What the picker changes

Top topics digest — the cards score the selected period against the prior 4 weeks.
12-week heatmap & outlet matrix — show the 12 weeks ending at the selected week (they slide back with the picker, they aren’t a fixed snapshot).
Per-topic volume / alias drift — same 12-week trailing window, anchored on the selected period.
Coverage gap quadrant — scores the selected period against the 12 weeks before it (not including it).
Anomaly cards — only show alerts the detector fired during the selected period. Quiet weeks legitimately show none.

What stays as-is

Outlet orientation strip / lean colours — context-only, drawn from the last 12 weeks of activity regardless.
Co-occurrence graph — recent-activity anchored, not picker-driven.
Source & topic profiles — all-time data for the topic; the picker doesn’t affect them.

Rolling 7 days is a sliding live window for “current vibes”; switch to Weekly to compare specific weeks side-by-side.

Weekly Rolling 7 days

live window

Topic · social

Ai Algorithm Transparency

49 posts · 547 engagement · first seen 11 May 2026 · last seen 23 Oct 2026

Debate over whether artificial intelligence systems and digital platforms should be legally or ethically required to disclose how their algorithms make decisions, curate content, and use data, and how New Zealand policy and regulation should respond to calls for greater openness.

Stance breakdown

How the per-edge framings classify across 24 classified edges. Each framing is labelled by a small AI stance classifier; see the methodology page for details.

50%

46%

Supportive 12 Critical 11 Neutral / explainer 1

Sentiment breakdown

Whole-post emotional valence across 24 classified posts that reference this topic. Distinct from stance: a post can be supportive of the topic and still negative in tone (an angry supporter) — the two together capture the texture short-form posts compress into very few words.

13%

50%

38%

Positive 3 Neutral 12 Negative 9

Platform mix

Where the conversation is happening. Engagement is volume-weighted (likes + 2×shares + 3×replies); a small number of high-impact posts can outrank a high-volume but low-engagement platform.

youtube

24 posts · 31 engagement

31

Ai Algorithm Transparency

Stance breakdown

Sentiment breakdown

Platform mix

Top posts by engagement