Lisa Singh

Social networks reveal cultural behaviour in tool-using dolphins

Animal tool use is of inherent interest given its relationship to intelligence, innovation and cultural behaviour. Here we investigate whether Shark Bay bottlenose dolphins that use marine sponges as hunting tools (spongers) are culturally distinct from other dolphins in the population based on the criteria that sponging is both socially learned and distinguishes between groups. We use social network analysis to determine social preferences among 36 spongers and 69 non-spongers sampled over a 22-year period while controlling for location, sex and matrilineal relatedness. Homophily (the tendency to associate with similar others) based on tool-using status was evident in every analysis, although maternal kinship, sex and location also contributed to social preference. Female spongers were more cliquish and preferentially associated with other spongers over non-spongers. Like humans who preferentially associate with others who share their subculture, tool-using dolphins prefer others like themselves, strongly suggesting that sponge tool-use is a cultural behaviour.

Can Friends Be Trusted? Exploring Privacy in Online Social Networks

In this paper, we present a case study describing the privacy and trust that exist within a small population of online social network users. We begin by formally characterizing different graphs in social network sites like Facebook. We then determine how often people are willing to divulge personal details to an unknown online user, an adversary. While most users in our sample did not share sensitive information when asked by an adversary, we found that more users were willing to divulge personal details to an adversary if there is a mutual friend connected to the adversary and the user. We then summarize the results and observations associated with this Facebook case study.

Visual analysis of dynamic group membership in temporal social networks

C-Group is a tool for analyzing dynamic group membership in temporal social networks over time. Unlike most network visualization tools, which show the group structure within an entire network, or the group membership for a single actor, C-Group allows users to focus their analysis on a pair of individuals. While C-Group allows for viewing the addition and deletion of nodes (actors) and edges (relationships) over time, its major contribution is its focus on changing group memberships over time. By doing so, users can investigate the context of temporal group memberships for the pair. C-Group provides users with a flexible interface for defining (and redefining) groups interactively, and supports two novel visual representations of the evolving group memberships. This flexibility gives users alternate views that are appropriate for different network sizes and provides users with different insights into the grouping behavior. We demonstrate the utility of the tool on a scientific publication network.

Visual Mining of Multi-Modal Social Networks at Different Abstraction Levels

Social networks continue to become more and more feature rich. Using local and global structural properties and descriptive attributes are necessary for more sophisticated social network analysis and support for visual mining tasks. While a number of visualization tools for social network applications have been developed, most of them are limited to uni-modal graph representations. Some of the tools support a wide range of visualization options, including interactive views. Others have better support for calculating structural graph properties such as the density of the graph or deploying traditional statistical social network analysis. We present Invenio, a new tool for visual mining of socials. Invenio integrates a wide range of interactive visualization options from Prefuse, with graph mining algorithm support from JUNG. While the integration expands the breadth of functionality within the core engine of the tool, our goal is to interactively explore multi-modal, multi-relational social networks. Invenio also supports construction of views using both database operations and basic graph mining operations.

Measuring Topological Anonymity in Social Networks

While privacy preservation of data mining approaches has been an important topic for a number of years, privacy of social network data is a relatively new area of interest. Previous research has shown that anonymization alone may not be sufficient for hiding identity information on certain real world data sets. In this paper, we focus on understanding the impact of network topology and node substructure on the level of anonymity present in the network. We present a new measure, topological anonymity, that quantifies the amount of privacy preserved in different topological structures. The measure uses a combination of known social network metrics and attempts to identify when node and edge inference breeches arise in these graphs.

Pruning social networks using structural properties and descriptive attributes

Scale is often an issue with understanding and making sense of large social networks. Here we investigate methods for pruning social networks by determining the most relevant relationships. We measure importance in terms of predictive accuracy on a set of target attributes of the social network. Our goal is to create a pruned network that models only the most informative affiliations and relationships. We present methods for pruning networks based on both structural properties and descriptive attributes demonstrate it on a network of NASDAQ and NYSE businesses and on a bibliographic network.

Visualizing node attribute uncertainty in graphs

Visualizations can potentially misrepresent information if they ignore or hide the uncertainty that are usually present in the data. While various techniques and tools exist for visualizing uncertainty in scientific visualizations, there are very few tools that primarily focus on visualizing uncertainty in graphs or network data. With the popularity of social networks and other data sets that are best represented by graphs, there is a pressing need for visualization systems to show uncertainty that are present in the data. This paper focuses on visualizing a particular type of uncertainty in graphs - we assume that nodes in a graph can have one or more attributes, and each of these attributes may have an uncertainty associated with it. Unlike previous efforts in visualizing node or edge uncertainty in graphs by changing the appearance of the nodes or edges, e.g. by blurring, the approach in this paper is to use the spatial layout of the graph to represent the uncertainty information. We describe a prototype tool that incorporates several uncertainty-to-spatial-layout mappings and describe a scenario showing how it might be used for a visual analysis task.

Exploring re-identification risks in public domains

While re-identification of sensitive data has been studied extensively, with the emergence of online social networks and the popularity of digital communications, the ability to use public data for re-identification has increased. This work begins by presenting two different cases studies for sensitive data re-identification. We conclude that targeted re-identification using traditional variables is not only possible, but fairly straightforward given the large amount of public data available. However, our first case study also indicates that large-scale re-identification is less likely. We then consider methods for agencies such as the Census Bureau to identify variables that cause individuals to be vulnerable without testing all combinations of variables. We show the effectiveness of different strategies on a Census Bureau data set and on a synthetic data set.

G-PARE: A visual analytic tool for comparative analysis of uncertain graphs

There are a growing number of machine learning algorithms which operate on graphs. Example applications for these algorithms include predicting which customers will recommend products to their friends in a viral marketing campaign using a customer network, predicting the topics of publications in a citation network, or predicting the political affiliations of people in a social network. It is important for an analyst to have tools to help compare the output of these machine learning algorithms. In this work, we present G-PARE, a visual analytic tool for comparing two uncertain graphs, where each uncertain graph is produced by a machine learning algorithm which outputs probabilities over node labels. G-PARE provides several different views which allow users to obtain a global overview of the algorithms output, as well as focused views that show subsets of nodes of interest. By providing an adaptive exploration environment, G-PARE guides the users to places in the graph where two algorithms predictions agree and places where they disagree. This enables the user to follow cascades of misclassifications by comparing the algorithms outcome with the ground truth. After describing the features of G-PARE, we illustrate its utility through several use cases based on networks from different domains.

Understanding actor loyalty to event-based groups in affiliation networks

In this paper, we introduce a method for analyzing the temporal dynamics of affiliation networks. We define affiliation groups which describe temporally related subsets of actors and describe an approach for exploring changing memberships in these affiliation groups over time. To model the dynamic behavior in these networks, we consider the concept of loyalty and introduce a measure that captures an actor’s loyalty to an affiliation group as the degree of ‘commitment’ an actor shows to the group over time. We evaluate our measure using three real world affiliation networks: a publication network, a senate bill cosponsorship network, and a dolphin network. The results show the utility of our measure for analyzing the dynamic behavior of actors and quantifying their loyalty to different time-varying affiliation groups.