With the increased adoption and use of smart home speakers in households, many young children have learned to use listening and speaking to interact with the voice assistants (VAs) for a variety of daily activities (e.g., asking questions, listening to music). Despite the proliferation of VAs, due to the technical limitations such as automatic speech recognition (ASR) and natural language understanding (NLU), children often experience communication breakdowns with VAs, limiting educational opportunities through these voice-based interfaces. Informed by prior literature on child-computer interaction, child language development, and audio and trivia game design, I first designed a trivia game Animal Actions using Amazon Alexa with interdisciplinary research teams. I then conducted an in-home user evaluation study with 18 children between 3 and 6 years old using a mobile phone. Through video analysis of children's gameplay, I examined children's interaction with Alexa from four areas: children's verbal behaviors, children's response errors, Alexa communication breakdowns, and children's communication repair strategies. I found that single-word responses were more accurately recognized and processed by Alexa than were multi-word phrase-level responses. When comparing children's response accuracy and errors, children's responses included a wide range of speech articulation, semantic, and syntactic differences leading to incidents of communication breakdowns, which were not accounted for during game design and game item development. Parent questionnaires also revealed similar usability issues with automatic speech recognition and natural language understanding when children use VAs at home for daily routines, and parents reported using VAs independently and collaboratively with their children to mitigate communication breakdowns. Based on these findings, I discuss sociotechnical, design, and ethical considerations to inform future research on designing accessible user interfaces in voice games to support children's communication skills, including those with communication disorders.

This dissertation endeavors to deeply understand the features of Minecraft servers explicitly created for youth through three studies using mixed methods research. Human-Computer Interaction (HCI) research shows that sandbox-style virtual world games like Minecraft operate as interest-driven spaces where youth can explore their creative interests, build technical expertise, and form social connections with peers and near-peers. Despite their popularity among youth (ages 6 - 14), we know little about the social and technological features of "in-the-wild" Minecraft servers that present themselves as "kid-friendly" or "family-friendly." The aims of this work are three-fold:1. To investigate the rhetoric of kid-/family-friendliness and the socio-technical mechanisms of such servers (Study I: 60 servers), 2. To understand the lived experiences of server staff who moderate on such servers (Study II: 8 youth and 22 moderators), and 3. To explore a design paradigm for technological mechanisms that leverage the strengths of a kid-/family-friendly server community while also supporting moderators' practices (Study III) I draw from interdisciplinary theories and structure this dissertation around two main arguments about kid-/family-friendly Minecraft server ecosystems. First, I argue that they are instantiations of play-based affinity networks created by adults that promote opportunities for youth to explore their interests and social connections. Second, I argue that the social and technological mechanisms reflected in the server rules and moderators' practices are characteristic of servers that self-describe as kid-/family-friendly. Study I contributes a taxonomy for understanding server rules and an empirical characterization of three server genres – kid-/family-friendly (n1 = 19); general-family-friendly (n2 = 20); and general (n3 = 20) in Minecraft. Study II reveals moderators' motivations and socio-technical practices in kid-/family-friendly servers. The findings show that adult moderators encourage youth-led creative roleplays, support the interests of young players (e.g., Hogwarts virtual world, virtual Pride Day celebrations, etc.), and offer mentorship to youth moderators on their servers. Study III theorizes the potential for automated prosocial tools in play-based spaces through a Discord Bot called "UCIProsocialBot" within OhanaCraft, one of the kid-/family-friendly server communities. Together, these findings provide a set of social and technological features that may substantiate a model for designing kid-/family-friendly online playgrounds. This work theorizes that kid-/family-friendly servers can actualize positive youth development when their self-narratives, social practices, and technological mechanisms are aligned with adolescent developmental needs.