Our simulation is based on Chris Starnes. original work by Reynolds [8] on the simulation of flocks of birds (or ‘Boids‘) in a manner not subject to the apparent combinatorial explosion of such calculations. We are extending the concept of flocking or clustering to be based on the data relationships (determined by textual comparison) between data points (our ‘Boids‘) rather than purely on adjacency of the Boids. We feel this simulation will be both intuitive and informative, as well as allowing for rich user interaction as Boids can be manually relocated (‘dragged around‘) by the user, and the simulation will react accordingly. Miguel Borromeo. Flock will take advantage of advanced computer graphics hardware and software. We will be using OpenGL to perform the rendering, and making use of OpenCL for parallelized Patrick Webster computation of data clustering and flocking. The implementation will consist of four core components: a storage and analysis module, the graphics engine, the flocking engine, and the user interface. The storage and analysis module defines data sets and storage facilities to easily find and represent facets about the information. The graphics engine will be tasked with rendering and computing the physical Nathan Clark. calculations necessary for our interactive environment. The flocking engine (with the aid of the clustering engine, part of the S&A module) will link Boids together to create ―flocks‖ of related data sets. We will explore what we call ‘implicit social networks,‘ in an effort to improve understanding of the interactions and relationships these networks present. While ‘explicit‘ networks are based on ‘friend lists,‘ bibliographies, and other such explicitly denoted user-user relationships, implicit networks are derived from the ‘activity networks‘ [13] of those users. Research has shown that full comprehension of a social network requires understanding these implicit links, as the explicit links rarely Luke Hersman. hold any correspondence to the actual strength of a relationship [3]. By mapping Twitter as an implicit social network, we will identify what aspects of a network correspond with relationships between Justin Kern. users, and be able to extend that knowledge to identifying corresponding relationships between topics, groups of users, and even individual tweets.

Miguel Borromeo

Chris Starnes

Nathan Clark

Luke Hersman

Justin Kern

Patrick Webster

Video

The problem is that this information is not easily apparent

Our goal was to create a way to easily interpret this information

Twitter was a prime choice for our project

A wealth of information can be gleaned from social web sites.

TwitterFlock

Problem

Design

MeritImplications

Impact

Problem

Initial Focus

Focused initially on technical aspects

Laid out component interactions in advance

Local store

Relationship engine

Flocking engine

GUI

Remote database

Application Components:

TwitterFlock

Problem

Design

MeritImplications

Impact

Evaluation

TwitterFlock

Problem

Design

MeritImplications

Impact

Flocking behavior is unclear

Which word influenced the flocking?

Interaction is limited and confusing

People just want to read the tweets

Evolution

Flocking behavior is unclear

Moved to a yes/no decision

Slowed down the simulation

Which word influenced the flocking?

Added glow lines between flocking tweets

Interaction is limited and confusing

Added tweet text box

Calculated and displayed the most meaningful word

People just want to read the tweets

TwitterFlock

Problem

Design

MeritImplications

Impact Added visual feedback when selecting a tweet

Added the ability to drag tweets

To See Profoundly

“There are some universal cognitive tasks that are deep and profound—indeed, so deep and profound that it is worthwhile to understand them in order to design our displays in accord with those tasks.” -Edward Tufte

Three levels of semantic metaphor

Depth rather than breadth of interaction

Simplicity vs. completeness

TwitterFlock

Problem

Design

MeritImplications

Impact

Sensemaking and Social Networks

Implications

How do we visualize data?

TwitterFlock

Problem

Design

MeritImplications

Impact

Numbers?

Graphs?

Charts?

Text?

This is easy, but what about relationships, semantics, and dynamic nature?

Implications

TwitterFlock

Problem

Design

MeritImplications

Impact

Used bird-based behavior to visualize Twitter content

Twitter has an inherent bird-theme

“Tweeting”, “following”, etc

Why not flocking?

Implications

TwitterFlock

Problem

Design

MeritImplications

Impact

Not limited to flocking:

What we’ve discovered:

Mapping behavior and content conveys the dynamic aspect of data well

Transcends making sense of numerically-based visualizations

Gravitation

Swarming

Broader Impact

Lack of tools to interpret data.

Meaning can be hidden through implicit connections.

Encourages the exploration of social networks.

May make it possible to create a more complete understanding of social networks and their interactions.

Easily expanded to any text-based data.

TwitterFlock

Problem

Design

MeritImplications

Impact

Questions?

TwitterFlock

Problem

Design

MeritImplications

Impact?