CSCW 2019 Workshop
We will be running a workshop on contestability at CSCW 2019! Please join us! Workshop website!
We study how users perceive and interact with potentially biased and deceptive opaque algorithms. What factors are associated with these perceptions, and how does adding transparency into algorithmic systems change user attitudes?
In this work, we study how users engage with difficult-to-validate control settings for social media, and find that they can function as placebos.
To understand consumer pain points and opportunities for tech interventions, this paper presents the results from two need-finding studies that explore the gold-standard of personalized shopping: interacting with a personal stylist.
How do users discover and behave around algorithmic biases? A cross-platform audit of online rating platforms revealed that users try to raise awareness of bias from within the platform.
Using polylingual topic models (PLTMs), we learn latent fashion concepts jointly in two languages: a style language describing outfits and an element language labeling clothing items. This model allows us to translate between the two languages, exposing the elements of fashion style.
Medical crowdfunding helps patients receive financial support, but little is known about who the patient's supporters are, what support they provide, and why. Interviews addressed these questions; we suggest making the variety of volunteering contributions more visible and discuss the associated design challenges.
Interviews revealed 10 "folk theories" of automated news feed curation, some quite unexpected. Providing users a probe into the algorithm's operation also helped users quickly develop theories. Foregrounding these automated processes may increase interface design complexity but may also add usability benefits.
CHI 2015 Best Paper
Users are often surprisingly unaware of the algorithms that permeate their digital lives. This work (with Motahareh Eslami) developed a system to reveal the differences between curated and unfiltered News Feeds to users and found that users had often inferred social meaning from algorithmic filtering.
NSF Trustworthy Algorithmic Decision-Making Workshop
How can and should users be able to appeal algorithmically generated decisions?
AAAI 2017 Spring Symposium
We propose an experimentation engine for fashion interfaces: leveraging social media platforms to run multivariate design tests with thousands to millions of users.
CSCW 2014 Workshop
We argue that the ACM Code of Ethics requirement to follow terms of service is problematic. While the reasons for following terms of service are clear, there are hidden costs. Using research into algorithm awareness transparency as an example, we argue that for some research problems the benefits of work violating TOS outweigh the harms.
Before Illinois, I worked at the MITRE Corporation, a federally funded R&D center. I focused on two areas: data mining/AI and signal processing. This included algorithm evaluation work, primarily dealing with identity management, social network analysis and NLP. I also worked on several signal processing projects.
I graduated from Reed College in Portland, Oregon. While there, I worked with Joel Franklin to develop a novel application of speech recognition to sociolinguistics. I developed a proof-of-concept tool that could detect socio-economic and other background features of speakers.
Please check out the webpages of some of the students I work with:
- Dylan Huang Website Github
- Sunaya Shivakumar Website Github
- Ziqiao Ding Github
- Lisa Huang Website
- Tanvi Agarwalla Github
- Kevin Ly Website
- Hanzi (Amber) Shen Website
If you are interested in working with our group, please email me!
I have TA'd a number of courses at the University of Illinois. I've also been selected to teach in the Graduate Academy for College Teaching, the University of Illinois' campus-wide TA training program. Teaching materials and/or information for each course below: