Which linguistic cues make people fall for fake news? A comparison of cognitive and affective processing

Bernhard Lutz, Marc T. P. Adam, Stefan Feuerriegel, Nicolas Pröllochs, Dirk Neumann

Introduction

Despite the pervasive impact of misinformation on modern society, the underlying mechanisms that make individuals susceptible to fake news remain insufficiently understood. Traditional research has largely relied on self-reported belief, a method that often fails to capture the rapid, unconscious processes occurring during the act of reading. This study addresses this gap by investigating how specific linguistic cues—ranging from grammatical structures to emotional valence—trigger distinct types of information processing: cognitive and affective.

Cognitive processing refers to the deliberate, "careful thinking" required to parse complex information, while affective processing involves the "automatic evaluation" or gut reactions that occur beneath the level of conscious awareness. By distinguishing between these two channels, the researchers aim to uncover why certain writing styles are more persuasive or deceptive than others.

Methodology

The researchers conducted a within-subject laboratory experiment involving 42 participants, resulting in 40 usable datasets and 1,598 individual observations. Subjects were presented with a series of 40 news articles (20 real and 20 fake) sourced from Politifact. Veracity labels were cross-validated against Snopes and FactCheck.org to ensure a "ground truth" with a 1.0 Fleiss kappa agreement.

To capture real-time processing, the team employed two primary neurophysiological measurements:

  1. Eye Fixations: Serving as a proxy for cognitive effort, where a higher number of fixations indicates more intensive "careful thinking."
  2. Heart Rate Variability (HRV): Specifically the SDNN metric, used to measure affective processing. A decrease in HRV indicates physiological stress or discomfort, characteristic of an autonomic "fight or flight" response.
Experimental workflow illustrating the transition from news sampling and linguistic cue extraction to neurophysiological measurement and regression analysis.

The text of the articles was analyzed using the Linguistic Inquiry and Word Count (LIWC) tool, which categorizes words into 33 dimensions. Mixed-effects regression models were then utilized to determine how these linguistic dimensions predicted changes in eye fixations and HRV.

Linguistic Dimensions

The study examined a wide array of linguistic cues to identify which features most strongly influenced reader response. These cues include structural elements like length and complexity, as well as psychological categories such as affective processes and analytic thinking.

Definitions and examples of the LIWC linguistic dimensions used to categorize the news article content.

Key Findings: Cognitive vs. Affective Dissociation

The results reveal a complex interplay between writing style and physiological response, often showing that cognitive and affective systems do not always align.

The "Analytic" Red Flag

A significant finding was that fake news utilizing high "analytic" language (e.g., words like think, know, because) triggered measurable affective discomfort, evidenced by a 2.64 millisecond drop in HRV for every standard deviation increase in analytic words. This suggests that when misinformation attempts to mimic a logical or scholarly tone, it may create a "gut feeling" of psychological dissonance in the reader.

Cognitive Engagement and Skimming

On average, participants performed 7.967 fewer eye fixations on fake news compared to real news, indicating a general tendency to "skim" or disengage more quickly from misinformation. However, this engagement varied based on specific cues:

  • Length: Longer fake news articles increased cognitive processing (more eye fixations), suggesting that "elaborate" lies may successfully force deeper mental engagement.
  • The Complexity Paradox: While higher complexity increased cognitive effort for real news, the opposite was true for fake news. Simple, "low-complexity" misinformation appeared to bypass mental filters, requiring less effort to process and potentially evading critical scrutiny.

Discussion and Implications

The study highlights a dissociation between how our brains "think" about news and how our bodies "feel" about it. A linguistic cue might trigger a physiological stress response without necessarily prompting the reader to think more critically.

These findings have significant implications for digital platforms. Social media algorithms could potentially be adjusted to prioritize content that encourages cognitive processing or to flag articles that exhibit high-analytic styles but lack veracity.

Limitations and Future Work

The researchers note that the laboratory setting may not perfectly mirror real-world social media consumption, where users are often in a "hedonic" mindset. Furthermore, the participant pool was limited to college students (mean age 26). Future research should explore broader demographics and "wild" environments using wearable technology to measure HRV in real-time social media contexts.

Summary updated

BibTeX
@article{lutz2024which,
  author = {Bernhard Lutz and Marc T. P. Adam and Stefan Feuerriegel and Nicolas Pröllochs and Dirk Neumann},
  title = {{Which linguistic cues make people fall for fake news? A comparison of cognitive and affective processing}},
  journal = {Proc. ACM Hum. Comput. Interact.},
  year = {2024},
  doi = {10.1145/3641030},
}
Figure