Question 1

How does LivingMeta keep up with new literature?

Accepted Answer

Continuous monitoring of OpenAlex (350M+ academic works), arXiv preprints, and twenty-nine thesis repositories. The instance owner triggers deploys after each collection cycle. Between deploys, the Lab is the live layer — extractions performed via the website appear immediately, gap data and resources refresh on the next deploy.

Question 2

Does the AI really help with the 200K-hit screening pile?

Accepted Answer

Yes. Papers with an abstract are scored on relevance, and a topic filter built from the instance's own classifications removes clearly off-topic papers up front. What's left for human review sits in the relevance 4-10 band, where judgement matters. A typical first-deploy filters more than 90% of records as not worth surfacing — the rest is what an attentive researcher should actually read.

Question 3

How do I know the AI isn't making things up?

Accepted Answer

Every extracted fact must include a verbatim quote from the source paper; claims without textual evidence are flagged. Six AI perspectives extract the same paper independently and a consensus is computed — disagreements between perspectives are visible on every card. The Lab agent is constrained to the corpus and cites every claim by paper identifier and quote. There is no external web search in the Lab; the agent cannot reach for a generic-internet answer.

Question 4

How does the multi-perspective extraction actually work?

Accepted Answer

Six AI perspectives read each top-relevance paper independently: a balanced reviewer, a methodological rigorist, a synthesizer, a skeptic, a resource scout, and a quantitative meta-analyst. Each extracts the same set of 75 fields but with a different cognitive emphasis. A consensus resolver computes field-by-field agreement. Where they disagree, the disagreement is shown.

Question 5

What happens when the perspectives disagree?

Accepted Answer

The agreement score is surfaced on the paper card, and the raw per-perspective extractions are visible behind the consensus. Low-agreement papers are flagged as needing human attention. The human reviewer is the final word; the AI never overrules.

Question 6

Who has the final word?

Accepted Answer

The instance owner and the people using their instance. AI decisions are decisions in the sense that they are visible and revisable; they are never authoritative. The pipeline is designed so a researcher can read the AI's work, see exactly what it did, and override anything that looks wrong.

Question 7

Why should reviewers trust a paper that mentions AI use?

Accepted Answer

Because the audit trail makes the AI's contribution legible. Janssen, Van der Kleij, and De Keijzer (Sociale Vraagstukken, 2026) argue that researchers' aversion to disclosing AI use is itself the problem: when AI use is treated as something to confess, transparency is discouraged. LivingMeta's stance is the opposite — every AI-mediated step is logged, queryable, and quotable in your own methods section.

Question 8

Can I use this for my MA thesis or PhD dissertation?

Accepted Answer

Yes. The Lab includes a Junior Researcher role with Socratic scaffolding across eight research phases (Explore, Understand, Evaluate, Synthesize, Design, Execute, Write, Reflect). Supervisors get a Coach role for course design and meta-level analysis. Most launchee instances are used in exactly this configuration — one anchor researcher, a handful of doctoral students.

Question 9

How is this different from Elicit, Consensus, or Cochrane?

Accepted Answer

Elicit does per-paper extraction through conversation but without aggregation across the corpus. Consensus searches abstracts and shows citation-backed summaries. Cochrane produces manual systematic reviews by expert groups. LivingMeta integrates structured extraction, multi-perspective validation, automated gap analysis, resource curation, and AI-coached investigation into one continuously updated workflow per field. The platform is the methodology — it defines a way of doing synthesis, not just one calculation step.

Question 10

Can I bring my own corpus?

Accepted Answer

If your field has a domain-specific corpus — a data archive, a public records body, a proprietary collection — we can plug it in alongside the literature via the BYOD pattern. The Lab agent will cite both sources with the same discipline. Two instances run in this configuration today: one with Dutch local-government documents, one with case law. Talk to us if this matters for your field.

Question 11

Is LivingMeta only for Dutch academia?

Accepted Answer

No. Anchor researchers are currently Dutch but the corpora are international — papers in any major academic language are collected. The website is in English. Lab threads can be authored in any language; the agent responds in the language of the question.

Question 12

Who owns the instance data?

Accepted Answer

The instance owner. The corpus is built from open academic sources; the extractions, gap analyses, and Lab threads belong to the instance. We host and maintain, but we do not claim ownership and we do not train models on instance data.

Question 13

What happens if LivingMeta disappears tomorrow?

