HomeResearchOverview Skill Acquisition Multi-tasking Time perception Learning from instructionsACT-R/LISA PublicationsPersonal

Multi-tasking

Note: for the latest on my multitasking research, check out: "Towards safe and productive human multitasking".

People have the remarkable ability to do many things at the same time. Despite the fact that they are good at it, they are not perfect. Studies of using a cell phone while driving have shown that the quality of driving suffers from using the phone, even if when using a hands-free model. The challenge of multi-tasking research is to have a theory on when multiple tasks interfere with each other, and when they don't.

Driving a car while using a navigation device, from Salvucci, Taatgen & Kushleyeva (2006).

Threaded cognition

Dario Salvucci and I have developed a theory that we call threaded cognition. The basic assumption about the human cognitive architecture that we use is that it consists of a set of modules. A module can be vision, or control of the hands, but also modules in the brain, like long-term memory access, a representation of what you are currently doing, or the perception of time. All of these modules can operate in parallel. However, a single module can only do one thing at a time, and is therefore serial. So, our conception of the human cognitive architecture is one of a set of serial modules that operate in parallel. Given this nature, the architecture should be able to do multiple tasks, as long as there are no conflicts between the modules. If there is a conflict, this has to be resolved in some manner. In that case, the goals start interfering with each other, and one goal has to wait for the other.

For example, if multiple goals need the eyes to look at something, only one of them can have control. Another example is memory: if multiple goals need memory, they have to wait for each other. This can explain interference in driving a car and using a cell phone: both processing a complex traffic situation and interpreting language put heavy demands on memory.

Heavy memory demands are also typical for novice behavior. That is why it is almost impossible to do two new things at the same time. As soon as we get more experienced, the need for declarative memory diminishes (through production compilation, see skill acquisition), giving us the ability to do something else at the same time. In existing theories of control in multi-tasking, resources are arbitrated by a central executive that is often ill-specified. In threaded cognition, there is no central executive: the different threads compete for resources using a greedy/polite strategy.

If you want to used threaded cognition in your own models, you can download it with examples from the threaded cognition webpage.

Individual Differences

Do individuals differ in their ability to do multiple things at the same time? There is ample evidence that this is indeed the case. Although various factors may play a role, our hypothesis is that a key factor in individual differences is the ability to optimally structure the internal control structure of a task. Control should be as minimal as possible, but just enough to do the task. Our idea is that people who are less proficient at multi-tasking have unsound control structures. This improper structure plays out when a task has to be combined with another task.

Funding

This project is funded by the European Research Council

Key references

Salvucci, D.D., & Taatgen, N.A. (2011). The multitasking mind. New York: Oxford University Press

Salvucci, D. D., & Taatgen, N. A. (2008). Threaded Cognition: An Integrated Theory of Concurrent Multitasking. Psychological Review, 115(1), 101-130. (pdf)

Taatgen, N.A. (2007). The minimal control principle. In Gray W. (Ed.), Integrated Models of Cognitive Systems (pp. 368-379). New York: Oxford University Press.(pdf)

Anderson, J.R., Taatgen, N.A. & Byrne, M.D. (2005). Learning to Achieve Perfect Time Sharing: Architectural Implications of Hazeltine, Teague, & Ivry (2002). Journal of Experimental Psychology: Human Perception and Performance, 31(4), 749-761. (pdf)