Category G: Communication Agents

Academic writing, dissemination, and humanization pipeline

Communication agents handle the entire publication lifecycle from journal selection through peer review response and preregistration, with advanced humanization capabilities to transform AI-generated text into natural academic prose while preserving scholarly integrity. 4 agents cover journal matching, publication management, and academic style optimization.

Core Principle

Clear communication, strategic publication, and natural academic writing

Journal Matcher

Differentiated journal recommendation using multi-dimensional matching beyond impact factor

MEDIUM • Sonnet

Light VS

Capabilities

•Multi-dimensional journal matching (scope, methodology, paradigm)
•Impact factor contextualization (field-normalized metrics)
•Open access vs. subscription analysis
•Review timeline and acceptance rate data
•Editorial board alignment check

VS Process

Avoid IF-centric recommendations, present differentiated options considering research paradigm, methodology fit, and strategic career goals

Example

Input

"Mixed methods study on AI in education - where should I submit?"

Output

Option A (High IF): Computers & Education (IF 11.2, 15% acceptance, 6-month review) | Option B (Methodology fit): Journal of Mixed Methods Research (IF 4.5, mixed methods specialty, 3-month review) | Option C (Open access): International Journal of Educational Technology (OA, faster review, broader reach)

Triggers:journalsubmissionimpact factorwhere to publishtarget journal

Publication Specialist

Comprehensive publication management — from manuscript writing and audience adaptation to peer review response, preregistration, consistency checking, reporting checklists, and reproducibility auditing

MEDIUM • Sonnet

Enhanced VS 3-Phase

Capabilities

•Abstract writing (structured, unstructured)
•Plain language summaries for non-specialists
•Conference presentation design
•Research poster layout
•Peer review response strategy and rebuttal drafting (absorbed G3)
•Preregistration template completion — OSF, AsPredicted, Registered Reports (absorbed G4)
•Manuscript consistency checking across sections
•Reporting checklists (CONSORT, STROBE, COREQ, etc.)
•Reproducibility auditing and documentation

VS Process

Phase 1: Analyze communication needs and audience | Phase 2: Draft content with consistency checks | Phase 3: Validate against reporting guidelines and reproducibility standards

Example

Input

"Write plain language summary of my RCT on AI chatbots"

Output

Technical abstract (250 words) for journal | Plain language summary (150 words) for general public: "AI chatbots helped students improve speaking skills by 23%. Students practiced more often because chatbots were available 24/7 and didn't judge mistakes." | Key visual: Before/after speaking proficiency chart

Triggers:abstractplain languagepresentationwrite papermanuscriptpeer reviewreviewerrevisionresponse letterrebuttalpreregistrationpre-registerOSFAsPredictedregistered reportconsistencyreporting checklistreproducibility

Academic Style Auditor

Detect AI writing patterns across 24 categories with probability scoring and risk classification

MEDIUM • Sonnet

Enhanced VS 3-Phase

CP_HUMANIZATION_REVIEW

Capabilities

•24-category AI pattern detection (based on Wikipedia AI Cleanup)
•Probability scoring (High/Medium/Low risk)
•Phrase-level granularity with context preservation
•Academic vs. generic AI pattern differentiation
•Citation integrity verification

VS Process

Phase 1: Scan for 24 pattern categories | Phase 2: Assign probability scores | Phase 3: Generate risk report with prioritized recommendations

Example

Input

"Check my methods section for AI patterns"

Output

HIGH RISK: 12 instances of hedging clusters ("it is important to note that", "it should be emphasized") | MEDIUM RISK: 5 instances of unnecessary semantic intensifiers | LOW RISK: 2 instances of list-heavy structure | Recommendation: Prioritize hedging removal, moderate intensifier reduction | Citations: All intact, no hallucinated references

Detection Patterns (Sample)

Hedging clustersUnnecessary semantic intensifiersList-heavy structureRobotic neutralityTransition word overusePassive voice dominanceGeneric academic phrasesRepetitive sentence structure

Triggers:AI patterncheck AI writingstyle auditAI detection

Academic Style Humanizer

Transform AI patterns into natural academic prose while preserving scholarly integrity and citations

HIGH • Opus

Full VS 5-Phase

CP_HUMANIZATION_VERIFY

Capabilities

•Pattern-specific transformation strategies
•Three modes: Conservative, Balanced, Aggressive
•Citation preservation and verification
•Meaning integrity validation
•Diff tracking for all changes

VS Process

Stage 1: Parse G5 audit report | Stage 2: Generate transformation alternatives | Stage 3: Human selects mode | Stage 4: Apply transformations | Stage 5: F5 verification

Example

Input

"Humanize: It is important to note that the results demonstrate a significant effect"

Output

CONSERVATIVE (high-risk only): "The results demonstrate a significant effect" (removed hedging) | BALANCED: "Results show a significant effect" (active voice + concise) | AGGRESSIVE: "Our data reveal a meaningful impact" (varied vocabulary, first person)

Transformation Modes

Conservative

High-risk patterns only • Journal submissions

Balanced

High + medium-risk • Most academic writing (recommended)

Aggressive

All patterns • Blog posts, informal writing

Triggers:humanizehumanizationnatural writingtransformremove AI patterns

Humanization Pipeline (G5 → G6 → F5)

Three-stage process for transforming AI-generated text:

Audit

Detect AI patterns, assign risk scores

Transform

Apply mode-specific humanization

Verify

Validate transformation integrity

Checkpoint Integration

Communication agents enforce publication readiness:

CP_HUMANIZATION_REVIEW

AI pattern audit complete

CP_HUMANIZATION_VERIFY

Transformation verified for integrity

Typical Publication Workflow

Select target journal

Write manuscript, preregister, and prepare reporting checklists

G5 → G6

Audit and humanize draft

CP_HUMANIZATION_REVIEW

Respond to peer review and verify reproducibility

Ethics Note: Responsible AI Use

Humanization helps express ideas naturally—it does NOT make AI use "undetectable." Researchers must follow institutional and journal AI disclosure policies. Diverga promotes transparent AI-assisted research communication, not deception.

Master Research Communication

Use Category G agents for publication success and natural academic writing.

Explore Category X: Cross-Cutting