Methodology
Evidence before inference.
Internet Pulse is designed to help people interpret a public conversation without turning probabilistic analysis into certainty.
Define the subject and source scope.
Every analysis begins with a bounded subject, timeframe and available source configuration. Current public-conversation collection centers on YouTube and Reddit. Web and news context can be added when discovery is configured; broader platform coverage is not represented as live unless the corresponding integration is available.
Keep context separate from reaction evidence.
Context helps define the subject and expand a search responsibly. Reaction analysis remains grounded in the collected public conversation, so contextual material is not silently treated as evidence of sentiment, emotion or stance.
Build connected analytical layers.
Relevance, sentiment, emotion, stance, semantic similarity, narratives and change over time are treated as related but distinct layers. The product records the provider and method used so an analyst can understand how an output was produced.
Treat signals as prompts for review.
Abrupt shifts, repeated behavior or high semantic similarity may warrant attention. They do not prove a bot farm, coordinated campaign or intent. Internet Pulse uses cautious language and keeps the underlying evidence available for review.
Leave consequential judgment with people.
Automated analysis helps organize volume; it does not replace analyst judgment. Lower-confidence material can be routed for review, and consequential conclusions should be checked against source evidence and wider context.