Each of the 170 people in the social graph carries a belief value Bi ∈ [0,1] — how convinced they are by AI-generated deepfakes / misinformation flowing along their social connections. It diffuses like a simplified epidemic (SIS) model along the graph's edges, opposed by fact-checking recovery:
dB_i/dt = β · (1 − B_i) · mean(B_j, j ∈ neighbors(i)) − γ · B_i
dJ/dt = α · J · (1 − J) · (1 − ρ)
- AI capability / realism — sets β, the transmission rate: more convincing generative AI makes deepfakes spread faster through the network, and also raises α, the pace of job automation.
- Media literacy / fact-checking — sets γ, the recovery rate at which an informed node stops believing a false claim and becomes resistant again.
- Regulation & oversight — the damping factor ρ on automation growth J (a logistic curve, since only so many roles can ever be automated), and adds a small extra recovery boost against misinformation.
- Election surge — temporarily multiplies β, modelling a burst of AI-generated political content and coordinated bot amplification during a campaign.