Congrats to @cqz@hci.social for graduating with his PhD from UW CSE!! Jim is the first PhD graduate from @socialfutureslab@hci.social and also my first PhD student! His dissertation was on "Understanding and Addressing Uncertainty of the Crowd", and you can read it here! ➡️ https://homes.cs.washington.edu/~cqz/dissertation.pdf
Remote
Amy Zhang
@axz@hci.social
mastodon 4.7.3UW CSE professor 👩🏻🏫 PI @socialfutureslab@hci.social making social tech better!
0 Followers
0 Following
50 Posts
Joined May 03, 2022
Homepage:
Lab website:
Twitter:
Whoo, my research lab @socialfutureslab@hci.social is on Mastodon finally! We decided to take the leap though our Twitter account is still up and will be updating as well for now. Follow for updates on research conducted in our lab at UW CSE on reimagining social and collaborative systems to empower people and improve society.
Congrats to Nouran Soliman on a fantastic talk on her best paper awarded work on meronimity (mitigating barriers to engaging in public conversations with partial, self-selected, and verified identity signals)! 👏👏 We also deployed a system for it right here in Mastodon at @LiTweeture@hci.social a while back 😌 Read the paper here: https://dl.acm.org/doi/10.1145/3613904.3642241 #chi2024
My dept at UW made a short video of the class I taught this spring on social computing with @kjfeng@hci.social https://youtu.be/9AWkEKRuDZg My students blew me away with their projects and insightful discussion! If curious, our syllabus: https://courses.cs.washington.edu/courses/cse481p/23sp/
Social Computing Capstone, CSE 481, Spring 2023
Please boost! We’re excited to invite applications for DUB summer REUs for US-based undergrads.
Interns will be matched with a
UW DUB faculty mentor, who will hire & lead the intern on an HCI research project. ~9 weeks, paid, on UW campus.
Deadline is 1/27: https://dub.washington.edu/posts/2023/dubreu.html
Replying to
After the interviews and ~2 years of iteration and feedback, we developed the FABLE framework of factors that influence magnitude of misinformation harms, encompassing: social (F)ragmentation, (A)ctionability, (B)elievability, (L)ikelihood of spread, and (E)xploitativeness.
Hey #FAccT2023! Please check out Teanna Barrett’s talk tomorrow on her paper:
"Skin Deep: Investigating Subjectivity in Skin Tone Annotations for Computer Vision Benchmark Datasets"
🔗 https://arxiv.org/abs/2305.09072
🗣️ Tuesday (7/13) @ 2:15pm CT in room W196A
📺 Talk video: https://drive.google.com/file/d/1Pn-q3xZjMNN4fLinb7DyGVCb1VtWZL3M/view
Teanna was an REU intern (!!) with us last summer, mentored by @cqz@hci.social (also attending!) and will be starting grad school next fall! If you’re at the conference, go talk with them both!
Question to academic researchers here: do you use generative models at all in your research (e.g., for literature review or helping with paper writing)? If so, how do you use it and how has it gone? If not, why not?
I was thinking about this q and realized some people might not be comfortable sharing publicly - if so, here’s an anonymous Google form: https://forms.gle/BjyvJ4HGMHsUNVdo7
Check out this article by @nickmvincent@hci.social and @hanlin@hci.social in Wired on ChatGPT and data theft: https://www.wired.com/story/chatgpt-generative-artificial-intelligence-regulation/
And Emelia Hughes giving her awesome talk! Good questions about novelty effects and possibility of reduced impact of citations or correction interventions over time in relation to systems like Community Notes #chi2024 https://arxiv.org/abs/2402.17218
Replying to
We show that this relatively low effort addition in annotation is deemed by medical imaging experts to be more interpretable than alternative uncertainty representations such as Bayesian maps, which have a smooth gradient, due to the way they capture structural uncertainty.
Leijie Wang giving a great talk (with a live demo!) of his no-code authoring system Pika, including grammar, compiler, component library, and form-based UI, for communities to create and carry out a wide range of governance online! #chi2024 https://dl.acm.org/doi/10.1145/3613904.3642012
@kjfeng@hci.social and Lawrence gave a great talk at #chi2024! Also some really interesting questions from the audience the work brought up - questions of ease and simplicity vs agency tradeoffs, dealing with as-of-yet unknown or revealed preferences, implications for content moderation. https://arxiv.org/abs/2401.14000
Replying to
With added issues like uneven rollout and adversarial usage by bad actors, we should design tools that do more curation, interpretation, and education around provenance chains so they are helpful for end users. Check out @kjfeng's thread for a summary of our findings and ideas! And link to paper is here: https://arxiv.org/abs/2303.12118
Replying to
@socialfutureslab Next @cqz gave a talk on Wed on targeted interventions to reduce uncertainty in judgments. Paper here: https://dl.acm.org/doi/10.1145/3610074 He also discussed how it fits into his broader research trajectory and agenda, as he's heading on to the job market this year!
