Why do users prefer community sources over encyclopedic references?
This reads the question as: why do people lean on lived, crowd-sourced material (reviews, forums, community discussion) over polished authoritative reference works — and the corpus doesn't tackle this head-on, but several notes on abstraction, comparison, and collective signal point at the same mechanism.
This explores why community sources feel more useful than encyclopedic ones — and while the collection has no paper aimed squarely at this, a few threads converge on a sharp answer: encyclopedic references systematically lose the specific, comparative, and lived detail that people actually use to decide things.
Start with the abstraction problem. Work on word frequency shows that general concepts (hypernyms) appear far more often than specific ones (hyponyms), and because language models and consensus writing both drift toward the common, they slide toward abstraction — erasing the expert-level specificity that makes an answer actionable Does word frequency correlate with semantic abstraction?. An encyclopedia entry is the endpoint of that drift: accurate, general, and stripped of the granular case detail a community thread keeps alive. The reader who wants 'will this work for *my* setup' is asking exactly the question abstraction can't answer.
Second, humans evaluate by comparison, not in isolation. Relational explanations that reference one item against another carry more decision-relevant information than standalone descriptions, because they match how people naturally size things up Do comparisons help users evaluate items better than isolated descriptions?. Community sources are saturated with exactly this — 'X vs Y, here's when each breaks' — whereas reference works describe each thing on its own terms. The same insight shows up in explainable recommendation: pulling aspect-level signal from reviews gives richer, more personalized grounding than generic descriptions, especially when you don't fit the average case Can retrieval enhancement fix explainable recommendations for sparse users?.
There's also a collective-knowledge angle the corpus surfaces from an unexpected place: aggregating many users' behavior exposes relationships that no single authoritative record contains Can cross-user behavior reveal news relations that individual histories miss?. Community sources are that aggregation in human form — thousands of partial, situated reports that together map territory an editorial reference never charts.
The twist worth taking away: preference here may be partly a trust *heuristic* rather than a quality judgment. People rate responses higher when they simply carry more citations — even irrelevant ones — because volume reads as credibility Do users trust citations more when there are simply more of them?. A community thread's many voices may trigger the same signal an encyclopedia's single authoritative voice doesn't. So the preference is real, but it's driven both by genuine information value (specificity, comparison, collective coverage) and by a quantity-equals-trust shortcut that's easy to game.
Sources 5 notes
WordNet analysis shows hypernyms (general concepts) occur more frequently than hyponyms (specific ones). Combined with LLMs' frequency bias, this means preferring common paraphrases systematically drifts toward abstraction, erasing expert-level specificity.
Relational explanations that compare items carry more decision-relevant information than isolated evaluations because they match how humans naturally assess products. A system extracting aspects from reviews and generating aspect-controlled comparisons produces sentences rated as both accurate and useful for purchase decisions.
ERRA combines model-agnostic review retrieval with personalized aspect selection to address data sparsity that embedded methods cannot solve. Retrieval augmentation provides richer signal when user history is sparse, while aspect personalization ensures explanations match user context rather than generic defaults.
GLORY constructs a global news graph from aggregated user clicks to discover article relationships invisible in any single user's sparse history. This population-level behavioral structure enables recommendations even when direct textual or per-user similarity fails.
Analysis of 24,000 Search Arena interactions shows irrelevant citations boost user preference (β=0.273) nearly as much as relevant citations (β=0.285), indicating citation count functions as a decoupled trust heuristic.