Source-First Information Practice / 源头优先的信息实践

Definition

Source-first information practice is the deliberate movement from recommendation feeds and repeated summaries toward the closest available evidence, combined with checks on provenance, incentives, omissions, and the decision for which the information will be used.

Current Synthesis

Source proximity reduces some distortion but does not guarantee truth. Papers, official white papers, financial filings, repositories, datasets, and direct statements expose more of the original claim, method, or accountable speaker than a chain of reposts. They can still contain selective disclosure, weak methods, strategic framing, technical barriers, or incentives that require interpretation.

AI lowers the cost of finding, translating, comparing, and summarizing source material, but it also makes plausible secondary synthesis abundant. A robust workflow therefore begins with purpose: name the decision or question, choose evidence appropriate to it, inspect how and why each source was produced, compare independent materials, and preserve a path back from the summary to the underlying evidence.

Key Claims

• Recommendation feeds optimize delivery and engagement, not necessarily evidentiary quality or user purpose.
• Moving closer to original evidence reduces retelling distance but does not remove bias, error, or strategic omission.
• Information collection should be driven by a stated question or decision rather than accumulation for its own sake.
• AI is useful for discovery, translation, extraction, comparison, and falsification, but provenance must remain inspectable.
• Author position, production context, likely audience, and avoided topics are part of interpreting a source.
• Independent comparison matters when source claims are high-consequence, contested, or self-interested.

Evidence

• Source hierarchy: EP391-当AI完成所有工作，一个人应当如何证明自己的价值？ contrasts papers, official reports, filings, repositories, and direct statements with media processing and fragmented reposting.
• Purpose before extraction: EP391-当AI完成所有工作，一个人应当如何证明自己的价值？ argues that pricing, product updates, demand discovery, and investment decisions require different collection and analysis questions.
• Context and incentives: EP391-当AI完成所有工作，一个人应当如何证明自己的价值？ asks users to examine why material was produced, the author’s position, and what may have been left unsaid.

Counterevidence & Qualifications

Primary or official does not mean neutral, complete, accessible, or methodologically strong. Secondary scholarship and specialist journalism can add comparison, expertise, leaked evidence, historical context, and error correction unavailable in a single original source. Some questions have no direct source, and source access can be constrained by language, paywalls, secrecy, safety, or technical complexity. The practice is therefore source-proximate and triangulating, not reflexively anti-secondary.

What Changed

• Created a purpose-driven workflow connecting source proximity with provenance and incentive checks.
• Added the qualification that primary and official materials still require triangulation and interpretation.

Related Concepts

• Deep Research - expands a question through multi-source investigation and synthesis.
• AI Answer Source Attribution - preserves the path from generated claims to supporting material.
• Research Report Reading Discipline - applies incentive and evidence checks to analyst reports.
• Information Cocoon / 信息茧房 - describes the narrowing risk of passive recommendation feeds.
• Problem Definition In Research - determines which evidence is relevant before collection begins.
• Human Judgment Under AI - retains responsibility for deciding whether retrieved evidence is sufficient.