Replying to
We found that users also have preconceptions about the respective capabilities of moderators at the platform versus volunteer community mods. For instance, users trust platform moderators more to not abuse their power but think community moderators have more time to review reports. 4/n
Replying to
We found that users expect that platforms are storing rich longitudinal reporting datasets, and recognize both their promise for better abuse mitigation and the privacy risk that platforms may exploit or fail to protect them. 3/n
Replying to
From there, we learned what are ppl's privacy concerns regarding the info they think is being shared. Our focus was E2EE apps, like WhatsApp, Signal, Messenger, etc, as reporting is even more important in those settings and ppl have heightened privacy needs, but we think our findings and ideas generalize. 2/n
Replying to
How do you get people to investigate provenance instead of relying on what they see? How do you convey malicious edits versus reasonable ones? What does it mean that a signed provenance chain is broken or has missing links? & how do you show users this without overwhelming them?
Replying to
@socialfutureslab And @kjfeng gave a talk on his collaboration with the C2PA UX task force investigating how end users interpret and react to media provenance indicators embedded in social media, particularly on trust and accuracy. https://dl.acm.org/doi/10.1145/3610061 Great job everyone!
Replying to
Read the full paper: https://arxiv.org/abs/2312.11678
Blog post: https://www.artt.cs.washington.edu/post/exploring-the-concept-of-misinformation-as-a-harm
Get just the worksheet here: https://www.artt.cs.washington.edu/docs/online-misinformation-harm-questionnaire
Work is led by Connie Moon Sehat, in collab w/ @socialfutureslab student Ryan Li and researcher Peipei Nie, and Tarunima Prabhakar of Tattle https://tattle.co.in/
Replying to
Also, instead of a 1-way 1-time interaction, imagine more flexible reporting interactions, including corroborations from bystanders, (consensual) escalations from community to platform mods and vice versa, or being able to progressively reveal information upon request from mods. 6/n
Replying to
Given these findings and ppl's privacy concerns expressed, we have ideas regarding how to improve reporting systems to be more privacy-sensitive! For instance, being able to decide to whom to report, or being able to select which messages to share in a report and obfuscate/remove sensitive info. 5/n
Replying to
I've had awesome REU summer interns come work with our lab
@socialfutureslab over the last several years, many of whom went on to publish papers based on their internship work! Such as: https://dl.acm.org/doi/abs/10.1145/3593013.3594114
https://arxiv.org/abs/2410.15527
https://dl.acm.org/doi/abs/10.1145/3626252.3630795
Replying to
Big thanks to the C2PA for supporting this work and the efforts and feedback from the C2PA UX Task Force who contributed to this research, especially
Pia Blumenthal, Nick Ritchie, and Andy Parsons!
Replying to
Coalitions like the C2PA (https://c2pa.org) are working on standards for attaching verified origin and edit provenance data to media. But there's lots of questions around how to convey a complicated topic like provenance to social media users and whether and how it will impact their perceptions!
Replying to
We propose a new annotation technique building on @cqz's work on using ranges to capture calibrated uncertainty for scales in the 2D setting by having annotators draw low and high "confidence contours" as opposed to asking multiple people to annotate a singular boundary.
Replying to
I also think there's interesting directions for new cryptographic techniques to verify aspects of reports that could be built to help guard against malicious reporting. This work is part of a larger collab with Cornell Tech researchers who are experts in crypto, so more to come from them! 7/n
Replying to
This work will be presented at #SOUPS2023 next week by Leijie Wang from @socialfutureslab along with @ruotongw @sanketh@wandering.shop and Sterling Williams-Ceci! 8/8
Replying to
We provide a series of questions to help with determining relative values. This can serve as a fuzzy indicator to be used and weighed however people/orgs find appropriate, recognizing that diff groups have diff goals & that people have limited time for investigation.
Replying to
From interviews w fact-checkers from orgs around the world, we learned that while prioritization is a big concern (too much to fact-check, not enough time or resources), triaging is often an ad hoc process. *Potential for harm* however is indeed a major factor in decision-making.
Replying to
We also discuss avenues for leveraging this framework towards other areas where harms-informed triage is useful (e.g., content moderation), and the possibility for automated tools to help w/ ranking items according to the dimensions or gathering supporting evidence. More coming!
Replying to
In addition, we will be organizing a guided visit to the George Floyd Memorial on Wed during lunch time. Lunch will be provided, and travel there takes about 30 min. If you're interested, please fill out the form, and we will get back to you ASAP: https://forms.gle/Y8n2FaRSt4Vtkcgs6
Replying to
Finally, inspired by fempower.tech's craft pop-ups at CHI (https://fempower.tech/) we will also be hosting some pop-up crafting stations during the coffee breaks on Tues, as a low key way to meet new people while doing some crafting! Look out for our sign in the open space where coffee will be served.
Replying to
Replying to
@pg @kjfeng no we didn’t…the thought never crossed my mind to either! I do have a student looking to write up her experience teaching social computing topics to middle and hs students though https://social.cs.washington.edu/sfl-curriculum/
Replying to
At the luncheon, we will have a keynote from Dr. Sheena Erete @sheenaerete@bird.makeup as well as presentation and discussion of the 15 papers recognized for contributions to diversity and inclusion (all papers are listed on the webpage linked above! Congrats winners!).
Replying to
Some things I want to call out: 1) the return of the Diversity Luncheon on Mon! We only had budget for 100 attendees (thank you to the SIGCHI Development Fund and CSCW chairs for funding), so if you were selected to attend from our raffle, you should have received an email.
Replying to
@imjane I am working on a project related to it, mostly looking at differences according to system designs. I'd be curious what other papers you find!
Replying to