Introduction
In the mid-2000s, the "Social Web" was characterized by isolated "walled gardens." Identities on platforms like LiveJournal, Tribe, and LinkedIn were technically distinct, creating fragmented digital personas. This 2008 paper from the AAAI Conference on Artificial Intelligence explores a Semantic Web solution: using the Friend of a Friend (FOAF) vocabulary to break down these social silos. By applying logic-based reasoning to distributed data, the researchers aimed to determine if the FOAF ecosystem could create a unified global social graph.
Methods
The researchers employed a three-step methodology to aggregate distributed social data:
- Data Collection: They accessed member lists from 11 social networks natively supporting FOAF, including LiveJournal, Buzznet, and Advogato.
- Semantic Reasoning: Using the Web Ontology Language (OWL), they leveraged the
owl:InverseFunctionalProperty. In FOAF, unique identifiers like email hashes (mbox), homepages, or chat IDs allow a reasoner to logically infer that two different profiles represent the same individual. - Custom Reasoner: To process approximately 4 million profiles, a task-specific OWL reasoner was developed. This tool focused on identity merging and
foaf:knowsrelationships while maintaining provenance tracking to identify the source network for each friendship assertion.
Datasets and Network Characteristics
The study analyzed 11 active social networking sites, ranging from large blogging platforms to niche professional communities. The total sample size reached roughly 3.8 million users.
As shown in the table above, the networks varied significantly in scale and purpose. LiveJournal was the largest component, while sites like Advogato and eCademy focused on business networking.
Key Insights: The Power of Hubs
The analysis revealed that a small group of multi-platform users—approximately 8,000 unique individuals—acts as "hubs" connecting otherwise disjointed social graphs. These users are not only tech-savvy but also more socially active, tending to have more friends who are also likely to maintain multiple accounts.
The impact of these hubs is most evident in the "Small World" effect. By merging profiles, the Average Shortest Path Length (APL) between users across different networks decreased. For instance, on Buzznet, the APL dropped from 4.43 to 2.76 after reasoning, demonstrating how cross-platform links significantly "shrink" the digital distance between communities.
Results: A Network of Networks
The study proved that no network was an island; every site studied had direct connections to at least four others.
Key findings include:
- Successful Merges: The reasoner identified 8,047 unique people holding 16,295 total accounts across the 11 networks.
- Friendship Overlap: Among pairs of friends who both held multiple accounts, 15.71% were connected on more than one network. On average, if two people were friends and shared multiple platforms, they linked on 57.75% of those platforms.
- The LiveJournal Bottleneck: LiveJournal's optional email field limited merging potential. Only 8.8% of its users provided a unique identifier, suggesting that cross-network connectivity would be an order of magnitude higher with stricter identity requirements.
Limitations and Ethical Considerations
While successful, the study relied on strict "inverse functional properties" (exact matches). Fuzzy matching (names or interests) would likely reveal more connections but increase false positives. Furthermore, the researchers noted significant privacy concerns. The "Plink" case study—a site forced to shut down due to user backlash over data centralization—highlights the ongoing tension between data utility and individual privacy in the Semantic Web era. Today, the trend toward proprietary data silos has only intensified, making the open vision of FOAF more challenging to realize.
Summary updated