Georgia Tech @ CSCW 2017

CSCW is the annual ACM Conference on Computer-Supported Cooperative Work and Social Computing

Feb. 25 – March 1, Portland, Ore.

EXPLORE

Georgia Tech Shapes Research in Computer-Supported Cooperative Work as ACM Conference Turns 20

Georgia Tech computing faculty, students and alumni will play a central part in the Association for Computing Machinery’s Conference on Computer-Supported Cooperative Work and Social Computing in Portland, Ore., where the main program runs Feb. 27 – March 1.

Six faculty from the School of Interactive Computing have a combined eight papers accepted at CSCW 2017, including two of six best papers at the conference. These Atlanta-based researchers’ work covers a range of challenge areas, including privacy for social media, fake news, online movements, health tracking and digital self-harm.

Georgia Tech alumni are also making considerable contributions to the field, with 17 papers, including 3 honorable mention papers, by 13 authors.

CSCW convenes its 20th conference this year – which took place biannually from 1986-2010 and annually since 2010 – having become the premier venue for research in the design and use of technologies that affect groups, organizations, communities, and networks. The conference explores the technical, social, material, and theoretical challenges of designing technology to support collaborative work and life activities.

Research Highlights

Finding Credibility Clues on Twitter

By scanning 66 million tweets linked to nearly 1,400 real-world events, Georgia Institute of Technology researchers have built a language model that identifies words and phrases that lead to strong or weak perceived levels of credibility on Twitter.  Their findings suggest that the words of millions of people on social media have considerable information about an event’s credibility – even when an event is still ongoing.

“There have been many studies about social media credibility in recent years, but very little is known about what types of words or phrases create credibility perceptions during rapidly unfolding events,” said Tanushree Mitra, the Georgia Tech Ph.D. candidate who led the research.

The team looked at tweets surrounding events in 2014 and 2015, including the emergence of Ebola in West Africa, the Charlie Hebdo attack in Paris and the death of Eric Garner in New York City. They asked people to judge the posts on their credibility (from “certainly accurate” to “certainly inaccurate”). Then the team fed the words into a model that split them into 15 different linguistic categories. The classifications included positive and negative emotions, hedges and boosters, and anxiety.

Read More

Likelihood of Dieting Success Lies Within Your Tweets

There is a direct link between a person’s attitude on social media and the likelihood that their dieting efforts will succeed.

In fact, Georgia Institute of Technology researchers have determined that dieting success ­– or failure – can be predicted with an accuracy rate of 77 percent based on the sentiment of the words and phrases one uses on Twitter.

“We see that those who are more successful at sticking to their daily dieting goals express more positive sentiments and have a greater sense of achievement in their social interactions,” said Assistant Professor Munmun De Choudhury, who is lead researcher on the project. “They are focused on the future, generally more social and have larger social networks.”

Read More

Most of Facebook is ‘Friends Only,’ But Public and Private Posts are Likely Similar

Social media content, while driving a sizable portion of today’s web traffic, is not all public, and according to a new study, about 75 percent of Facebook posts, or three in four, are shared only with friends or subsets of friends. This translates into billions of daily online conversations that are seen by only a few.

Researchers from the Georgia Institute of Technology enlisted almost 2,000 Facebook users – who shared their most recent posts – and used machine learning methods as well as qualitative hand coding to determine content types and topics for roughly 11,000 public and private posts. They analyzed patterns of choices for privacy settings and found, contrary to expectations, that content type is not a significant predictor of privacy settings. They did find however that some demographics such as gender and age are predictive, suggesting that privacy choices may be driven more by the attributes of the person rather than by the content of the posts.

