Computing tie strength

Relationships make social media social. But, not all relationships are created equal. We have colleagues with whom we correspond intensely, but not deeply; we have childhood friends we consider close, even if we fell out of touch. Social media, however, treats everybody the same: someone is either a completely trusted friend or a total stranger, with little or nothing in between. In reality, relationships fall everywhere along this spectrum, a topic social science has investigated for decades under the name tie strength, a term for the strength of a relationship between two people. Despite many compelling findings along this line of research, social media does not incorporate tie strength or its lessons. Neither does most research on large-scale social phenomena. In social network analyses, a link either exists or not. Relationships have few properties of their own. Simply put, we do not understand a basic property of relationships expressed online. This dissertation addresses this problem, merging the theories behind tie strength with the data from social media. I show how to reconstruct tie strength from digital traces in online social media, and how to apply it as a tool in design and analysis. Specifically, this dissertation makes three contributions. First, it offers a rich, high-accuracy and general way to reconstruct tie strength from digital traces, traces like recency and a message’s emotional content. For example, the model can split users into strong and weak ties with nearly 89% accuracy. I argue that it also offers us a chance to rethink many of social media’s most fundamental design elements. Next, I showcase an example of how we can redesign social media using tie strength: a Twitter application open to anyone on the internet which puts tie strength at the heart of its design. Through this application, called We Meddle, I show that the tie strength model generalizes to a new online community, and that it can solve real people’s practical problems with social media. Finally, I demonstrate that modeling tie strength is an important new tool for analyzing large-scale social phenomena. Specifically, I show that real-life diffusion in online networks depends on tie strength (i.e., it depends on social relationships). As a body of work, diffusion studies make a big simplifying assumption: simple stochastic rules govern person-to-person transmission. How does a disease spread? With constant probability. How does a chain letter diffuse? As a branching process. I present a case where this simplifying assumption does not hold. The results challenge the macroscopic diffusion properties in today’s literature, and they hint at a nest of complexity below a placid stochastic surface. It may be fair to see this dissertation as linking the online to the offline; that is, it connects the traces we leave in social media to how we feel about relationships in real life.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Computer Science

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2011

Author dc:creator

• Gilbert, Eric E.

Contributors dc:contributor

• Karahalios, Karrie G.
• Bailey, Brian P.
• Zhai, ChengXiang
• Sandvig, Christian E.
• Terveen, Loren
• Grudin, Jonathan

• tie strength
• social media
• social networks, computer-mediated communication
• collapsed contexts
• streams
• diffusion

Statement dc:rights

• Copyright 2010 Eric Edmund Gilbert

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/18285

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/18285