The challenges of children and voice assistants

As voice based interfaces and conversational interfaces are becoming increasingly popular, there is one group that the applications utilizing speech recognition software cannot really comprehend; children. According to research made by Axios children’s screen time during the pandemic surged by as much 50% to 60% more. Meaning that children 12 and younger were spending upwards of 5 hours or more on screens per day. But as kids sometimes struggle to grapple with the user interfaces (often designed for adults in mind), voice assistants such as Siri and Alexa could be helpful. But here is where the problems lie: Siri and Alexa do not always understand what the kids are saying.

The problems stem from the fact that the speech recognition software never was designed with children in mind, and children’s voices, language and behaviours are far more complex than that of adults. Children can change volume very fast, going from whispering to screaming in the same sentence. They do sometimes (or oftentimes) talk incoherently, and grammatically incorrect, over-enunciating words, elongate certain syllables, punctuate each word, or skip some words entirely. They talk in different ways during different phases of their development, and they also have slightly different bodies than adults. Children’s vocal tracts are thinner and shorter, their vocal folds smaller, and their larynx has not yet fully developed which results in very different speech patterns. All of this is very complex for technology, and means that in order to be compliant with kids, recognition systems need to be built to intentionally learn from the ways kids speak.

There are also more cognitive challenges within children’s use of conversational interfaces that needs to be addressed. Research shows that one of the obstacles young children face when using voice search is not fully understanding what the system can and cannot answer (i.e., what the system knows) and how much context to provide. A child could potentially ask Siri or Alexa where their parents are, or how old their grandfather is, without understanding that these are not questions suitable for voice searches. Another aspect is the child’s understanding of where the answers are coming from. Judith Danovitch, an associate professor of psychological and brain sciences at the University of Louisville, one day found her 4-year old son interacting with Siri. Danovitch started to record his questions in order to use them in her own research. She also made inquiries to who the child thought was answering his questions, and his response; a small person that was living inside of the device. This also ties into questions regarding whether or not the answers are trustworthy. Danovitch research has shown that Preschoolers’ trust in technology sources largely is based on previous experience, as it is with people, a behaviour that evolves with age. If Siri or Alexa misunderstands children due to their inability to understand their way of talking, the children might loose trust in it.

But even if the voice assistants ability to understand children is developed over the coming years, there are still more ambiguous questions related to the voice assistants ability to answer questions. Should they answer all the children's questions just because they can? When children ask adults questions the adults sometimes ask the child why they are asking that specific question, or what the child might think the answer is. Which in turn invites the child to figure out the answer, partly on their own. And through a dialogue like this children not only develop understanding, but also language and the ability to reason. All of which is important steps on their journey towards adulthood.

References

As Kids’ Screen Time Surges During The Pandemic, Here’s What Research Suggests
Voice assistants don’t work for kids: The problem with speech recognition in the classroom
Young Children and Voice Search: What We Know From Human-Computer Interaction Research
When AI becomes childsplay