The role of memory in correcting misinformation

Abstract

People learn about the world by continuously obtaining new information. When this information is inaccurate or purposefully deceptive, it can have devastating consequences on individuals, groups and societies. Today, most people encounter an abundance of information online, some of which may be factually incorrect.

Chapter 2 investigates properties that may influence belief in false headlines accompanied by AI-generated images and examines the effectiveness of corrections. Results show that images that are realistic and provide strong evidence to a headline increased initial beliefs in headlines. However, the memorability of corrections and their effectiveness at reducing headline beliefs were not associated with image properties.

Chapter 3 compares corrections that include AI-generated images to those that do not, and found that re-exposure to images during corrections improves memory and belief reduction relative to corrections that do not re-expose people to images. Corrections to headlines accompanied by probative images were more memorable than corrections to headlines paired with non-probative images, and improved memory for corrections was associated with decreased belief in false headlines in all conditions.

While Chapters 2 and 3 explored the relationship between memory and beliefs for AI-generated visual misinformation, Chapters 4 and 5 examined the potential neural mechanisms behind corrections to causal misinformation.

In Chapter 4, a novel misinformation correction paradigm was designed to facilitate electroencephalography (EEG). Neural activity was recorded to identify mechanisms underlying successful and unsuccessful corrections to misinformation. Results suggested that corrections were more difficult to process than affirmations, but there were no differences in recorded neural activity between subsequently remembered and forgotten corrections. An investigation of variables linked to delayed memory for corrections revealed that initially strong memories for corrections were associated with improved subsequent memory, while was no evidence that increasing misinformation fluency influenced memory for corrections.

Chapter 5 examines how mnemonic processes are influenced by providing an alternative explanation to misinformation and measuring electrophysiological activity when participants retrieved corrections. When corrections were accompanied by alternative explanations, neural activity associated with memory recollection was enhanced, suggesting that alterative explanations effectively reduce belief in misinformation by improving recollective processing.

Taken together, this thesis investigates the relationship between memory for corrections and misinformation belief in the context of visual and causal misinformation. Our findings indicate that while realistic AI-generated images have the potential to mislead, re-exposure to these images during corrections can increase belief reduction and correction memorability. EEG recordings show that while corrections may be more difficult to process than affirmations, providing an alternative explanation in addition to corrections can reduce reliance on misinformation by enhancing strategic monitoring processes.