In Vaccaro and colleagues’ framework, human augmentation means a human–AI system performs better than humans working alone on a specified task and performance measure. Human–AI synergy requires the combined system to outperform both humans alone and AI alone, so it is compared with the stronger solo baseline. Improvement over humans alone therefore does not establish synergy. For example, average accuracies of 70% for humans, 90% for AI, and 80% for the combination illustrate augmentation without synergy; these are hypothetical numbers, not study results. Evaluating synergy requires all three conditions on a comparable task and measure. Empirical comparisons also require reporting uncertainty. These definitions concern the selected performance measure and do not guarantee that collaboration improves outcomes.

Source: Vaccaro, Almaatouq and Malone (2024)