Not a poll
CANZUK International has released a comprehensive public opinion analysis examining attitudes toward a proposed CANZUK alliance among citizens in Canada, Australia, New Zealand, and the United Kingdom. …
New Zealand: 75% Support
In New Zealand, 75% favored participation in a CANZUK alliance, with only 18% opposed and 7% unsure.
I was interested in which polling company was used for the NZ portion, as I assumed it was a poll. But you had to download the full report to find the following:
The methodology employed digital sentiment analysis, a modern, large-scale approach that synthesises weighted favourability signals from a broad range of public discourse. This included social media comments, likes, shares, reposts, news articles, op-eds, editorials, informal online polls, and digital petition activity.
Advanced artificial intelligence tools—incorporating natural language processing (NLP) and machine learning models—were used to classify sentiment, detect contextual nuance (including sarcasm and mixed views), and weight contributions from verified geographic sources. Unlike traditional probability-based polling (e.g., random-digit-dial telephone surveys or stratified online panels), this method draws on organic, unprompted conversations rather than prompted responses.
Sentiment analysis can be very useful and interesting. I know the Trump campaign relied on it greatly. I am a fan of sentiment analysis.
However I am not a fan of presenting sentiment analysis as what most people would think is a poll result. Saying in NZ 75% favoured participation implies a scientific poll.
What they should have said is something along the lines of
“Online sentiment analysis in New Zealand found 75% of comments were favourable towards CANZUK”. You shouldn’t have to download the full report to find it it wasn’t actually a poll.