Faculty Bibliography

Audio annotation is key to developing machine-listening systems; yet, effective ways to accurately and rapidly obtain crowdsourced audio annotations is understudied. In this work, we seek to quantify the reliability/redundancy trade-off in crowdsourced soundscape annotation, investigate how visualizations affect accuracy and efficiency, and characterize how performance varies as a function of audio characteristics. Using a controlled experiment, we varied sound visualizations and the complexity of soundscapes presented to human annotators. Results show that more complex audio scenes result in lower annotator agreement, and spectrogram visualizations are superior in producing higher quality annotations at lower cost of time and human labor. We also found recall is more affected than precision by soundscape complexity, and mistakes can be often attributed to certain sound event characteristics. These findings have implications not only for how we should design annotation tasks and interfaces for audio data, but also how we train and evaluate machine-listening systems.

Peer production communities often struggle to retain contributors beyond initial engagement. This may be a result of contributors' level of motivation, as it is deeply intertwined with activity. Existing studies on participation focus on activity dynamics but overlook the accompanied changes in motivation. To fill this gap, this study examines the interplay between contributors' fun motives and activity over time. We combine motivational data from two surveys of Wikipedia newcomers with data of two periods of editing activity. We find that persistence in editing is related to fun, while the amount of editing is not: individuals who persist in editing are characterized by higher fun motives early on (when compared to dropouts), though their motives are not related to the number of edits made. Moreover, we found that newcomers' experience of fun was reinforced by their amount of activity over time: editors who were initially motivated by fun entered a virtuous cycle, whereas those who initially had low fun motives entered a vicious cycle. Our findings shed new light on the importance of early experiences and reveal that the relationship between motivation and participation levels is more complex than previously understood.

Recent advances in genetic testing and Internet technologies have led to a dramatic increase in the access non-experts have to their own personal genomic data. Such data are complex and sensitive, involve multiple dimensions of uncertainty, and can have substantial implications on individuals' behavior, choices, and well-being. Personal genomic data are also unique because unlike other personal data, which might change frequently, genomic data are largely stable during a person's lifetime; it is their interpretation and implications that change over time as new medical research exposes relationships between genes and health. Future progress in genetic research and technologies is likely to further increase the availability of interactive personal genomic information to non-experts. This trend raises technological, ethical, and regulatory concerns related to how people make sense of, engage with, and rely on their personal genomic data. Such concerns are not only of paramount importance for health professionals and policymakers, but are also a pressing issue for human-computer interaction (HCI) research. HCI tools, methods, and practices can help make genomic information more accessible and understandable to non-experts. We argue that the complexity, importance, and personal relevance of this type of information makes understanding, informing, and empowering non-experts' interaction with personal genomics a key challenge that lies ahead for the HCI community. In this article, we explore the roles HCI can play in helping nonexperts contribute, understand, engage with, and share their personal genomic information. This article is also a call-to-action for those of us interested in the intersection of personal informatics and HCI, and, more broadly, in facilitating non-expert interaction with large amounts of complex, personal, and uncertain information.