This piece warns that AI chatbots often prioritize flattery over factual accuracy, which can distort decision-making, erode trust, and have serious psychological and political consequences, especially in high-stakes contexts.
How the framings classify across 3 articles. Each framing is labelled by a small AI stance classifier; see the methodology page for details.
Stacked weekly counts; colour by lean. “n/a” covers government and iwi-Māori sources where lean isn't applicable.
How this topic has been named, week by week. A new alias winning out is usually a framing shift.
How the news corpus has covered this same topic over the last 12 weeks. 6 articles from RNZ, Stuff, NZ Herald, ODT, 1News, Newsroom and The Spinoff. Click through to the press view for the full panel.
Up to 12 framings spread across orientations. Each framing is a short phrase the topic extractor generated to characterise the piece's stance — not a quote from the source. Click through to read the original.
ai reproduces and amplifies training data flaws
Artificial intelligence’s limited ‘intelligence’reinforced by human agreeableness in feedback
AI chatbots can prioritize flattery over facts – and that carries serious risksSocial-media signal on the same topic, drawn from the social lens. Engagement is likes + 2×shares + 3×replies, the same weighting used across the digest cards. View on /social →
Spotted something wrong on this page? Report a correction.
