| Humans tend not to spend excess energy when they perform a task. However, subjective factors, such as comfort, impact how they behave. Some studies use optimization to explain peoples' decision making process. A task's goal represents a constraint on behavior, and the chosen, or optimal, behavior is that which minimizes some cost function. Researchers often assume energy or kinematic variability costs, which may miss important subjective motivations. In this work, we leverage constrained optimization to predict and control behavior based on a general subjective cost. The ability to quantify and control decisions for behavior could benefit motor learning research, rehabilitation, and strength training. We developed feedback to uncover subjective costs associated with a number of exercise tasks. We alter task constraints, and their associated subjective costs, by unevenly weighting limb power toward a goal sum of this weighted power. The unknown subjective cost function may thus be characterized by sampling the preferred strategies for a range of different constraints. This method can be used to both characterize subjective costs and provide a framework to direct effort toward specific limbs. Results indicate that healthy subjects split effort between limbs based on more than economy alone, such as grip type, or reach length. These factors can alter the subject's effort distribution between limbs by about 15% of the mean net power performed. Implicitly weighted feedback was used to unveil the trade-off between excess mechanical power generated and factors beyond economy. The relationship between weightings and limb use allowed us to predictably shift effort toward arms or legs up to 37%. We compared implicit feedback with an explicit alternative, which did not weight effort contributions, but rather specified power and effort distribution targets. When they used implicit feedback, subjects displayed 74% less error relative to their feedback goals, and were able to perform simultaneous cognitive tasks 4.2% faster. Finally, subjective costs inform behavior outside of multi-limb exercise. In a drop landing experiment, subjects performed up to 32% less excess work when they dropped on more cushioned surfaces. These experiments allowed us to quantify subjective costs and predict their effect on human behavior. |
