CORE GUIDE
Triangulate independent sources
Confidence should increase when independent evidence paths converge, not when many links repeat the same upstream claim.
Mental model
Count independent evidence paths, not citation count. Five pages copying one original source are still one underlying piece of evidence.
Why it matters
AI-assisted research can produce a long bibliography that looks rigorous while remaining epistemically thin. Syndicated articles, derivative summaries and repeated press coverage may all inherit the same unverified claim. Triangulation asks whether the evidence has genuinely independent origins.
01
Track evidence lineage
Identify primary sources and group documents that share the same upstream evidence. For consequential claims, seek a different evidence path such as an official record, independent measurement, separate dataset or expert analysis. Preserve disagreement when independent sources do not converge.
02
Example: five articles, one statistic
Five news stories say that 70% of teams adopted AI, but each article quotes the same company blog. That is one data point repeated five times. An independent survey with a different sample would add a second evidence path; another rewrite of the blog would not.
Common failure modes
- Treating syndicated coverage as independent confirmation.
- Using search-result count as a proxy for confidence.
- Smoothing over disagreement to produce a cleaner narrative.
Engineering heuristics
- Track the original source behind every consequential claim.
- Seek at least one independent evidence path for high-impact conclusions.
- Preserve conflicts and uncertainty in the synthesis.
Takeaways
- 01Citation quantity is not evidence diversity.
- 02Independence is about source lineage.
- 03A strong synthesis can explicitly say that credible evidence conflicts.
Related concepts from the Knowledge Graph
These relationships come from the canonical graph, not a separate Guide taxonomy.