Proceedings of the Annual Meeting of the Cognitive Science Society

People have a common-sense notion of intelligence and use itto evaluate decisions and decision-makers. One can attributeintelligence by evaluating the strategy or the outcome of agoal-directed agent. We propose a model of intelligence at-tribution, based on inverse planning in Partially ObservableMarkov Decision Processes (POMDPs) in a probabilistic envi-ronment, inferring the most likely planning parameters givenobserved actions. The model explains the agent’s decisionsby a combination of probabilistic planning, a softmax decisionnoise, prior knowledge about the world and forgetting, estimat-ing the agent’s intelligence by a proxy measure of efficientlyoptimising costs and rewards. Behavioural evidence from twoexperiments shows that people cluster into those who attributeintelligence to the strategy and those who attribute intelligenceto the outcome of the observed actions. People in the strat-egy cluster attribute more intelligence to decisions that min-imise the agent’s overall cost, even if the outcome is unlucky.People in the outcome cluster attribute intelligence to the out-come, judging low-cost outcomes as a sign of intelligence evenif the outcome is accidental and make neutral judgements be-fore they observe the result. Our model explains human in-telligence judgements better than perceptual cues such as thenumber of revisits or moves.