Accepted Answer

Every public-tier instance is published as static JSON and Next.js files. The complete dataset is downloadable. An instance can be forked and re-hosted on any CDN. Private-tier instances follow the same pattern but behind authentication; the owner can request a full export at any time.

Question 14

Is the code open?

Accepted Answer

The pipeline scripts and instance schema are publicly documented in the project repository. The Lab agent runs on our infrastructure for operational and rate-limiting reasons; the data it serves is open.

Question 15

What does the Relevance (Rel) score mean?

Accepted Answer

Every paper is scored 0–10 by an AI classifier for how central it is to the instance's field — higher is more on-topic. The score is field-specific, so a paper can be central in one instance and irrelevant in another. Papers below 4 are treated as off-topic and left out of the browsable corpus; papers scoring 6 and up get abstract-level extraction, and the most central (9–10) get the full six-perspective read wherever the full text is available.

Question 16

What are the evidence models (Empirical, Theoretical, Design, Interpretive, Computational)?

Accepted Answer

Not every paper produces knowledge the same way, so each is tagged with the kind of evidence it offers: Empirical (experiments, surveys, trials), Theoretical (proofs, formal derivations), Design (prototypes, artifacts), Interpretive (textual or qualitative analysis), or Computational (simulations, ML models). Extraction adapts to the model — an empirical paper is read for sample and effect size, a design paper for its artifact and evaluation — so the fields captured are the ones that matter for that kind of work, and the quality score is tuned per model.

Question 17

How do I find the right papers?

Accepted Answer

The Papers view searches titles, authors, topics, and AI summaries, and filters by domain, methodology, theme, evidence model, publication year, and full-text availability. Filters combine, so you can narrow to, say, experimental studies in one theme that report an extractable effect size. Every result shows its relevance score, evidence model, and citation impact — and, where extracted, its methodology and key findings inline.

Question 18

What does the full-text / PDF badge mean?

Accepted Answer

It marks papers whose full text was available (as an open-access PDF or XML) and could be read in depth rather than only from the abstract. Those papers get the richest six-perspective extraction; abstract-only papers get a lighter single pass. The badge doubles as a filter for when depth matters.

Question 19

Can I export the papers?

Accepted Answer

Yes — the current, filtered set exports to CSV for a spreadsheet or BibTeX for a reference manager, so an instance becomes a place to screen and shortlist before taking the results into your own workflow. (On the open public demo this action is disabled; it is part of every full instance.)

Question 20

What is the LMQS quality score?

Accepted Answer

LMQS — the LivingMeta Quality Score — is a per-paper measure of methodological transparency, shown as a fraction such as 3/4. It is adaptive: for each evidence model it checks the signals that actually matter (for an empirical study, sample transparency and statistical rigor; for a computational one, whether code and data were shared and results reproduced). Green means most signals are present, amber some, red few. It captures how transparently the work was reported — deliberately separate from how often it has been cited.

Question 21

What is FWCI?

Accepted Answer

Field-Weighted Citation Impact, from OpenAlex, normalizes a paper's citation count for its field, year, and document type. 1.0 is the world average for comparable papers; above is more-cited than average, below is less. It is a signal of reach and attention, not of quality or correctness — read it alongside the LMQS transparency score, not instead of it.

Question 22

What does the Meta-Analysis view show?

Accepted Answer

For every paper that reports a numeric effect size, the value is extracted and converted to a common metric (Pearson's r) where the statistics allow. Where a single paradigm has enough comparable studies, they are pooled into a random-effects estimate shown as a forest plot with a heterogeneity measure. Effects are only ever pooled within one paradigm — never across different phenomena — and papers without an extractable effect are listed transparently rather than dropped.

Question 23

Are those proper meta-analyses?

Accepted Answer

Treat them as exploratory. They are pools of AI-extracted effect sizes from the corpus, not systematic reviews with a pre-registered search and formal inclusion criteria. They are excellent for hypothesis generation and for seeing where the evidence clusters — but for a confirmatory meta-analysis you would verify the extracted effects against the source papers and apply your own protocol.

Question 24

How is the Priority Research Agenda generated?

Accepted Answer

Automatically, from the instance's own evidence. The 'further research is needed' statements across all extracted papers are clustered into themes and scored on frequency, source quality, feasibility, recency, and how much of the field they touch. An AI then filters the top themes against the field's context, so only genuinely relevant priorities survive. The list updates as new papers arrive, and each priority links straight to an investigation in The Lab — a defensible, evidence-based answer to 'where should we be working next?'.

