Example 1

Headline from National Review, The Deeply Flawed Racism Index, originally labelled negative with an overall sentiment score of -1.61
Text classifier score: Neutral 51%, Positive 29%, Negative 19%

Lack of context leads our text classifier to label it mostly neutral. This makes sense. The headline does not actually attempt to elicit an emotional reaction, even though it does seem to withhold important information. Sometimes withholding information is better than adding information without context. A lack of context makes us unable to say whether negative implications were technically inferred. It also means intentionality is impossible to derive objectively. Our model errs on the side of caution. It only analyzes truly volatile news headlines material within its margin of error.

Example 2

Headline from Times of India, Over 60 dead, dozens missing as severe floods strike Europe, originally labelled negative with an overall sentiment score of -1.5
Text classifier score: Negative 100%

This headline is a negative anchor case. Our text classifier was able to differeniate between the previous headline which had a higher original sentiment score with standard sentiment analysis. From a visual inspection, this headline is also clickbait material. Even if the article contained information about floods everywhere in Europe, deaths from floods would only represent a fraction of European localities. If you wanted to see if these floods affected your specific locality, you would feel an urge to click on it because deaths and a lack of specificity heighten panic, fear, and anxiety. Volatility is a measure of subjective intentionality from objective statements of intention. Our negativity indicators are actually better at determining clickbait than training a model on low quality news-related material to find clickbait.