Multi-Agent · Evaluation

Evaluating maPO Explanations

Two complementary evaluations — automatic readability (cognitive load) and a human user study across RL, CBS & ICBS scenarios.

94.4Flesch Reading Ease“very easy” · grade ~1–5

95.2%Preferred maPO159 / 167 judgements

4.40Clarity / 5p < .001

N = 28Participantsgraduate & undergraduate

Cognitive-load metrics (FRE · ARI · CLI) confirm the explanations are linguistically simple and easy to comprehend — even for complex 7-agent scenarios.

* maPO & OMEGA work additionally with Ritirupa Dey (University of South Carolina).