The answer to this question is increasingly "no." Multitasking is endemic in American culture today. The opportunity to surround ourselves with screens and sounds continues to grow and is prompting behavioral and social changes. Multitasking is fundamentally changing the rules for how we interact with one other, the way we work, and even the way we think. In spite of this, multitasking is not often a direct consideration for those of us who design speech technology applications, so I've been sorting through the research on multitasking to understand what implications there may be for voice interaction design.
First, as a matter of clarification, the term multitasking is somewhat of a misnomer. There are very few instances of truly engaging in multiple tasks at once, at least for tasks that require active attention. Instead, we rapidly switch our focus from one task to another, which does not always lead to the efficiency implied by the term multitasking.
Studies on multitasking make it clear that the propensity to multitask is linked to age. Those born after 1980 are much more likely to multitask than members of Generation X or Baby Boomers. This may be attributable to changing social norms and the ubiquity of multiple media sources, but it may also reflect cognitive limitations. Older users are more susceptible to distractions when multitasking, which can result in errors and delays in completing tasks. The difference between older and younger people is observable at the neuronal level in the patterns of blood flow while multitasking, so this is a solid finding.
There are also differences in how successfully different individuals are able to multitask. It's a sad truth that people who routinely multitask are less able to filter out irrelevant information as they switch from one task to another, which makes them more prone to making mistakes. Similarly, even short interruptions can affect our ability to complete tasks accurately and efficiently. Even more troubling is the fact that people who believe they're good at multitasking, and who therefore do so often, are in fact worse at it than people who rarely multitask.
So what does all of this mean for voice interaction design? Clearly, we have to acknowledge that users of speech applications may well be multitasking, which means we're not getting their full attention. In-vehicle speech applications have had to wrestle with the issue of the user's attention being divided from the start. Automotive speech interfaces tend to be more streamlined and less chatty, which will limit the amount of irrelevant information that may distract users. This same lesson may be useful in the design of IVRs and mobile apps, whose users are also likely to be multitasking. A case in point is Web jumpers, who are simultaneously or in quick succession interacting with an organization across multiple channels. These customers may not be paying full attention to the IVR because they're still trying to complete their task on the Web site.
Another implication is that it may take time for distracted users to switch from one task to the other. This means that speech applications must tolerate pauses without penalizing users, allowing them to jump back into the conversation when they're ready. Depending on the length of the pause, users may require some reorientation on where they are in the interaction when they come back.
A final implication concerns how we test speech applications. Traditional usability tests, whether in the lab or remote, create a situation in which users focus solely on the application being tested. In years past, we recognized that usability tests didn't account for real-world scenarios, such as boisterous children and ringing doorbells in the background, but today, we're also failing to account for reading emails, checking Facebook, and receiving text messages during the interaction. Single-focus usability tests may still have diagnostic value in helping uncover problematic interactions, but the results of such tests are less predictive of real-world performance. Running usability tests in the field (in which we observe users interacting with applications in their daily lives) can provide a rich source of data, but can also be logistically challenging. We can borrow a technique from experimental psychology to make traditional usability tests more predictive of the multitasking environment: giving participants a distractor task to complete while they interact with the application. Counting backwards from 100 by threes or circling all the letter As in a paragraph is probably not what's distracting customers in real life, but it can mimic the sort of divided focus that's common among today's multitaskers.