WEBVTT

1
00:00:00.370 --> 00:00:03.320
Step 1 of 9. Topic labels: a keyword dictionary and a language model label the same gold-labelled excerpts.

2
00:00:03.320 --> 00:00:06.970
Step 2 of 9. The keyword baseline runs for everyone: agreement and Cohen's kappa with 95% intervals.

3
00:00:06.970 --> 00:00:11.770
Step 3 of 9. The model is optional and runs on the visitor's own key. Open AI settings.

4
00:00:11.770 --> 00:00:18.450
Step 4 of 9. Paste a key (here a placeholder, not a real key). It stays in this browser tab.

5
00:00:18.450 --> 00:00:22.800
Step 5 of 9. Save. Requests go straight from the browser to the provider, never to this site.

6
00:00:22.800 --> 00:00:31.230
Step 6 of 9. A seeded sample of 40 excerpts, sent in four requests. Mocked AI response for illustration.

7
00:00:31.230 --> 00:00:36.480
Step 7 of 9. Both labellers scored on the same excerpts: intervals, paired difference and McNemar's test.

8
00:00:36.480 --> 00:00:42.780
Step 8 of 9. Every label is marked AI-generated. Correct one, then accept the run.

9
00:00:42.780 --> 00:00:53.600
Step 9 of 9. Every call is in the AI audit log with prompt, answer, latency, tokens and decision.