Discover and Cross-Check Web Sources

Today's Goals

D1 gave Maya a question map for Jo's fictional Harborfield night-bus decision. Today you will turn search results into an evidence register. You will distinguish a result, a candidate source, a reviewed source, and an accepted claim. You will record who produced each source, the period it covers, the exact passage you rely on, and what it cannot establish. You will also test whether two apparently different pages are independent evidence or two copies of the same origin. The day's deliverable is a register and one documented cross-check, not a list of links.

The Harborfield excerpts in the lab are wholly invented. Their source IDs are stable across the course: HF-01 is a transit count, HF-02 listening notes, HF-03 an operations estimate, and HF-04 a news item from another town. Treat them as offline practice originals. A live Perplexity search is optional, must use a separate real-world practice question, and must never be described as having generated these cards.

Plain-Language Walkthrough

A result is a lead, not an admitted source

Think of a search result as a name on a library catalog card. It tells you where to look, not whether the book answers your exact question. Perplexity can present a fluent synthesis with linked citations. That format is useful because you can follow a claim outward. It is also dangerous if the presence of a citation is mistaken for a completed check. The linked page may say something narrower, refer to a different place, quote a supplier, or point to an inaccessible report. Your job is to open the source, locate the exact statement, and decide what, if anything, that statement supports.

Use four statuses in the register. A lead is a result worth opening. A candidate has an identifiable original and appears relevant. A reviewed source has a documented creator, date, method, scope, and passage. An accepted claim has a specific claim-to-source match with a limitation. A single page may be reviewed yet fail to support the claim you hoped for. HF-04 can be a reviewed article and still be excluded as proof of Harborfield demand. Keeping these statuses separate prevents a copied link from becoming a fact merely because it stayed in a document long enough.

An effective register holds a stable ID, title, producer, original URL or file location, publication date, observation period, accessed date, evidence type, exact supporting passage, method or sample, permitted use, excluded use, and decision reason. In this course the location is a local lab excerpt rather than a real URL. In a real project, save the original link and note if it redirects or changes. Do not assume a search snippet is equivalent to the page it summarizes. A snippet can omit a footnote, table caption, or qualification that reverses the interpretation of a number.

The two dates are easy to confuse. A page published on August 5 might describe counts collected in July. If you need to know whether a new policy has changed behavior in September, the August publication date does not make July observations current. Record both publication and coverage periods. Also record your access date because a web page can be updated or removed. D3 will use these fields to handle a later correction. Without them, “newer source” can accidentally mean “newer webpage wrapping the same old data.”

Sign in and link WeChat to read

Sign in, then link your WeChat mini program account on the account page. Linked accounts share one free article per day. Unlock additional articles with rewarded videos in the mini program.

Scan in WeChat to open Dayward AI. Open the same course and chapter, choose the rewarded video, then refresh this page after completion.

Unlocked — refresh to readSign in / scan with WeChat →