The use of deceptive language: a new look at president Donald Trump

In 2022, Sophie van der Zee achieved significant results by developing a personalised language model to analyse Donald Trump’s tweets. Tweets containing misinformation were found to use more adverbs, negative emotions, and negations compared to truthful tweets. Van der Zee also launched a website where users can explore Trump’s tweets and see whether the model predicts them to be correct or incorrect based on linguistic patterns. A recent article in the largest Dutch daily morning newspaper De Telegraaf, highlighted her work as a key example of lie detection. The article discusses how deception research points to various methods but acknowledges disagreements in the literature about the ease of detection. It advocates for a more tailored approach: analysing an individual’s specific strategies.

Personalised lie detection model

Sophie van der Zee, together with co-researchers Ronald Poppe, Alice Havrileck, and Aurélien Baillon, conducted the study, published in Psychological Science, using a comprehensive database of fact-checked tweets from Donald Trump’s X account, compiled by The Washington Post. They applied linguistic analysis tools to examine a sample from February to April 2018. The findings revealed notable differences in language use between factually correct and incorrect tweets, suggesting intent. Incorrect tweets showed increased word count, more words expressing negative emotions, greater use of cognitive processes, and more frequent negations. These patterns were tested in reverse, enabling the model to predict tweet correctness with 74% accuracy.

A new unique opportunity

‘Trump’s new presidency gives us a unique opportunity to compare his deceptive language use from four years ago to now, and to evaluate its consistency,’ van der Zee comments. If the model remains accurate in identifying Trump’s language patterns, it could serve as a highly valuable assessment tool. ‘Especially now that Meta has recently discontinued its fact-checking programme, tools like these are more crucial than ever,’ she adds.