An RCT assigns participants to interventions randomly, which helps reduce confounding and supports causal inference when well-designed.