Stakeholders:
State: Catastrophe Incoming
Monitoring and deep learning lead to lead to fairly accurate predictions roughly 5 days in advance
*→ Accountability: Who darms? Agent or human?*
+Agents in Checks and Balances
+Danger clustering (semi autonomous warnings) → Hawaii nuke case must be prevented
Agent Task: Defining degree of action: impact space and prediction / heatmap
*→ Selection Issue (Bias)*
+Utilitatian selection based on values (eg. social scoring)
+Random selection
*→ Freedom of Choice "What if I don't want to go?"*
+Personal choice must overrule utilitarism
+Provide backup plan for non-rescuables