Anna Osepayshvili
Contact me at |
I am currently splitting my time between studying energy markets as a Visiting Researcher at the Palo Alto Research Center (Embedded Reasoning Area) and working at Google as a contractor. I am on the job market for full-time (applied) research or teaching positions, preferably in South Bay Area, CA. My research interests include digitally mediated markets such as Internet auctions, information-good markets, and, more recently, electricity markets and smart-grid technologies. My research methodology is a combination of game-theoretic and computer-science techniques, human-subject experiments, and social-network and quantitative data analysis. I received a PhD in Information in 2009 from the University of Michigan School of Information with the focus on the Incentive-Centered Design (ICD). The ICD approach emphasizes the importance of individual incentives in the design of systems that rely on information, communication and collaboration technologies to mediate interactions. For more about the projects of the ICD group at the University of Michigan, see the ICD twiki page. I also earned an MA in Economics from the University of Michigan Department of Economics in 2006. Here is my complete PhD dissertation on strategic decision making in digitally mediated markets. You can also download separately Chapter 2 (Bundling Information Goods: A Study of Competing Firms Facing Heterogeneous Consumers) and Chapter 3 (Learning Bayesian Nash Equilibrium: An Experimental Study). Chapter 1, which is joint work with Michael P. Wellman, Jeffrey K. MacKie-Mason, and Daniel M. Reeves, was published in The B.E. Journal of Theoretical Economics in 2008. You can also download my dissertation oral-defense slides. My dissertation committee members were Professor Jeffrey K. MacKie-Mason (chair), Professor Yan Chen, Professor Michael P. Wellman, and Professor Rahul Sami. |
Quick Links: CV | Resume | Research Projects: (1) Auctions: SAAs | (2) Info Goods | (3) Experiment: Bayes Nash | (4) Soc. Networks: EBay | (5) Experiment: Network Externalities | Publications
CURRICULUM VITAE
I am currently on the job market for (applied) research or teaching positions, preferably in South Bay Area, CA. Click on a link below to view my CV.
 CV [html] |
 CV [pdf] |
RESEARCH PROJECTS
Below are some projects that describe my area of expertise and training background.| Back to top |
Research Question: How should one bid in simultaneous ascending auctions (SAAs) in which goods are complements or substitutes?
Motivation: Many real-world auctions are SAAs: eBay, FCC spectrum auctions, energy markets in many countries, business-to-business transactions, etc. Although SAAs are relatively simple to implement, they are strategically complex. Auction theory does not provide analytic solutions for optimal bidding strategies in non-trivial environments.
Method: Empirical game-theoretic methods, which combine game-theoretic and heuristic approaches to selecting bidding strategies. The methods involve agent-based modeling and computer simulation of auctions.
My key contributions: Development and analysis of the performance/profitability of price-predicting bidding strategies; programming a large part of the simulation and data-processing software.
Coauthors: Michael P. Wellman, Jeffrey K. MacKie-Mason, and Daniel M. Reeves.
| Bundling Strategies for Information Goods (Thesis, Chapter 2) | Back to top |
Research Question: Evaluate performance of various bundling strategies (pure bundling, pure unbundling, and mixed bundling) for two competing firms. Compare results to the case of a monopolistic producer.
Motivation: Classical marginal-cost pricing does not recover high first-copy costs of producing information goods, because their reproduction and distribution costs are negligible relative to first-copy costs. Many theoretical and empirical studies have shown that bundling is a promising alternative, especially for a monopolistic producer. I focus on the case of two firms and consumer preferences specific to (digital) information goods.
Method: I adopted the empirical game-theoretic methods (see above) to search for pricing strategies of the firms. Due to the complex structure of consumer preferences, the problem of finding equilibria appears to be analytically intractable.
My key contributions: Analysis of equilibrium prices, profits, social welfare, and efficiency; programming market-simulation and data-processing software.
Coauthors: Jeffrey K. MacKie-Mason and Scott A. Fay.
| Learning Bayesian Nash Equilibrium: An Experimental Study (Thesis, Chapter 3) | Back to top |
Research Question: In games of incomplete information, when do people converge to Bayesian Nash equilibrium (BNE)? What factors affect learning of BNE?
Motivation: BNE has been the main solution concept in games of incomplete information. In particular, mechanism-design theory relies on this concept (esp. Bayesian implementation in mechanism design). Whether people actually converge to BNE-play is important for practical applications of game theory.
Method: Human-subject laboratory experiment.
My key contributions: Design and software implementation of the experiment; recruitment of subjects and data collection; computer simulations of the dynamics of learning models; statistical data analysis.
Supervisors: Yan Chen and Jeffrey K. MacKie-Mason.
| A Networks Approach to Understanding Shilling on EBay (Course Project, 2007) | Back to top |
Research Question: Identify network characteristics of shillers on eBay.
Motivation: Reputation systems in many online markets are easy to manipulate. For example, reputation shillers (fraudsters or otherwise) acquire a good reputation in low-price markets in order to reap benefits in high-value transactions. Such systems can be made more robust to manipulation if shillers are automatically identified and flagged to warn potential buyers of inflated reputation scores.
Method: Social-network analysis of participants in low-price and high-price transactions on eBay.
Coauthors: Tapan Khopkar. Project Supervisor: Lada Adamic.
| Network Externalities: An Experimental Study (Course Project, 2004) | Back to top |
Research Questions: First, why are network economies populated by temporary monopolies? Is this due to network externalities or due to supply-side economies of scale or both? Second, when do firms become motivated to invest in compatibility devices? Finally, do firms facing network externalities engage in predatory pricing?
Motivation: Understanding network economies is important for designing effective regulation policies. For example, antitrust principles have to take into account the role of network externalities in the market share of firms and incentives to engage in anti-competitive practices. Similarly, compatibility affects social welfare, and therefore may require careful regulation if the incentives of the firms are not aligned with the public interest.
Method: Human-subject laboratory experiment.
Project Supervisor: Yan Chen.
PUBLICATIONS
Here's a list of my publications. Click on paper title for more details, including abstract and BibTex.
| BEJTE-08 |
Bidding Strategies for Simultaneous Ascending Auctions Michael P. Wellman, Anna V. Osepayshvili, Jeffrey K. MacKie-Mason, and Daniel M. Reeves In B.E. Journal of Theoretical Economics, 8(1). 2008. |
| |
| Bidding Strategies for Simultaneous Ascending Auctions: Appendix |
| UAI-05 |
Self-Confirming Price Prediction for Bidding in Simultaneous Ascending Auctions Anna V. Osepayshvili, Michael P. Wellman, Daniel M. Reeves, and Jeffrey K. MacKie-Mason In Proceedings of the 21st Conference on Uncertainty in Artificial Intelligence (UAI 2005). July 2005. |
|
| DSS-05 |
Exploring Bidding Strategies for Market-Based Scheduling Daniel M. Reeves, Michael P. Wellman, Jeffrey K. MacKie-Mason, and Anna V. Osepayshvili In Decision Support Systems. Pages 67--85. 2005. |
| ICAPS-04 |
Price Prediction Strategies for Market-Based Scheduling Jeffrey K. MacKie-Mason, Anna V. Osepayshvili, Daniel M. Reeves, and Michael P. Wellman In Proceedings of the Fourteenth International Conference on Automated Planning and Scheduling. 2004. |
|
Get Adobe Reader