Study finds algorithmic aversion in military AI decision-support systems, mitigated by explainable AI features

A recent study titled “Black Box Warfare: Human Judgment and Military Decision-Making in the Age of AI” empirically examined how AI decision-support systems (DSS) influence military targeting decisions. Researchers reconstructed a high-fidelity replica of a real-world military AI system and evaluated its impact on combat decisions through two experiments involving 2,015 Israeli military personnel.

Contrary to widespread concerns about automation bias—where humans over-rely on algorithmic recommendations—the study found strong evidence of algorithmic aversion. This skepticism was most pronounced in scenarios involving high collateral damage, where soldiers were less likely to approve AI-suggested strikes.

The research also tested the effect of integrating explainable AI features into the DSS interface. When such transparency features were present, algorithmic aversion decreased, and soldiers engaged in more deliberate evaluations of the system’s recommendations. The study concludes that trust in military AI is not static but varies with individual predispositions, perceived operational stakes, and the informational features of the interface.

The authors argue that these findings ground normative debates about AI in warfare in empirical evidence, emphasizing the enduring importance of human agency in high-stakes military decision-making. They suggest that interface design—particularly the inclusion of explainable AI—can help calibrate trust and improve the responsible integration of AI in conflict environments.