Information integrity from source via retweets

Lamplight Research Direction Proposal

Agenda

Problem Statement
Hypothesis
Flow
Information Cascade Tree
Concept of Speed, Scale, Range
Questions
Next Steps

Problem Statement

How does the integrity of

entertainment, political & technology

topical information on Twitter evolve via

retweets and comments?

Hypothesis

*Dependent on the data

*Depends on the experimental design

Information diverge as they travel through the networks

Preservers

- maintain the original intent, context, and content

- a continuum between preserving content and meaning of a tweet

- e.g. @DanFan: I shorten the words, del unnecessary [puctuation,] ... but don't change meaning or attribution

Adapters

- willing to remove various parts of the tweet to suit their own purposes

- e.g. 1. remove some or all of the original tweeter's comment, leaving just the URL

- e.g. 2. Write their own text or paraphrase the original tweet

- e.g. 3. truncates original msg to make it fit ( - context)

Reasons for Info Loss/Gain

Boyd D. et al. "Tweet, Tweet, Retweet: Conversational Aspects of Retweeting on Twitter" (2010, IEEE)

Twitter etiquette

For all retweets, follow one the following formats

1. RT @user URL

2. retweet @

3. via @

Implications of Info Loss/Gain

1. Change in intellectual ownership of the substantive content of the message

2. Meanings of the original message changes as they spread across the network

Flow

Current Topical Data from Lamplight

(Star Wars, Uber, Paris Attack, US Election)

Identify hops

Investigate integrity of diffused info from Seed

Seed

Social Influence

Way of Transmission

Topic

Global influence: celebrity, epidemic

Local influence: immediate circle of followers

Retweet (RT)

RT + Comment (RT + C)

Entertainment

Politics

News

Technology

Method

2 * 2 * 4 matrix comparative study

User graph: scale and range of a seed

Evaluate integrity of tweet at node (n) against seed or source

Info Cascade Tree

Speed

Range

Scale

Integrity of Info

RT + C

Questions

Does Lamplight's current nlp algorithm cover the ability to differentiate the tiers of users?

-- construction of the user graphs (seed and nodes)

Does lamplight's current algo include a way to verify and present a confidence level of the integrity of the information?

-- assessment of the integrity of information between nodes and between node (n) and seed

Next Steps

With lamplight topical data, start to form

1. User tree

2. Hypothesis around information integrity and social influencers (Global and Local)

3. Ways to measure integrity maintained and sentiment change