Binge's recommendation system rests on a hypothesis: the way reviewers write reveals what they're looking for in a show. To test that hypothesis, we clustered 5,514 IMDb reviews of prestige TV drama into six rhetorical styles — and then ran a separate, independent analysis on a sample of each style, using the Close Reader's playbook of reasoning, rhetoric, and audience modules. The two systems agreed.
Method. 5,514 reviews across 293 shows were embedded with BGE-base, clustered into 6 groups by writing style, then named by Claude Sonnet 4.6. Twenty reviews per cluster were independently analyzed by the Close Reader's 200-module taxonomy (Reasoning · Rhetoric · Audience). The aggregated module frequencies, shown below, were computed without knowledge of the original Binge style label. Where they line up, that's cross-system validation: two independent typologies seeing the same patterns.
Universal patterns
What every style shares.
Two reasoning camps.
The clearest divide the cross-system analysis surfaced.
Verdict voices
"I felt it, so it's true."
Reasoning leans on ampliative:inductive — generalizing from a personal viewing experience to a universal claim about the show. The verdict arrives first, the explanation follows. The writer is the warrant.
Reasoning leans on revision:defeasible — explicit position-updating as new evidence (episodes, seasons) accumulates. The verdict is provisional; the reader watches the writer think.
Casual Consumer Voice · Pragmatic Plot Assessor
Six profiles.
For each style, the top reasoning, rhetorical, and audience modules — with sample passages from real reviews.