As voice-operated AIs become part of everyday living, children talk to them, they ask questions, they make jokes, they test their intelligence, or they demand to hear a specific tune. These technologies answer, interact with them and collect enormous amounts of personal and highly contextual data. Yet we have very little knowledge or understanding about the social and political implications of voice data.
We tend to believe that voice is ephemeral. Yet a voice print, like a fingerprint, is a uniquely identifiable piece of data, and once that data is linked to an individual it can be used to infer many things about them, from their health to their socio-economic and educational background. This raises critical questions about what companies actually do with our voice data and the data of children.
In the last few years, campaign groups and lawmakers have paid particular attention to how voice-operated virtual assistants or AI toys could impact on children’s privacy. In 2017 Mattel cancelled its Aristotle AI assistant for kids amidst privacy concerns. In 2018, we witnessed a growing debate from lawmakers in the U.S. about Amazon’s use of children’s data in the Amazon Dot Echo for Kids. In 2019, the Campaign for a Commercially Free Childhood has launched a new investigation into Dot Echo for Kids that has shown that Amazon retains children’s data also after parents tried to delete it.
Yet when we think about the implications of children’s voice prints the question at heart is not only one about privacy but it is about human rights and social justice.
It is for this reason that, in May 2019, I decided to join forces with a leading expert in the field Dr Patricia Scanlon, who is the founder and CEO of SoapBox Labs, a ‘privacy by design’ business that specialises in children’s speech recognition.
Together, we submitted a report titled ‘Voice Prints and Children’s Rights’ to the Call for Submissions on the UN General Comment on Children’s Rights in Relation to the Digital Environment, which asked researchers, businesses and experts to provide evidence on how the current digital environment threatens children’s rights.
In our report, we contend that in the current digital environment, speech recognition technologies do not only pose great risks to children’s right to privacy, but they threaten other rights, such as their right to self-expression, freedom of thought, optimum development, and non-discrimination.
One of the greatest threats posed by speech recognition is represented by the fact that biometric data can be extracted from voice data, and hence companies can use this speech data from children to potentially create unique ID profiles; extract emotion and sentiment data and catalogue other sensitive information from transcriptions of their speech – that can follow them across a lifetime.
In our report we also show that speech recognition technologies – that are not designed for children – but nevertheless are technologies with which children interact on a daily basis expose them to adult-centred, digital environments threatening their best interests.
We see this report as a very first step towards the development of a worldwide call for action on children’s voice data use, and we hope to involve researchers, law-makers, businesses and non-for-profit organizations who work on the issue. We believe that much work needs to be done to raise awareness on the voice data of children, and how this data can impact on their rights.
VOICE PRINTS AND CHILDREN’S RIGHTS (by Dr Veronica Barassi and Dr Patricia Scanlon).