Towards optimal payoff manipulations

Introduction Multitasking typically requires people to make performance trade-offs: paying more attention to one task can improve performance there, but might lead to performance decrements on other unattended tasks. In our work we try to gain a better understanding of how people make such tradeoffs. One difficulty in this effort is that performance is typically expressed in different units across tasks (e.g., “accuracy” of keeping a car inside a lane and “speed” of performing a secondary task such as dialing). How do people trade-off these different units? Explicit payoff functions have been proposed as a way to achieve the desired trade-off (e.g., Howes, Lewis, & Vera, 2009; Janssen, Brumby, Dowell, Chater, & Howes, 2011; Payne, Duggan, & Neth, 2007; Schumacher, et al., 1999). They can be used to translate performance on multiple tasks into a single score. The participant and the modeler can then use this feedback to objectively compare performance for different strategies (Howes, et al., 2009; Janssen, et al., 2011). If successful, payoff functions can be used as a formal way to manipulate a user’s priorities. Different strategies can be made optimal through changes of the payoff function. In ongoing work we are exploring under what conditions such “ideal payoff manipulations” can be made. We report some of our intermediate findings here.