Basic Research

Human categorization
In order to understand the cognitive processes underlying how we make sense of information I conducted a number of experiments examining human categorization. While most studies of categorization have focused on concepts defined by simple features, I examined categories based on the relations between features. For example, understanding that water flowing through pipes is in some ways the same as electricity flowing through wires requires putting both into the same category despite water and electricity sharing few surface features -- instead, the key is the relation they share. Through a number of empirical experiments I’ve shown that such relational categories are learned, used, and remembered differently from traditionally studied feature-based concepts. For example, some of the most robust findings in the last thirty years of categorization research -- such as typicality, central tendencies, and family resemblance structures -- appear to be very different in relational categories. I’ve also used Bayesian statistical modeling to quantitatively capture these differences and to demonstrate the computational constraints needed to model them. These experiments suggest that we may need to reconsider how we categorize and organize information in the real world.

Human memory modeling
The flip side of encoding information is the ability to retrieve it. Understanding how the human memory system retrieves information could have important implications for computer-based information environments. Starting from a rational statistical analysis of human memory, I developed a simple yet powerful computational model that captures a wide range of human retrieval dynamics. In addition to accounting for many different, sometimes unintuitive memory dynamics, the model can easily be implemented as part of a larger cognitive or computational architecture. Current applications include cognitive modeling for interface design, algorithms for server caching, and context-aware retrieval.

Applied Research

Social collaborative systems
Recently the web has seen massive growth in the use of distributed collaboration in developing large knowledge bases, with Wikipedia proving one of the biggest success stories for online collaborative information environments. However, as users and information are added to the system, conflict and coordination costs inevitably grow. Combining statistical, machine learning, and visualization techniques, I characterized the growth of conflict and coordination costs in Wikipedia and developed tools to help deal with conflict at the global, article, and user levels. I also characterized the evolution of user influence on Wikipedia and del.icio.us, finding that for both there has been a decline in the power of elite users and a corresponding rise of novice user influence. I continue to investigate how large scale collaborative dynamics can be harnessed for optimal benefit by smaller collaborative groups.

Collaborative ontology building
The rapid growth of medical and neuroscientific data require structured ontologies which can help people find and make use of large data stores. However, traditional bioinformatics approaches to ontology building are insufficient for many domains where rapid advances mean that knowledge is both dynamic and distributed among many sources. Other drawbacks include high participation costs, limited longevity, and adoption issues. In contrast, I am developing a system for ontology building that combines the bazaar-like aspects of emerging collaborative online paradigms such as Wikipedia with cathedral-like support for structured data. The goal of this research is to enable researchers to make robust cross-disciplinary inferences (such as tying together cognitive, behavioral, and neuroimaging data) in a system that brings together knowledge from many researchers while maintaining low participation costs.

Literature discovery
The recent rapid advancement of science has been accompanied by an explosion in the accompanying scientific literature. Nowadays to make a significant advance it seems necessary to study a single area of science for many years; yet some of the most important discoveries are made by bridging between fields. In order to help people more quickly understand a new field I developed a web application which spiders and maps a scientific literature information space. With the program automatically aggregating thousands of citation links in meaningful ways, a user can quickly make sense of the literature space around a given article. A novel contribution of the tool is its algorithm for using both global and local relational space metrics to find highly relevant content. Built using a python web framework, users can easily perform iterative querying online.

Optimizing personal information retrieval on the web
The amount of information available online is rapidly growing, with many companies and search engines focusing on getting even more information to people. However, what to do with that information once it has been found remains an open question. Popular online strategies for storing and categorizing content include tagging (e.g., on del.icio.us) and classification (e.g., bookmarking). I am currently running experiments which directly compare tagging, classification, and search to understand the cognitive costs and benefits of each at encoding and retrieval, and to determine in what situations each strategy is most appropriate. The experimental platform also allows for controlled benchmarking and comparison of novel and hybrid strategies.

Publications

Kittur, A., Chi, E., Suh, B. (in prep). Can You Ever Trust a Wiki? How Article Stability Impacts Perceived Trustworthiness in Wikipedia.

Kittur, A., Hummel, J. E., Holyoak, K. J. (in prep). Feature-Based vs. Relational Category Learning: A Dual Process View.

Kittur, A., Chi, E., Suh, B. (2008). Crowdsourcing User Studies With Mechanical Turk. CHI 2007: Proceedings of the ACM Conference on Human-factors in Computing Systems. Florence: ACM Press

Suh, B., Chi, E., Kittur, A., Pendleton, B. (2008). Lifting the Veil: Improving Accountability and Social Transparency in Wikipedia with WikiDashboard.CHI 2007: Proceedings of the ACM Conference on Human-factors in Computing Systems. Florence: ACM Press

Suh, B., Chi, E., Pendleton, B. A., Kittur, A. (2007). Us vs. Them: Understanding Social Dynamics in Wikipedia with Revert Graph Visualizations. VAST 2007: IEEE Symposium on Visual Analytics Science and Technology.

Kittur, A., Suh, B., Pendleton, B. A., Chi., E. (2007). He Says, She Says: Conflict and Coordination in Wikipedia. CHI 2007: Proceedings of the ACM Conference on Human-factors in Computing Systems. San Jose: ACM Press

Kittur, A., Chi, E., Pendleton, B. A., Suh, B., Mytkowicz, T. (2007). Power of the Few vs. Wisdom of the Crowd: Wikipedia and the Rise of the Bourgeoisie. Alt.CHI 2007

Kittur, A., Holyoak, K. J., & Hummel, J. E. (2006). Using Ideal Observers in Higher-order Human Category Learning. Proceedings of the Twenty Eighth Annual Meeting of the Cognitive Science Society. Vancouver, CA.

Kittur, A., Hummel, J. E., & Holyoak, K. J. (2006). Ideals Aren’t Always Typical: Dissociating Goodness-of-Exemplar From Typicality Judgments. Proceedings of the Twenty Eighth Annual Meeting of the Cognitive Science Society. Vancouver, CA.

Green, C., & Kittur, A. (2006). Retrieval-Induced Forgetting in a Multiple-Trace Memory Model. Proceedings of the Twenty Eighth Annual Meeting of the Cognitive Science Society. Vancouver, CA.

Kittur, A., Hummel, J.E., & Holyoak, K.J. (2004). Feature- vs. Relation-Defined Categories: Probab(alistic)ly Not the Same. Proceedings of the Twenty Sixth Annual Meeting of the Cognitive Science Society. Chicago, IL.

Green, C., & Kittur, A. (2004). A Multiple-Trace Memory Model Exhibiting Realistic Retrieval Dynamics. Proceedings of the Twenty Sixth Annual Meeting of the Cognitive Science Society. Chicago, IL.

Hummel, J.E., Holyoak, K.J., Green, C., Doumas, L.A.A., Devnich, D., Kittur, A., & Kalar, D.J. (2004). A Solution to the Binding Problem for Compositional Connectionism. In S.D. Levy & R. Gayler: Compositional Connectionism in Cognitive Science: Papers from the AAAI Fall Symposium. Menlo Park, CA: AAAI Press.

Lee, M. K., Borchelt, D. R., Kim, G., Thinakaran, G., Slunt, H. H., Ratovitski, T., Martin, L. J., Kittur, A., Gandy, S., Levey, A. I., Jenkins, N., Copeland, N., Price, D. L., & Sisodia, S. S. (1997). Hyperaccumulation of FAD-linked presenilin 1 variants in vivo. Nature Medicine, 3, 756-60.