 

Georgia Tech Papers

 

A Parsimonious Language Model of Social Media Credibility Across Disparate Events

Social media has increasingly become central to the way billions of people experience news and events, often bypassing journalists—the traditional gatekeepers of breaking news. Naturally, this casts doubt on the credibility of information found on social media. Here we ask: Can the language captured in unfolding Twitter events provide information about the event’s credibility? By examining the first large-scale, systematically-tracked credibility corpus of public Twitter messages (66M messages corresponding to 1,377 real-world events over a span of three months), and identifying 15 theoretically grounded linguistic dimensions, we present a parsimonious model that maps language cues to perceived levels of credibility. While not deployable as a standalone model for credibility assessment at present, our results show that certain linguistic categories and their associated phrases are strong predictors surrounding disparate social media events. In other words, the language used by millions of people on Twitter has considerable information about an event’s credibility. For example, hedge words and positive emotion words are associated with lower credibility.

Tanushree Mitra 
Georgia Institute of Technology
Graham Wright
Georgia Institute of Technology
Eric Gilbert
Georgia Institute of Technology

 

Computational Approaches Toward Integrating Quantified Self Sensing and Social Media

The growing amount of data collected by quantified self tools and social media hold great potential for applications in personalized medicine. Whereas the first includes health-related physiological signals, the latter provides insights into a user’s behavior. However, the two sources of data have largely been studied in isolation. We analyze public data from users who have chosen to connect their MyFitnessPal and Twitter accounts. We show that a user’s diet compliance success, measured via their self-logged food diaries, can be predicted using features derived from social media: linguistic, activity, and social capital. We find that users with more positive affect and a larger social network are more successful in succeeding in their dietary goals. Using a Granger causality methodology, we also show that social media can help predict daily changes in diet compliance success or failure with an accuracy of 77%, that improves over baseline techniques by 17%. We discuss the implications of our work in the design of improved health interventions for behavior change.

Munmun De Choudhury
Georgia Institute of Technology
Mrinal Kumar
Georgia Institute of Technology
Ingmar Weber
Qatar Computing Research Institute

 

 

Defining Digital Self-Harm (Best Paper)——————————————

Self-harm is the infliction of pain or injury onto oneself. Though historically these behaviors were relegated to the fringes of communities, information technology now enables new ways to foster and encourage these dangerous activities. This paper defines the concept of digital self-harm asξthe online communication and activity that leads to, supports, or exacerbates, non-suicidal yet intentional harm or impairment of an individual’s physical wellbeing. We outline a research agenda for the CSCW community to understand the correlation and possible causation of offline self-harm behaviors due to online activities, and to design and assess technologies focused on prevention, mitigation and treatment. CAUTION: This paper includes media that could potentially be triggering to those dealing with an eating disorder or with other self-harm related illnesses. Please use caution when reading or disseminating this paper.

Jessica A. Pater
Georgia Institute of Technology
Elizabeth D. Mynatt
Georgia Institute of Technology

 


Gender and Cross-Cultural Differences in Social Media Disclosures of Mental Illness (Best Paper)

Cultural and gender norms shape how mental illness and therapy are perceived. However, there is a paucity of adequate empirical evidence around gender and cultural dimensions of mental illness. In this paper we situate social media as a “lens” to examine these dimensions. We focus on a large dataset of individuals who self-disclose to have an underlying mental health concern on Twitter. Having identified genuine disclosures in this data via semi-supervised learning, we examine differences in their posts, as measured via linguistic attributes and topic models. Our findings reveal significant differences between the content shared by female and male users, and by users from two western and two majority world countries. Males express higher negativity and lower desire for social support, whereas majority world users demonstrate more inhibition in their expression. We discuss the implications of our work in providing insights into the relationship of gender and culture with mental health, and in the design of gender and culture-aware health interventions.

Munmun De Choudhury
Georgia Institute of Technology
Sanket Sushil Sharma
Georgia Institute of Technology
Tomaz Logar
United Nations Global Pulse
Wouter Eekhout
Leiden University
Ren Clausen Nielsen
United Nations Global Pulse

 

Growing Their Own: Legitimate Peripheral Participation for Computational Learning in an Online Fandom Community

Online communities dedicated to the creation of fanworks (e.g., fiction or art inspired by media such as books or television shows) often serve as communities of practice for learning communication, artistic, and technical skills. In studying one successful fan fiction archive that was designed and built entirely by (predominantly women) fans, we observed processes of legitimate peripheral participation (LPP) in which some of these fans began in peripheral roles and came to be more involved in the technical aspects of the archive over time. In addition to outlining positive outcomes, we discuss the challenges of supporting learning within this CoP, particularly with respect to the burden on experts. We discuss potential implications and solutions for the problem of expert scarcity in CoPs, and propose that LPP within fan communities can be leveraged for broadening participation in computing among women.