Question 25

What counts as a 'research gap'?

Accepted Answer

A specific place where the evidence is missing, thin, or contradictory. Gaps are classified with an established taxonomy (Miles 2017) into eight types — evidence, knowledge, practice, methodological, empirical, theoretical, population, and integration — so a gap in long-term outcomes reads differently from a gap in translating findings to practice. They are drawn from what the papers themselves flag as unresolved, then aggregated to show which are field-wide versus one-off.

Question 26

What makes it a 'living' review, and what does Investigate do?

Accepted Answer

New papers are pulled from OpenAlex automatically every night, then classified and, where relevant, extracted; the site refreshes when the instance is next deployed. So the classifications, gaps, and priorities keep tracking the field as it publishes, instead of freezing on a publication date. From any priority, an Investigate button opens a Lab thread pre-loaded with that gap, so you go straight from 'here is an important gap' to working it through with the agent.

Question 27

What are Curated Resources?

Accepted Answer

Datasets, validated questionnaires, measurement instruments, tools, code repositories, and APIs discovered across the extracted papers, then curated and deduplicated into one registry. Each links back to the papers that use it — and, for instruments, to what it measures — so you can weigh the evidence base before adopting it. Link health is checked, and the registry grows as papers are added.

Question 28

What exactly is the Lab agent?

Accepted Answer

A server-side AI research agent you work with in plain-language chat. Ask a question and it searches the instance's papers, extractions, resources, and gaps to answer, pulling in the right specialist automatically — evidence mapping, gap hunting, skeptical review, resource scouting, or academic writing. There is nothing to install and no API key to bring; every claim it makes is tied to a paper by identifier, with the source quote.

Question 29

Does the Lab agent need my API key, and does it cost me?

Accepted Answer

No key, and no per-question charge to the user. The operator covers the model usage, and where a balance applies it is shown in the account as a token allowance so people know where they stand; if it runs low the agent says so plainly rather than failing silently. A credit wallet and rate limits keep costs predictable per instance.

Question 30

Can the coaching match my level?

Accepted Answer

Yes — each Lab thread is set to a role (layperson, junior researcher, senior researcher, or supervisor/coach), and the agent adjusts how much it scaffolds and across which of the eight research phases (Explore, Understand, Evaluate, Synthesize, Design, Execute, Write, Reflect). A layperson gets plain-language explanation; a junior gets Socratic, step-by-step guidance; a senior gets fast peer-level discourse; a coach gets meta-level help and teaching materials.

Question 31

Can I bring my own documents into a thread?

Accepted Answer

Yes. Attach PDFs, Word documents, or text and Markdown files, and the agent reads their content — a dataset description, a questionnaire, notes, or a draft — and works with it alongside the corpus, so the conversation is grounded in your own material as well as the literature. For a whole domain corpus rather than a few files, see the bring-your-own-data option above.

Question 32

Can it help me write the study up?

Accepted Answer

There is a dedicated writing mode: the agent co-writes against a target journal's guidelines, keeps every claim tied to a cited source, and deliberately avoids the tell-tale 'AI polish' reviewers notice. It is built to help you draft and structure honestly, with the evidence attached — not to produce a finished paper to submit unread.

Question 33

How do I get an instance for my own field?

Accepted Answer

The fastest way is a free trial: describe your field and a couple of key papers, and the platform builds a small private instance — a few hundred papers with extractions, resources, a priority agenda, and a Lab agent — live for a week, with only a small card check to keep bots out. From there it can grow into a full, maintained review of your field. Use the Try free link in the navigation, or the request form on the home page.

Question 34

Can AI agents or other tools use an instance directly?

Accepted Answer

Yes. Every instance publishes machine-readable instructions (an agent.json describing its endpoints, taxonomy, and how to contribute) and follows the llms.txt convention, and there is an MCP server for querying the corpus, extractions, gaps, and resources from your own tools. The same discipline applies to agents: cite papers by identifier, and never present outside work as evidence.

Frequently asked questions

My workflow keeps stalling

How do I trust the output?

Who is this for?

Ownership and open data

What's inside — the papers

Quality signals and meta-analysis

Gaps, the agenda, and resources

The Lab agent, in detail

Getting your own