Abstract
Multitasking occurs frequently. For example, office workers switch tasks every two to three minutes. However, little is known about the efficiency of such frequent switching. Do people interleave tasks in a way that maximizes rewards? We investigated this in an experiment. Participants had to divide their time between two identical tasks: "whacking moles" that appeared in a 3 by 3 grid by pressing corresponding buttons on the numeric keypad. Each mole contributed points towards the total score and a trial ended as soon as 50 moles were hit. The two tasks differed in how rewarding each mole was, as defined by mathematical functions. A model showed how the exact combination of rewards affects the difficulty of finding the optimal interleaving strategy. Three unique scenarios emerged. In easy scenarios, many strategies (for interleaving tasks) achieved optimal performance. Participants therefore varied in the strategies that they applied. In difficult scenarios, the optimal strategy was hard to identify (e.g., due to local maxima), and participants spent longer searching for it. This also resulted in high variation of strategies. Finally, in constrained scenarios there was a well-defined optimal strategy, which participants consistently applied. This work demonstrates how reward functions influence the ability to optimize task interleaving in general. Future work will investigate how this ability changes under different circumstances (e.g., time pressure). By identifying when multitasking is challenging, we can inform the design of safe multitasking systems that maintain productivity, while reducing dangerous situations.
Original language | English |
---|---|
Publication status | Published - 2015 |
Event | NVP Winter Conference 2015 - Egmond aan Zee, Netherlands Duration: 17 Dec 2015 → 19 Dec 2015 |
Conference
Conference | NVP Winter Conference 2015 |
---|---|
Country/Territory | Netherlands |
City | Egmond aan Zee |
Period | 17/12/15 → 19/12/15 |
Keywords
- multitasking
- rewards
- cognitive model
- performance trade-offs