How Woke Movie Scores Are Calculated

The pipeline combines TMDB metadata with cultural research produced by Gemini using Google Search grounding. The research pass looks for reporting and public discussion about representation, social themes, cultural impact, and related controversies. When grounded research finds little relevant material, the next stage can work from the film metadata instead of treating missing evidence as fact.

Web research is a snapshot of what Google could retrieve when the film was scored. It can be incomplete, and pages or search results may change later. The source URLs and the research summary are not retained in published score records. Published records retain scoring timestamps, but a timestamp does not mean that the underlying claims have been continuously checked since that date.

A language model uses the research and movie metadata to produce 10 marker assessments, including a score and short evidence description for every marker. No-evidence markers are instructed to receive zero. A separate model pass then synthesizes those marker assessments into the published Woke Score from 0 to 100.

That holistic pass is constrained by the rubric below and calibration examples. It is still a generative model judgment, not deterministic arithmetic that will always reproduce the same output from a hand calculation.