浏览代码

Add research workflow guide

yila-research-145f 1 月之前
父节点
当前提交
322dc27a6a
共有 1 个文件被更改,包括 63 次插入 和 2 次删除
  1. 63 2
      README.md

+ 63 - 2
README.md

@@ -1,3 +1,64 @@
-# yila-source-credibility
+# Judging Source Credibility When Searching Beyond the Big Databases
 
-Judging source credibility beyond the big databases - notes from Yila https://yila.ai
+Researchers increasingly search beyond curated databases such as Web of Science, Scopus or PubMed. Google Scholar, preprint servers, institutional repositories, conference proceedings, policy reports and the open web all contain relevant work, and in many fields some of the most current evidence appears there first. The cost of this breadth is that quality control shifts from the database to you. Predatory journals, retracted papers, unreviewed preprints and opinion pieces dressed as research all turn up in the same results list. This guide sets out a practical way to judge source credibility quickly, which signals are reliable and which are not, how to record your judgements, and how a source-linked assistant such as [Yila](https://yila.ai) fits into a workflow that keeps those judgements visible.
+
+## Credibility is about the claim, not just the venue
+
+It is tempting to judge a source entirely by where it appears: a well-known journal is trustworthy, a blog is not. Venue matters, but it is a rough proxy. Strong journals publish weak papers, and some excellent analyses appear in working papers or technical reports. The more useful question is whether this particular source supports this particular claim. A preprint may be perfectly adequate evidence that a method exists, but weak evidence that it outperforms alternatives. A government report may be the best available source for a national statistic, but not for a causal claim.
+
+Keep this distinction in mind as you assess sources. Your notes should record not only "is this credible?" but "credible for what?"
+
+## Check the venue quickly
+
+For journal articles, a few checks catch most problems. Is the journal indexed in a major database such as Scopus, Web of Science, DOAJ for open-access journals or MEDLINE for biomedicine? Does it have a clear editorial board with verifiable members, a stated peer review process and a publisher you can identify? Do the fees, timelines and scope look reasonable, or does the journal promise acceptance within days across every discipline?
+
+Lists of predatory journals exist, but they are incomplete and sometimes contested. Treat them as one signal among several. Think. Check. Submit, a resource for authors, offers a useful checklist that works just as well for readers.
+
+For conference papers, check whether the conference is established in the field and whether it is peer reviewed. For preprints, note the server and whether a peer-reviewed version has since been published.
+
+## Check the paper itself
+
+Whatever the venue, look at the paper's internal signals. Are the authors identifiable, with affiliations and a track record you can verify? Is there a methods section detailed enough to evaluate? Are data or code available, or at least described? Do the conclusions follow from the results, or do they overreach? Are conflicts of interest and funding declared?
+
+Look for retractions and corrections. Retraction Watch maintains a database, and many publishers now display retraction notices prominently, but retracted papers still circulate through copies on other sites. If a paper seems surprisingly important, search for its title together with "retraction" or "correction".
+
+For preprints, look for subsequent discussion: comments on the server, published responses or a later peer-reviewed version. A preprint that has been public for years without being published or cited may have problems that reviewers found.
+
+## Check grey literature on its own terms
+
+Reports from governments, international organisations, think tanks and NGOs are often essential, especially in policy-relevant fields. Judge them on transparency: is the method described, are the data sources named, is the organisation clear about its mission and funding? Advocacy organisations can produce rigorous work, but you should know their position. Prefer the primary source for statistics, for example the national statistics office rather than a news article quoting it.
+
+## Use citation context, carefully
+
+Citation counts and citation context can help. A paper cited many times by other work in strong venues is at least visible in the field. Tools that show whether citations support or contradict a paper can reveal controversy. But citation counts reward age and popularity, not correctness, and highly cited papers are sometimes highly cited because they are disputed. Use citations to find discussion of a paper, not as a quality score.
+
+## Record your judgement
+
+Credibility judgements should be recorded, not just made. In your reference manager or evidence table, add a short field for source type (peer-reviewed article, preprint, report, conference paper, other), a credibility note and the claim the source supports. Something as simple as "preprint, not yet reviewed, methods adequate; use only for existence of method" is enough.
+
+This record matters for three reasons. It stops you from upgrading weak sources over time as you forget their status. It lets co-authors and supervisors see and challenge your judgements. And it makes your final text more honest, because you can match the strength of your wording to the strength of each source.
+
+## Match wording to credibility
+
+The final step is in the writing. When you cite a preprint, say so if the claim depends on it. When a finding comes from a single small study, avoid wording that implies consensus. When a statistic comes from an advocacy report, consider naming the source in the text. Readers rely on your phrasing to judge how much weight a claim can bear, and careful wording is often more effective than excluding sources altogether.
+
+## Where an assistant helps
+
+Assistants that search and summarise across many sources can make it harder, not easier, to judge credibility if they present everything in the same confident voice. The safeguard is traceability: being able to see exactly which source supports each sentence, so you can apply your judgement to it.
+
+Yila is designed to turn papers and data into cited reviews, figures and slides, keeping each statement tied to the passage it came from. When you work from a set of sources you have already assessed, its drafts carry links back to specific passages, so you can check whether a sentence rests on a peer-reviewed trial, a preprint or a report, and adjust the wording accordingly. The credibility judgement remains yours; the links make it practical to apply it to every claim.
+
+## A quick credibility checklist
+
+- Ask "credible for what claim?" rather than "credible or not?"
+- Check venue indexing, editorial transparency and peer review.
+- Check authors, methods, data availability, conflicts and conclusions.
+- Search for retractions, corrections and published versions of preprints.
+- Judge grey literature on transparency and prefer primary statistics.
+- Use citations to find discussion, not as a quality score.
+- Record source type, credibility note and supported claim.
+- Match the strength of your wording to the strength of the source.
+
+## Conclusion
+
+Searching beyond the big databases is often necessary and frequently rewarding, but it moves responsibility for quality onto the researcher. A quick, consistent set of checks, a recorded judgement for each source and wording that reflects that judgement keep breadth from becoming a liability. A source-linked workflow, whether manual or with an assistant like Yila, makes those judgements easy to apply and easy for others to see.