Media attention shapes perceived social influence and, through that channel, individuals' intentions to ride autonomous vehicles; conversely, early intention to ride predicts later increases in perceived trustworthiness, creating potential feedback loops in AV adoption.

Main Finding

Cross-lagged analyses of two-wave panel data (wave 1 n = 1,306; wave 2 n = 650) provide causal evidence that (1) viewers’ attention to media content about AV risks/benefits influences their presumptions of how that content affects others (projection effect and impersonal impact), and (2) there is a consonance effect in which intention to ride an AV predicts later increases in perceived trustworthiness of AVs.

Key Points

• The study adapts the Presumed Media Influence (PMI) model to autonomous vehicles (AVs).
• Projection effect: viewers’ own attention to AV-related media predicts their assumptions about media influence on others.
• Impersonal impact: those presumed influences on others feed back to shape the viewer’s own attitudes and intentions regarding AVs.
• Consonance effect (reverse/feedback path): intention to ride an AV at wave 1 predicts higher perceived trustworthiness of AVs at wave 2 — i.e., behavioral intention can shape subsequent beliefs about technology trustworthiness.
• Analyses support directional (causal) paths rather than only cross-sectional associations.
• Attrition: sample reduced from 1,306 at wave 1 to 650 at wave 2.

Data & Methods

• Design: Two-wave panel survey with cross-lagged panel analysis to test directional relationships among attention to media content (risks/benefits), presumed influence on others, perceived trustworthiness of AVs, and intention to ride.
• Sample sizes: wave 1 n = 1,306; wave 2 n = 650 (panel subsample).
• Analytical approach: cross-lagged models that estimate lagged effects between constructs across waves to infer causal ordering (controls and model specifics not detailed here).
• Measures (as described): self-reported media attention to AV risk/benefit content, presumed media influence on others, perceived trustworthiness of AVs, and intention to ride AVs.
• Limitations inherent to two-wave observational panels: sample attrition, reliance on self-report, and limited temporal span — though cross-lagged models strengthen causal interpretation versus cross-sectional analyses.

Implications for AI Economics

• Demand formation and diffusion: media narratives shape not only individual beliefs but also presumed social influence, which then feeds back into adoption intentions — models of AV demand should incorporate mediated social perceptions and indirect media-driven externalities.
• Strategic communication and policy: regulators, firms, and advocates can influence uptake by shaping media content about risks/benefits; managing perceived societal reactions may be as important as changing individuals’ direct beliefs.
• Feedback effects in adoption models: the consonance effect implies intention can alter perceived trustworthiness, creating positive feedback loops (self-reinforcing adoption) or negative loops if early intentions are low — important for forecasting uptake trajectories and tipping points.
• Pricing, insurance, and investment decisions: expectations of adoption speed and public trust (partly media-driven) affect market size projections, network externalities, and risk pricing for insurers and investors.
• Targeting interventions: influencing presumed media impact on particular social reference groups (not only individuals) could be a cost-effective lever to raise adoption; economic interventions should consider heterogeneous social influence channels.
• Research needs: incorporate media-driven social influence into structural adoption models, use experimental/quasi-experimental designs to estimate causal effect sizes for policy simulation, and link stated intentions to revealed behavior for welfare and market predictions.