The Challenger and Columbia disasters were not caused by missing data. Engineers had accurate risk information in both cases, and it was filtered or diluted before it reached the people who could act on it. My research asks whether agentic AI can help human decision-makers see those warnings, even when the organization pushes against them.
I am designing a two-agent, decision-support framework. The first agent monitors mission data and detects dangerous anomalies. The second escalates those signals through a channel that does not depend on the normal chain of command. It also logs every override attempt and flags when what engineers report differs from what management passes up the line. People keep the final decision at every step. The framework's job is to make sure the warning cannot disappear silently.