DERI Social Network Webcamp
From Technolotics
Notes - Webcamp - Social Networks - NUIG, 07/03/07
- Due to spam this is a closed mediawiki server, but if you'd like to contribute directly, mail me and I'll create a login for you.
- These are my personal notes, and not in any way a substitute for official available available at the webcamp site
Topics, tags and trends in the blogosphere
- Looked at (inter)relationship between blogs, and their topics
- sought topic related interlink structure
- 7k blog data set
- Muhammad cartoon controversy tracked
- Blogosphere user centric vs topic centric Usenet
- Clutters of topic sharing blogs are short lived
- Strong clusters hold users
- but users drift from topics over time
- Extreme topic drift increases user falloff
- Tags poor method of tracking blog post content
- Concept clustering better
- A list blogs more similar to one another
- more topic focused, consistent - authorities
- tags chosen for comprehensibility
- tight networks of clusters
- C list blogs have higher user entropy
- little or no inward linking
Notes
How were A list bloggers defined?
What about Maven Bloggers / influential but narrow popularity - Malcolm Gladwell's Idea articulated in 'The Tipping Point'
Does user topic drift account for primacy of link aggregator blogs / communities
Collecting community wisdom: integrating social search and social navigation
2 Aspects - social browsing and social search, combined to augment search of Association for Computing Machinery Digital Library (popular computer science pub database - with browsing and search navigation)
KnowledgeSea - social browsing
- Collates anonymized browsing history
- Provides the capacity for digg like ratings and annotation
iSpy - social search
- Reranks results based on previous keyword searches by user group
- Icon based visual cues on rating etc
Combined and integrated both systems - increased search efficacy
Notes
where's the hook for students to build this system - are legal restrictions keeping the content closed ?
would it not be better to use off the shelf digg / del.icio.us like solutions ? alternately light weight sem web client on top of existing networks of databases ?
Quirky mystery meat interface - why do academic database sits have to be so unusable
Annotations could be used to increase utility of abstract
No information provided on stat significance of results
- Control group browsing history is artificially useful seed for social group
- How to compare trained systems fairly?
The demand for search in a social network
- Founders of Bigulu - Bebo search, 2 Phd students
- May expand to other networks
- Crawls Bebo to allow superior profile search based on keyword, age, gender, location
- Estimate Bebo at 7m users not 25m as claimed !
- Slow linear growth of Bigalu
- Sem web API would aid search and data migration
- New features - inc user popularity (like Facebook radar)
- Inbound links as social distance cue to search
- limitations on social scraping due to private profiles etc
Notes
What is user persistence like?
Could they be underestimating Bebo network due to islands (isolated nodes or clusters - social darknets) ?
Is 'low attention span' really a function of high efficacy expectation
Are assumptions about viral growth of Bebo correct? Implied signup active network discrepancy
Where is the knowledge: reflections on social networking in corporate environments
- (Virtual) ethnographic study of collaborative work practices
- IBM Dublin software lab
- Technology as tool inseparable from culture
- technofetishism (sysadmin) vs fluffy-bunnies (people persons)
- tools should be tailored to users
- human guides to knowledge repositories vital
- social networks facilitate knowledge aquisition
- Lotus Messenger tool
- IM - integrates: blog, wiki, org chart, social bookmarks, presence, contact details
- Used to request ad-hoc meeting attendance
- Used to reduce steps to data acquisition
- Documents can link to individuals
- Real time back channel in meetings
- Data Retention
- Legal and practical requirements to archive data
- creates and enforces etiquette
- no interdiction on personal contact - social honor enforces
Successful campaigns on the Bebo social network
- Small list of clients - A list Irish sites
- Daft, Eircom, RTE, Bebo etc
- Sell to ad agencies and brands
- Internet only 1% Irish advertising market
- UK internet ad budget = 2X Irish ALL media budget
- Ireland € 24 per user, UK € 80 per user
- Main income to Bebo (other than VC) = ads
- All major Irish net spending brands advertise
- Brand budget (overall) - 140k - 600k (top ten)
- Some brands shy away from user content areas - use frontpage etc
- Bebo top site for under 17 in Ireland (comscore)
- Average visit 75mins ! (of 3 hour session)
- 95% reach
- drops off list of top sites in older categories
- Use by corps like Mcdonalds under pressure not to kidvertise on TV
- Relevance, customer interaction / relationship to brand
- Options
- Sidebar ads
- Branded profiles (eg: Disney Cars)
- Branded Bebo mail window
- Clickable in video advertisement tickers
- Texter widget - Moco - € 2 a week
- Anti sponsor brand comments / groups (eg: killer coke) would probably be removed
Notes
Ethics of Bebo spread? Via address book spam?
Growth of Facebook a threat?
Lack of direct access in school has not limited popularity
Corruption of utility of social media in 'brand interaction'
Interesting to note bebo added comment moderation not in response to cyber bullying but concerns of Coke - insistence on stifling branded profile criticism
Texter widget is misleading
- initially video states that service is free
- only mentions charges towards end of video
Keynote: Social network analysis: 1987-2007
Any kind of network can be mathematically mapped
- from social networks to computer networks
- social / organisational network analysis
- Inflow software
'Highly between person' connects networks
- analogous to Gladwells superconnector
David Krackhardt - organisational structure related to power
- Hierarchy, network position, and network knowledge - knowing the map
Six Degrees of Separation = Golden myth
- original study flawed (response bias)
- Noah Friedkin - UC Santa Barba - Social networks have a horizon
- 2 step clarity - blind after 4 connections max
- outer reaches of network are invisible
Mark Granovetter - one or at most 2 intermediaries useful
- in context of information (job) seeking
- direct ties most useful of all
IBM and Valdis Krebs - adaptive organisations have higher average reach
- data acquired through survey and interview, document sharing etc
Pharmaceutical firms market to opinion leaders
- use social network analysis to target doctors who influence drug prescription of contemporaries
Formalising of common interest networks
- Bounded vs unbounded (not task / job delimited) networks
Network weaving = analyzing networks to build communities
9/11 hijacker network - built on media published data
- conclusion - project teams are similar
- task focused group structures do not vary based on task
False positives can emerge from social network analysis
- limited inferences can be reasonably drawn
Infrastructures are built on effectiveness not resilience
- useful for network attacks
Notes
Evolution of social networks over time? - differ based on organisation evolution? - loose voluntary network migration
Richness of less visible ties?
Does similarity of group structure apply cross culturally?
Group Discussion - Future of social networks
Existing Networks
- tribe, livejournal -> myspace -> bebo -> facebook
Do super connectors lead network migration / adoption?
Do teens socialise on social networks, or just mirror existing real world acquaintance?
- do users actively resist socialising
- are there different socialisation patterns culturally, network determined
- Book - republic.com - reinforcement of preexisting connections
- social networks not scale free
Social networks as dating substitute - exhibitionism
- display, personality branding
- rating and identity definition
- utility of maintenance of connections over time
- forking elitist / domain specific networks
Social networks -> open id -> light sem web clients -> device based ad hoc attentional aggrigators
- carrying groups between networks
- open attentional data
- open identity data
- balance of privacy, security, reputation and freedom
Hardware versions of my blog log / twitter
- mobile ubiquitous anonymised positional aggregation
- variety of possible consequences
- loss of privacy vs great utility
- branded flash mobs - ala Dublin's black parade
- social network manipulation and politicisation
- SNA used to infer guilt by association / track dissidents
- pattern matching by employers
Comparative safety of blogging - distance from 'online friends'
Exaggerated fears of predation
Reputation as currency
