Task Switching
Abstract
This review examines what happens when people must shift between cognitive tasks. There is a robust 'switch cost' in both reaction time and error rate. The cost persists even with long preparation time and reflects both reconfiguration of task-set and proactive interference from the competing task. The cost increases with task complexity and the number of active task-sets.
★ Curator Summary
Monsell's research is the cognitive science foundation for the N* collaborative staffing model. When agents monitor multiple concurrent AI interactions and intervene as needed, each context switch has a measurable cost that increases logarithmically with portfolio size: gamma(N) = gamma_0 + gamma_1 * ln(N). This isn't just response time — it's error rate too. The research proves there's a fundamental cognitive ceiling on how many concurrent interactions a human can effectively oversee, and that ceiling is lower than most AI-augmentation business cases assume.
Why It Matters
The collaborative pool — where humans oversee multiple concurrent AI interactions — only works if you model switching costs correctly. Monsell's findings mean you can't just divide total intervention time by available agent hours to get staffing. You need to account for the logarithmic degradation in human performance as portfolio size increases. Get this wrong and your 'AI-augmented' agents will be slow, error-prone, and burning out.
Caveats
Laboratory cognitive psychology research with controlled tasks — not validated specifically in contact center multitasking environments. Real-world switching costs may differ from experimental estimates. The logarithmic cost function is an approximation — individual variation is significant. The research predates AI-assisted workflows specifically.
Discussion
Loading...