academic-review

Academic Review

Purpose

Conduct structured multi-phase literature reviews using ike corpus documents via MCP tools. Produce a synthesised, evidence-grounded analysis that maps the landscape of available sources, identifies key themes and contested areas, and provides traceable citations for every claim.

Before Starting

Call get_server_instructions before doing anything else. The server may return corpus-specific research guidance — domain-specific instructions such as "always note jurisdiction for legal documents" or "recommend professional consultation for medical sources". Apply that guidance for the rest of the session.

Five-Phase Workflow

Phase 1 — Discovery

Map the available source landscape before committing to any particular reading strategy.

• Call list_corpora to understand available collections, their scope, and coverage areas.
• Run 3–5 search_documents queries using different angles to cover the topic space:
  • Broad topic query: natural-language description of the subject area — casts the widest net.
  • Specific term query: names, dates, technical terms, or concepts likely to appear verbatim in sources.
  • Methodological query: combine research methods or analytic approaches with the topic (e.g. "empirical evidence", "case study analysis", "meta-analysis").
  • Contrarian query: combine "criticism", "limitations", "counter-argument", or "contested" with the topic to surface dissenting views.
  • Cross-disciplinary query (optional): adjacent fields that may offer different perspectives on the topic.
• Collect a candidate pool of 10–20 results across all queries.
• Note which corpora and documents appear frequently across multiple queries — these are likely central sources.
• Use list_documents to check for documents whose titles suggest relevance but did not surface in search results.

Consult references/search-strategy.md for detailed query construction patterns and decision points.

Phase 2 — Source Assessment

Evaluate candidates before committing to deep reading.

• For each candidate document, call get_document_metadata to retrieve: publication year, document type, authors, publisher, and any available journal or series information.
• Call get_page_index to understand the document's scope and internal structure — this reveals chapter or section titles and helps assess depth of coverage before reading.
• Apply inclusion criteria to select the final reading set:
  • Relevance: how directly does the document address the review topic?
  • Recency: more recent sources take precedence unless earlier work is foundational or canonical.
  • Authority: consider source type (primary research vs. secondary analysis), institutional affiliation where available, and peer-review status where indicated in metadata.
  • Depth: does the document cover the topic substantively, or only in passing?
• Select 5–10 sources for deep reading. If the user has specified a preference for source count, honour it. If the corpus does not yield sufficient sources, note this limitation explicitly rather than padding with peripheral material.

Phase 3 — Deep Reading

Extract the substance from selected sources systematically.

• For each selected source, use get_document_section to retrieve the sections identified from the page index that are most relevant to the review topic. Prefer section-level retrieval over get_document_content (full document retrieval) to stay within context limits and maintain precision.
• Use get_document_chunk for specific passages when a precise extract is needed for direct quotation or close citation.
• Take structured notes for each source covering:
  • Key claims: the main arguments or findings advanced by the source.
  • Evidence: the type and strength of evidence used to support those claims.
  • Methodology: how the source approaches the topic (empirical study, theoretical analysis, case study, archival research, etc.).
  • Page references: specific locations for claims that will be cited in the review.
  • Relation to other sources: whether this source corroborates, contradicts, or extends other sources in the reading set.

Phase 4 — Synthesis

Construct the review by integrating findings across sources, not summarising them source-by-source.

Structure the review with the following sections:

Research Landscape Overview Describe what the corpus covers on this topic: the range of source types, time period represented, disciplinary perspectives present, and any notable gaps in coverage. Set expectations for what the review can and cannot address based on available material.

Key Findings and Themes Identify the major conclusions, arguments, and points of agreement across sources. Organise findings thematically rather than by source. Each theme should draw on multiple sources where possible.

Methodological Approaches Describe how sources approach the topic differently — empirical vs. theoretical, qualitative vs. quantitative, archival vs. analytical. Note where methodological differences may explain divergent findings.

Contested Areas Identify where sources disagree and present the evidence from both sides. Do not silently resolve conflicts by selecting one side — present the disagreement, characterise the strength of evidence for each position, and let the evidence speak.

Gaps and Future Directions Identify what the corpus does not cover: unanswered questions, underexplored angles, topics adjacent to the review that lack coverage. Be honest about limitations — overstating corpus coverage undermines the review's credibility.

Quality requirements during synthesis:

• Every factual claim must be traceable to a specific source (author, title) and page reference.
• Distinguish direct findings (what a source explicitly states) from synthesis (the reviewer's interpretation or cross-source inference). Mark interpretations as such.
• Flag conflicting evidence explicitly using language that signals disagreement (see references/synthesis-guide.md for hedging patterns).
• Do not blend corpus findings with general knowledge without clear attribution.

Phase 5 — Citation

Produce a complete and accurate citation apparatus.

• Call get_citation_link for every source referenced in the review.
• Present citations in a consistent format. Default to Chicago Notes & Bibliography unless the user specifies a preference (Chicago Author-Date, Taylor & Francis Harvard, and MHRA are also supported — see the citation-formatting skill for full style guidance).
• Include a full bibliography at the end of the review with viewer links for all corpus sources.
• Distinguish verified corpus sources (retrieved via MCP tools, with viewer links) from any manually supplied sources (labelled [unverified]).

Quality Standards

• No facts without a source (author, title) and page reference.
• Distinguish primary claims from synthesis and interpretation.
• Flag conflicting evidence explicitly rather than silently choosing one side.
• Note gaps in corpus coverage honestly — do not overstate what the sources say.
• Prefer direct quotes with page references for key claims.
• If the corpus has insufficient coverage for a meaningful review, say so and describe what is available rather than producing a thin review that overstates the evidence.

Additional Resources

• references/search-strategy.md — query construction patterns, decision points for too few or too many results, and guidance on iterative search refinement.
• references/synthesis-guide.md — handling conflicting sources, hedging language patterns, evidence weighting, and thematic organisation.
• citation-formatting skill — full citation style templates and edge cases for all supported formats.