Text- and voice-based conversational agents (CAs) provide interactivity beyond the messages public health agencies traditionally use to share information and correct misperceptions. Yet, little research has compared how interactivity and modality influence misperception correction and behavior change. This study examines how modality and information quantity shape cognitive load, misperception change, and behavioral intention.
A 3 (modality: non-interactive message vs. text-based CA vs. voice-based CA) × 2 (information quantity: short vs. long) between-subjects experiment was conducted with 1,157 U.S. adults.
Both CA modalities were associated with lower post-intervention misperceptions than non-interactive messages. Relative to text-based CAs, voice delivery was linked to higher cognitive load, which in turn was associated with greater post-intervention misperceptions and lower intention to adopt healthy behavior. Nevertheless, when information quantity, cognitive load, preexisting and post-intervention misperceptions, and preexisting behavioral intention were controlled for, voice-based CAs were directly associated with the intention to adopt recommended behaviors. Information quantity did not moderate the relationship between modality and cognitive load.
To our knowledge, this study is the first controlled experiment to compare text-based CAs, voice-based CAs, and social media messages used to correct health misinformation, while clarifying the mechanisms through which they shape behavior change. The results show that neither CA modality is consistently more effective than the other across all outcomes. As such, practitioners should adopt a “fit-for-purpose” strategy, i.e. use text-based CAs when demands for comprehension are high and voice-based CAs when the main goal is to motivate behavior change.