Casey Fiesler
University of Colorado Boulder
Shannon Morrison
Syracuse University
R. Benjamin Shapiro
University of Colorado Boulder
Amy S. Bruckman
Georgia Institute of Technology

 

Mobile Collaboration for Human and Canine Police Explosive Detection Teams

We designed a communication system for law enforcement officers to use when conducting explosive detection searches with multiple agencies. Dogs trained in explosive detection work alongside human handlers to form a K9 team, which are an integral part of these searches. Officers in K9 teams have a strong bond and communication with these dogs, but noisy locations, long distances, and crowded spaces present challenges. In addition, other officers assigned as backup often lack the experience to read the cues from the canine, which hinders the speed and effectiveness of the team. Coordinating a search with teams from different municipalities presents challenges due to a lack of standard collaboration tools. Getting the right information as quickly as possible saves lives, whether this information is about the areas that have been searched or the location of an explosive device. We hope that in addition to increasing public safety, our system will make working conditions safer for law enforcement officers and their canines.

Joelle Alcaidinho
Georgia Institute of Technology
Larry Freil
Georgia Institute of Technology
Taylor Kelly
Georgia Institute of Technology
Kayla Marland
Georgia Institute of Technology
Chunhui Wu
Georgia Institute of Technology
Bradley Wittenbrook
Georgia Institute of Technology
Giancarlo Valentin
Georgia Institute of Technology
Melody Moore Jackson
Georgia Institute of Technology

 

What (or Who) Is Public? Privacy Settings and Social Media Content Sharing

When social networking sites give users granular control over their privacy settings, the result is that some content across the site is public and some is not. How might this content-or characteristics of users who post publicly versus to a limited audience-be different? If these differences exist, research studies of public content could potentially be introducing systematic bias. Via Mechanical Turk, we asked 1,815 Facebook users to share recent posts. Using qualitative coding and quantitative measures, we characterize and categorize the nature of the content. Using machine learning techniques, we analyze patterns of choices for privacy settings. Contrary to expectations, we find that content type is not a significant predictor of privacy setting; however, some demographics such as gender and age are predictive. Additionally, with consent of participants, we provide a dataset of nearly 9,000 public and non-public Facebook posts.

Casey Fiesler
University of Colorado Boulder
Michaelanne M. Dye
Georgia Institute of Technology
Jessica L. Feuston
Northwestern University
Chaya Hiruncharoenvate
Georgia Institute of Technology
C.J. Hutto
Georgia Institute of Technology
Parisa Khanipour Roshan
Georgia Institute of Technology
Umashanthi Pavalanathan
Georgia Institute of Technology
Amy S. Bruckman
Georgia Institute of Technology
Munmun De Choudhury
Georgia Institute of Technology
Eric Gilbert
Georgia Institute of Technology

 

When the Internet Goes Down in Bangladesh

We present a study of internet use and its forced non-use in Bangladesh. In light of current initiatives on state and industry actors to improve internet access and bridge the ‘digital divide’ for underserved, under-resourced, and underrepresented communities across the world, we offer a situated, qualitative perspective on what the current state of internet use looks like for select social groups in Bangladesh. We analyze how a state-imposed ban attempted to effect the nonuse of particular web-based services and how the affected populations found or did not find workarounds in response. We also discuss takeaways for researchers as well as industry and state actors studying and working towards more equitable access to the internet in the ‘developing’ world.

Mehrab Bin Morshed
Georgia Institute of Technology
Michaelanne M. Dye
Georgia Institute of Technology
Syed Ishtiaque Ahmed
Cornell University
Neha Kumar
Georgia Institute of Technology