Supporting Data Literacies Through Online Interactions
In today's increasingly data-driven world, the abilities to ask and answer questions with data are beyond important. However, formal data science training resources often only emphasize technical skills while disregarding the development of critical capacities that's essential for working with data. How can we design to help people learn these critical skills via informal, accessible, and creative ways? In this series of projects, we will study online data science communities and design to support people to effectively ask questions about and explore data. Check later for more publications!
Community-generated Learning Resources and Creativity around Data Structures
In this mixed-method analysis of data from Scratch, we examined how novices learn to program with simple data structures by using community-produced learning resources. Surprisingly, we discovered that unstructured interest-driven content creation can result in limited opportunities for informal learning. We present a qualitative study that describes how community-produced learning resources discourage exploration and diverse interests in the community and find broad support for this dynamic through quantitative hypothesis tests. Our findings identify a social feedback loop that can restrict learning opportunities over time by limiting sources of inspiration, posing barriers to broadening participation, and confining learners’ understanding of general concepts. We conclude by suggesting several approaches to mitigate these dynamics.
Knowledge Building in Online Data Science Competitions
Knowledge building is a prevalent feature in open online systems, but it is challenging to motivate participants to contribute and to maintain quality in their contributions. In the context of Kaggle, an online data science community where community knowledge is curated in the form of shared computational notebooks, we explored how competitions could be used as a potential design model to incentivize contribution. By interviewing 14 Kaggle participants, we found that competitive mechanisms could be used to motivate experts to contribute to shared knowledge, but could create challenges for beginners to do so. We offered design implications for effectively implementing competitive mechanisms that could benefit both expert and beginner participants in future knowledge building systems.
Mapping the Design Space of Human-AI Interaction in Text Summarization
Automatic text summarization systems commonly involve humans for preparing data or evaluating model performance, yet, there lacksa systematic understanding of humans’ roles, experience, and needs when interacting with or being assisted by AI. From a human-centered perspective, we map the design opportunities and considerations for human-AI interaction in text summarization and broader text generation tasks. We first conducted a systematic literature review of 70 papers, developing a taxonomy of five interactions in AI-assisted text generation and relevant design dimensions. We designed text summarization prototypes for each interaction. We then interviewed 16 users, aided by the prototypes, to understand their expectations, experience, and needs regarding efficiency, control, and trust with AI in text summarization and propose design considerations accordingly. Check later for more publications!
Many Destinations, Many Pathways: Legitimate Peripheral Participation in Scratch
While informal online learning communities have been increasingly proliferating, what exactly are users of these communities expected to learn? Guided by the theory of communities of practice, we identify three distinct types of learning goals common to informal learning communities and a set of explanations that describe how different types of legitimate peripheral participation by newcomers may be associated with these three learning goals. Using data from the Scratch online community, we test these explanations as hypotheses in a quantitative analysis and find support for all. Our study contributes both theoretical insights and empirical evidence on how different kinds of learning happen in informal online environments. Check later for more publications!
Feedback Exchange and Online Affinity
Feedback is a critical piece of the creative process. However, creators run into socio-psychological challenges when engaging in online critique exchange with people they have never met. In this study, we step back and take a different approach to investigate the space of online feedback by casting our attention to creators in online affinity spaces, where feedback exchange happens naturally as part of the interest-driven participatory culture. We interviewed 29 fanfiction writers and identified distinct feedback practices specific for different stages of creativity and unpack the socio-emotional needs around feedback in each stage. Our results emphasize affinity and trusting relationships in feedback exchange and map out the amateur creator’s journey in building up those relationships. Check later for more publications!
Feedback Seeking in Online Creative Community
Creative workers frequently turn to online critique communities for feedback on their work. While past research has focused primarily on how to yield better feedback from providers, less is known about the strategies feedback seekers use to engage providers and request feedback. In this study, we took a mixed-method approach to study how creators publicly request feedback in an online critique community. We first identified creators’ strategies to request feedback in the community through 12 interviews with members of r/design_critiques. We then performed qualitative and computational analyses on requests and feedback posts to investigate the effectiveness of the strategies. We offered design implications for future online feedback systems.
Problem Framing in Online Distributed Design Team
In 2017-2018, I worked with Prof. Steven Dow in UCSD Protolab and Dr. Joel Chan on how to incorporate experts’ knowledge to improve problem framing in distributed collaborative design. In order to find a systematic way to select effective frames in distributed online design platforms, we collected opinions on frames from domain experts and conducted online experiments to test whether those opinions predict the effectiveness of frames by evaluating quality of resulting ideas generated from real users of our design platform.
Plug-N-Talk
In 2017, my team and I developed Plug-N-Talk, a smartphone application prototype that serves as an affordable solution to the pricey hearing aids market. By using our app, a pair of earphones and a microphone, users can enjoy a real-time in person conversation with background noises effectively reduced. Users can also easily adjust the enhancement of specific frequencies according to their hearing needs. This project was part of UCSD Design Competition: Design for Senior Citizens in 2017. I worked as the leading UX researcher along with other 8 inspiring students. We spent 20+ hours observing in a retirement community, and interviewed a group of seniors. We regularly followed-up with a senior member of the community and studied difficulties and needs in her daily life using a rapid ethnographic approach. We identified the problem that current hearing aids on market are all very pricy, and usually cannot meet the user's needs for reduced background noise during conversations and enhancement for a specific frequency band of sound. We held brainstorming sessions with seniors and tested our low-fidelity prototype in a wizard-of-oz fashion. The final high-fi product, an ios app prototype equipped with a pair of earphones and a microphone, is the result of countless iterations and coordination between the design team and the engineering team.
Data Science Narrative in Computational Notebooks
In 2017, I worked with Prof. Jim Hollan and Dr. Adam Rule in the UCSD Design Lab to study the narrative features among the Jupyter Notebooks using by the scientific data analysts community. I helped with building a dataset of 1.5 million computational notebooks and metadata and conducting thematic analysis and topic modeling. The resulting paper won an Honorable Mention award in CHI2018! We also built Jupyter Notebook extensions to support narrative in data analysis.
Application of EEG Neurofeedback on Working Memory Defect Treatment for Schizophrenia
In 2016 - 2018, I worked with Prof. Jaime Pineda and Dr. Fiza Singh at the UCSD Cognitive Neuroscience lab to study the effect of EEG-gamma neurofeedback on working memory in patients with schizophrenia. In this study I wrote MATLAB scripts to adapt a machine learning algorithm created by a collaborator in Spain to the data we collected from patients, which provided solid evidence to the effects of our neurofeedback.