‘A really smart guy plugged himself into Alexa and now she’s smart too’ – children’s understanding of smart speakers.
By Valentina Andries and Judy Robertson.
Smart speakers – such as Alexa, Siri and Google Home – have entered our lives and are now comfortably installed on kitchen tables across the land. A recent OfCom report notes that 83% of children have at least tried using a smart speaker, while our own research in Edinburgh found that 91% of the children in our sample had a smart speaker in their home. Because smart speakers typically use voice interactions and are designed to be conversational assistants, they emulate human-human interactions. But could talking to a smart speaker as if it is human be confusing for children? Might they then assume the technology is more intelligent and human-like than it actually is? What consequences might this have for their interactions with technology and their attention to privacy? Our current research aims to document a baseline of children’s knowledge about smart speakers and address potential misconceptions about these systems through educational materials for schools.
We have employed a mixed-methods approach with primary-school children to investigate themes such as children’s general knowledge of Artificial Intelligence and how smart speakers work, user data management and security aspects, as well as children’s verbal interactions with their smart speakers. So far, we collected and analysed data from two state-funded primary schools in the Edinburgh area.
It’s fascinating to learn what the children think about their smart speakers. The younger children were more inclined to assume that smart speakers are intelligent – more intelligent than themselves but not so intelligent as scientists! They gave a variety of explanations for what makes smart speakers intelligent, ranging from the view that their intelligence is the result of programming to the slightly alarming idea that a “really smart guy… plugged himself into it and then it [Alexa] got really smart and then he plunged himself into multiple [Alexas] and now there’s an army”. Not all children agreed that smart speakers were intelligent but rather that humans make them work – “I think Alexa’s made by humans who are on a computer and then they listen to what you’re saying. And then when it says, ‘Alexa’, it turns on and then they type something in.”
In terms of privacy, most of the children did not realise that others in their home, and people who work for the company which makes the smart speaker can review their commands and interactions. In fact, they were often horrified at the prospect. Some of them imagined unpleasant (but often technically inaccurate) scenarios in which sharing private information could go wrong -“Alexa and Siri work together and then they could spread … then it could go viral.” It is clear that the children value their privacy; we owe it to them to clearly and accurately explain the real privacy risks of using smart speakers and how they can mitigate these. As part of this, we need to educate children about the basic facts of how such technology works.
The ultimate goal of this project is to produce a range of educational materials as well as to collate and curate relevant resources that can be used by STEM teachers in Scotland to educate young children on how smart speaker technology works, the implications of its use, as well as foster AI literacy learning. We are adopting a participatory approach to developing such materials together with young learners and other relevant stakeholders such as specialists in education and teachers.
We are collating a list of resources and approaches that can be used and adapted by teachers in their classrooms. We have organised a workshop with teachers, during which we presented our findings, and we invited the teachers to a group discussion on different teaching approaches that they would adopt to address some of the children’s misconceptions or open-ended questions about AI and CAs. The insights that we gathered from that session are useful in further refining educational materials and approaches. We are also collaborating with Informatics students at the University of Edinburgh to share their knowledge of AI with young learners.
More about the project ‘Designing conversational assistants to reduce gender bias’