professional norms.","Authority and career dependence may suppress disagreement despite formal consent.","Mentor selection may reproduce favoritism or exclude responders with different legitimate styles.","Confidential incident details may leak through examples or debriefs.","A mentee may become dependent on mentor reassurance instead of developing independent judgment.","Simulated competence may be mistaken for authorization to act in production.","Additional relationship work may displace documentation improvements that remain necessary."]},"next_evidence_step":"Run one bounded retrospective pilot with two screened mentors and six consenting new or prospective on-call engineers over three sanitized incident cases. Before mentorship, each mentee independently records a response, rationale, uncertainty statement, escalation threshold, and proposed blast-radius limit for one case. The mentee then observes a mentor walkthrough and completes a structured debrief on a second case, followed by an independent transfer response to the third. Two reviewers who did not conduct the mentorship score predeclared features of the rationales without seeing session order, while participants separately report whether they could question, seek a second opinion, and withdraw. Treat the exercise only as a feasibility and mechanism check; do not change production privileges or infer operational effect size.","prior_art_status":"UNSEARCHED","diversity_from_prior_proposals":"Not assessed against other proposals because runtime isolation forbids inspecting them; this one-shot candidate is defined solely by mentor-anchored transmission of tacit production-incident judgment.","revision_record":{"parent_version":null,"progress_targets_addressed":["Initial one-shot construction from the supplied archetype and computer-science domain card"],"conceptual_changes":[],"operational_changes":[],"evidence_changes":[],"claim_changes":[]}}