Projects
Smarter Blogroll
Emergence and Abstraction
Scenario Visualization
Normative Echoes
EcoRaft
Synthetic Social Construction
Smarter Blogroll
This was a project that I did with Danyel Fisher during a summer internship with the Community Technologies Group at Microsoft Research. We started looking at interesting ways to link different bloggers together, considering a bipartite social network tool where one type of node was individual blogs and the other type of node was the topics they were discussing. Through the course of several discussions, paper mock-ups, and prototypes, we ended up with a somewhat different tool designed not for authors of blogs, but for readers of blogs.
Blogs often contain a blogroll, a sidebar with links to other blogs. Ostensibly, a blogroll is the list of other blogs that the author reads. However, when many blog rolls number in the hundreds, some linking to over one thousand blogs, it becomes clear that the blogroll serves functions other than the author's reading list. A blogroll can be a declaration of interests, a means of referring readers to related material, a list of social ties, or a statement of approval. However, blogrolls generally list only the names of other blogs, which can be quite inscrutable. For instance, what to Eschaton, Crooked Timber, and Apartment 11D all have in common? By the names, one would never guess that all three feature political discussion. Our approach was to scrape the RSS feeds for all the blogs on a user's blogroll, run a simple topic extraction algorithm over the content, and present the user with the Smarter Blogroll, wherein blog names are annotated by a list of up to the five most salient recent topics. By clicking on any of these topics, the list expands to show the titles of recent posts about that topic, the titles themselves being permalinks directly to the post. Thus, it becomes possible for the user of the blogroll to quickly scan through the list and get a general gist of what his or her blogs are talking about. Similarly, it enables a new visitor to a blog to get a general impression about the interests of the creator(s) of the blog, as well as quickly navigate through the blogroll to find posts about topics of interest.
Publications:
Baumer, E. and Fisher, D. Smarter Blogroll: An Exploration of Social Topic Extraction for Manageable Blogrolls Microsoft Research Technical Report MSR-TR-2007-18, Microsoft Research, Redmond, WA, 2007.
Emergence and Abstraction
This work centers around the study of emergent phenomena in social systems and the premise that, in social systems, the process of emergence is intrinsically linked with the process of abstraction. By emergence, we mean the process of beginning with simple rules and generating complex phenomena that emerge from those rules but are not explicitly described in those rules. By abstraction, we mean the process of beginning with a set of complex phenomena and creating a set of simple rules that describe the phenomena. Thus, we can see that emergence and abstraction move in opposite directions along a continuum of complexity. Moreover, in the case of societies, the two processes mutually inform one another. The abstractions an individual agent makes about the social interactions it observes and in which it participates will serve to guide its future interactions, thus influencing what patterns emerge. As the agent interacts with other members of the society, complex social structures emerge, which then serve to influence what abstractions the individual agent makes. Thus, there is a feedback cycle, an abstraction-emergence loop, that contributes to the formation of complex structures of social interaction.
Publications:
The Interconnected Roles of Abstraction and Emergence in Artificial Societies. AAAI Fall Symposium (AAAI FS-06-06).
We're also working on developing some social simulations based around study the role and effects of the abstraction-emergence loop, with the goal of being able to better understand and manipulate emergent phenomena.
Scenario Visualization
This was a collaboration between myself, my advisor, and Thomas Alspaugh, another Informatics professor who works in the area of requirements engineering. It came out of a series of discussions about how stories and narrative structure are similar to the scenarios used in requirements engineering. Using the codebase that my advisor has developed since his days at the media lab, we built a tool to visualize scenarios that describe the functioning of a software system as a social interaction among groups of animated characters. We did a small evaluation, showing that the technique helped users to find various sorts of errors in the scenarios more readily than the text of the scenarios alone.
Publications:
Using Social Agents to Visualive Software Scenarios. ACM Symposium on Software Visualization (ACM SOFTVIS 2006).
On a Mixed-Methods Evaluation of a Social-Agent Scenario Visualization. Fourth International Workshop on Comparative Evaluation in Requirements Engineering (CERE 2006).
Normative Echoes
Normative Echoes is an installation I developed on top of the functionality provided by Virtual Raft (below). It's an exploration of a lot of the ideas that I've been pursuing lately, especially the application of sociological theory to multiagent systems. Communities of animated autonomous agents communicate with one, while those communications are sonified so as to be comprehensible by human participants. As the agents interact with one another, patterns of communicative behavior emerge as norms within these communities. When one agent is transferred to another community, it takes with it all the norms and institutions it knows and attempts to use those to communicate in the new community. There is also a level of user interaction, in that participants may communicate with the agents via a microphone. The words and phrases uttered by the participant are learned and repeated by the agents, which then incorporate those utterances into their communications.
This installation explores a number of ideas. One of them is in the possibility for multiple interpretations to allow for user (re-)design. In computational systems, it is often the case that the meaning of a certain portion of the system is inherent to the system itself. That is, meaning is posited by the designers and implementers of the system. However, this ignores the fact that any design will be appropriated by its users. For example, the phone book on a teenager's cell phone has just as much functional utility as it does social capital in its ability to show to whom the individual is connected. Such uses were neither intended nor predicted for such a device. How can we as designers support and encourage this implicit redesign to allow for the creation of more useful and more meaningful artifacts? I argue that one was is to remove some of the inherent meaning from computational devices. In Normative Echoes, the utterances spoken by the agents do not, for them, carry any inherent meaning; any comes from the actions and interpretations of participants. In this way, participants implicitly take part in the design of the system and define for themselves its meaning.
Publications:
"Normative Echoes: Use and manipulation of player generated content by communities of NPCs." AI in Interactive Digital Entertainment (AIIDE 2006), Demonstration Program. (video)
The EcoRaft Project
EcoRaft Project Website
The EcoRaft Project is an interactive installation designed to teach children about restoration ecology. It uses much of the technology developed in the Virtual Raft Project for transferring of autonomous animated agents among networks of heterogeneous computational devices. However, rather than transferring humanoid characters, participants use the tablet PCs to move various plant and animal species between three islands with the goal of restoring the islands' ecosystems.
This project was shown in the Emerging Technologies venue at ACM SIGGRAPH 2005 and was temporarily deployed at the Discovery Science Center, a hands-on science museum for children of all ages, in Santa Ana, CA.
Publicaitons:
Virtual Raft Project. ACM SIGGRAPH 2005, Emerging Technologies.
Synthetic Social Construction: Norms and Institutions in Societies of Autonomous Agents
Some of my early research in grad school focused on applying ideas from social construction to multiagent systems.
In social construction, an individual experiences society both as an objective reality, in that society dictates certain norms and patterns to the individual, and as a subjective reality, in that the individual's actions work to definet the nature of the society of which he or she is a member. Thus, the individual enters into a dialectic process of mutual definition with society.
Current work is on using these concepts to implement a framework that allows agents in a multiagent system to engage in this dialectic process with the systems of which they are a part. When groups of agents form patterns of behaviors, and each agent in the group has a subjective interpretation of that behavior that corresponds with the subjective interpretation of the other agents, an institution is formed. These ideas helped in part to guide the development of the Normative Echoes installation.
Publications:
Institutionalization through Reciprocal Habitalization and Typification. Second NASA Workshop on Radical Agent Concepts (WRAC 2005).
Synthetic Social Construction for Autonomous Characters. Workshop on Modular Construction of Human-Like Intelligence (AAAI 2005).
back to main