Inferential decision-making algorithms developed to date have assumed thatan underlying probabilistic model of decision alternatives and outcomes may be learned \emph{a priori} or online. As a result, when these assumptions are violated, they fail to provide solutions and are limited in their ability to modulate between optimizing and satisficing in the presence of hard time or cost constraints, adverse environmental conditions, or other unanticipated external pressures. Cognitive studies presented in this dissertation demonstrate that humans modulate between near-optimal and satisficing solutions, including heuristics, by leveraging information value of available environmental cues. Using the benchmark inferential decision problem known as a "treasure hunt", this dissertation develops a general approach for investigating and modeling active perception solutions under pressure, learning from humans how to modulate between optimal and heuristic solutions on the basis of external pressures and probabilistic models, if and when available. The result is an active perception approach that allows autonomous robots to modulate between near-optimal and heuristic strategies, tested via high-fidelity numerical simulations and physical experiments. The effectiveness of the new active perception strategies is demonstrated under a broad range of conditions, including decision tasks in which state-of-the-art sensor planning methods, such as cell decomposition, information roadmap, and information potential algorithms fail due to adverse weather (fog) or significantly underperform because of time or cost limitations.