Investigating User Expectations on the Roles of Family-shared AI Speakers

The Last Decade of HCI Research on Children and Voice-based Conversational Agents

“…Another set of 23 studies collected and analyzed audio, video, and/or text-based interactions logs interactions logs (e.g., [15,52,154]). Five studies analyzed data collected from participatory design activities (e.g., design outcomes and audios/videos of design sessions) (e.g., [51,95]). Other data collection methods included drawings (𝑁 = 1; [149]), surveys, where-in the questionnaires did not comprise of pre-defined or customized scales to measure a phenomenon (𝑁 = 2; [25,78])), diary studies (𝑁 = 1; [50]), and a corpus for machine-learning modeling (𝑁 = 3; [24,72,151]).…”

“…France (𝑁 = 3) and the UK (𝑁 = 3) were the next two most common countries where studies took place. Four studies [50,95,117,125] did not collect data directly from children, but instead collected data about them from their parents. We also collated the age groups of the children in the studies covered by the reviews (see Table 3).…”

“…Garg and Sengupta [51] also referred to the persona effect. • Method (2): Park and Lim [95] used cultural dimension theory [57] to justify the peculiarities of conducting research in a specific contexts and their selection of participants. Serban et al [119] used the concept of the uncanny valley [88] to justify their decision of designing their own Wizard of Oz platform and interactive narration scenario.…”

“…Eighteen papers involved existing systems (e.g., Google Home [52], Amazon Echo [15]) or working prototypes developed by the researchers (e.g., [148]). In addition, seven papers focused on engaging children in co-design or participatory design work (e.g., [51,95]) to understand their preferences concerning CAs. Rader et al [104] explained how the design of a virtual peer could be informed by findings obtained from the analysis of a corpus of child-to-child language use; however, this paper did not fit into our categories and so has been omitted from Table 4.…”

“…Another four studies in this category involved design work aimed at achieving an understanding of the expectations of children and parents towards CAs within the context of home and family. Park and Lim [95], after conducting a participatory user study, suggested that devices should be able to develop an understanding of family dynamics and rituals and adapt their features to social versus private situations (e.g., avoid revealing personal and private information in shared spaces). In a similar vein, Luria et al [81] conducted speed dating and group interviews with families containing children who were aged twelve or older.…”