Observatory
YOUPOL / ExploreThemes and arguments.
The arguments put forward or contested in the videos, grouped by theme and by period. Our local model draws them from a sample of political sentences, then they are checked against the sentences; each quote links back to its video.
Technology
Political sentences about technology, technique or innovation, or about tech companies and their leaders.
The last seven days are read again every day. * provisional, part of the period’s videos are still to be transcribed. Greyed out, reading pending.
- 15
- political sentences on the theme
- 4
- sentences read
- 72%
- of the period’s YouTube videos transcribed
- Gemma 4 31B
- read on 5 October 2026 at 14:22
- Conservative right 20.0% 1 videos
Mean share, per video, of the political sentences on the theme; each video counts once, from 20 political sentences.
- Conservative right 100% 15 sentences
This split follows the size of each orientation’s corpus. It describes the channel selection, not the weight each side gives the theme.
Arguments come from a sample shared by the territories. Only excerpts from channels in the selected country and their associated arguments are kept here; counts are recalculated.
The videos, in their own words.
The sentences the classifier ties most clearly to the theme, taken from different orientations and channels; each one comes out of its video, which opens in a side panel.
-
Ben ShapiroConservative right · 3 October 2026
« This is the logic behind the DSA's demand for public ownership of major corporations. If private profit represents value taken from workers, then transferring the ownership back to the workers corrects the injustice. »
Where the theme weighs most.
Videos of the last four months, ranked by the share of their political sentences the classifier assigns to the theme; each record opens in a side panel.
The percentage is the share of the video’s political sentences assigned to the theme (at least 20 political sentences and 5 on the theme).
The theme, year by year.
Political sentences on the theme by period; click a period to show it. The number also depends on the share of videos transcribed, which varies a lot from year to year.
The theme’s political sentences are drawn at random by period, with quotas per orientation and a cap per channel and per video. Our local model reads the sentences without knowing their channel or orientation, groups those making the same argument, separates those reporting it in order to contest it and sets aside those off the theme. Orientation shares and quotes come from the sentences, never from the model. The last seven days are read again every day.
The detail in the method