Table 2.
Message assessment framework (MAF).
| Component | Description | Supporting theory/reference |
|---|---|---|
| Behavioral theory | Core theoretical models provide the foundation for message framing and agent decision-making. This study used EPPM and TPB, but the framework can also incorporate models such as the Health Belief Model, Social Cognitive Theory, and Elaboration Likelihood Model. | EPPM; TPB; Health Belief Model; SCT; ELM |
| Message design (theory-based messaging) | Experimental messages were framed using constructs from EPPM (fear, efficacy) and TPB (subjective norms), tested individually and in combination. A neutral control was also included. | Witte and Allen (2000), Ajzen (2020), Tannenbaum et al. (2015) |
| Agent Design (theory-based population agents) | Agents were embedded with constructs aligned with message framings: perceived threat (severity/susceptibility), efficacy (response/self-efficacy), and subjective norms. These shaped how agents processed messages and made evacuation decisions. | Morss et al. (2024), DiCarlo and Berglund (2020) |
| Belief–desire–intention (BDI) architecture | BDI architecture allowed agents to engage in goal-directed reasoning and planning, beyond simple “if–then” rules, improving realism in decision-making. | Archibald et al. (2024) |
| Behavioral outputs (evacuation vs. non-evacuation) | Outcomes depended on agents’ internal states and message framing. For example, high threat + high efficacy increased evacuation, while high threat + low efficacy led to fear control or inaction. | Schmidt-Colberg et al. (2024), Morss et al. (2024) |