Processing of veracity cues: how processing difficulty affects the memory of event description and judgment of confidence

Abstract

In today's rapid dissemination of information, discerning truth from falsehood is crucial. We investigated how cues signaling information veracity influence memory accuracy and confidence in coherent narratives. Two studies manipulated perceptual difficulty in distinguishing true-labeled from false-labeled information in event descriptions using font color as a cue. Study 1 (N = 198) revealed that the presence of veracity cues reduces memory accuracy for the event description. Study 2 (N = 248) showed that when differentiating veracity cues became more challenging, false-labeled information was more frequently misidentified and less accurately remembered. Confidence ratings decreased with the presence of veracity cues (Study 1) but resulted in disproportionately high confidence for sentences labeled as false that were confused as true (Study 2). False-labeled information was less retained, yielding initially more accurate event representation. However, once stored, false-labeled information was recalled with confidence as true, leading to a false representation. Therefore, mechanisms such as highlighting the veracity of information within coherent news articles on social media should be used with consideration.

Keywords: Confidence; Event description; Memory; Processing fluency; Truth bias; Veracity cue.

Fig. 1

Example stimuli material. The figure depicts in (A) an example of the randomized assignment of the font color as a veracity cue representing true or false in the high and low conditions. (B) shows the two introduction sentences always depicted as true, and (C) shows the first target sentence depicted as false

Fig. 2

Study Procedure for Study 1 and Study 2. The figure depicts in (A) the start of the reading task with a brief reminder showing the veracity cues in the high and low conditions, while in the control condition, an example statement is shown without any veracity cue. This is followed by two introduction sentences (B) and a first target sentence (C). The encoding phase was similar for Study 1 and Study 2. The retrieval phase (D) differed between the studies (Study 1: content memory; Study 2: veracity memory). Study 1 included the control condition (dark gray surface). Note that the colors representing true and false were randomized within low-discriminability (blue and gray) and high-discriminability (purple and gray) conditions, while in the control condition, one of the three colors (purple, gray, blue) was randomly used. See also Fig. 1 for examples regarding A, B, and C for the high- and low-discriminability conditions

Fig. 3

Percentage of Correct Recognition of Event Sentences as a Function of Discriminability and Veracity Cue. In the discriminability condition "control," participants did not see a veracity cue. Black dots represent the mean. Whiskers represent the interquartile range of 95%. N = 198

Fig. 4

Calibration Plot for Confidence and Correct Recognition separated by Discriminability for Study 1. Confidence was collected on a 6-point Likert scale with answer options "50% (= I guessed.)" to "100% (= I am certain.)". The blue line represents the optimal calibration of confidence and correct content recognition of sentences labeled as true or false. In the control condition, the blue line represents optimal calibration and correct recognition of all sentences, as no veracity cue was provided. Optimal calibration would exist when high confidence predicts high accuracy ("100% (= I am certain.)"), while low confidence predicts low accuracy ("50% (= I guessed.)")

Fig. 5

Percentage of given responses (classified as correct, confused, and misidentified) for Study 2: source memory as a function of discriminability and veracity cue. The plots (A), (B), and (C) represent the descriptive data for the responses given by the participants. Black dots represent the mean. N = 248

Fig. 6

Calibration plot for confidence and responses (classified as correct, confused, and misidentified) separated by discriminability and veracity for Study 2. Confidence was collected on a scale from "50% (= I guessed.)" to "100% (= I am certain.)" The blue line represents the optimal calibration of confidence and response classification of true-labeled and false-labeled sentences. For the responses classified as confused or misidentified, the slope is negative as with higher confidence, fewer errors should be made