 |
Home > Research > Research Experience for Undergraduates (REU) > Projects
SI REU Summer Program: Projects
Below are the research projects available during the summer 2009 REU program. You will choose three of the titles put in rank order as your preferred projects in your REU program application when the 2010 project listing is available.
Characterizing and Evaluating Online Question-Answer Communities
Professor Mark Ackerman
Computer Science and Engineering and School of Information, U-M
Professor Lada Adamic
School of Information, U-M
The goal of the project is to construct tools in order to understand online communities and create new interfaces for those communities. We'll be looking at online question-and-answer forums like webboards, Yahoo Answers, and the like.
One or two students will develop two pieces of software. First, they will develop Web crawlers for obtaining data. With assistance, they will design a general package to crawl and scrape web-based forums. These crawlers need to be able to handle large data volumes and be highly robust. Second, they will extend a simulator to model these online communities.
The students will learn about online communities, Web-based protocols, distributed systems, and user interfaces. The simulator is written in Java. The crawler package can be written in Java, Ruby, Python, or a mix of languages. The students will also be expected to become part of the research team, including attending team meetings. They may participate in publications.
Requirements: Junior standing in CSE or equivalent programming experience.
For more information about the project, please contact Kevin Nam at ksnam@umich.edu.
Social Identity and Mechanism Design
Professor Yan Chen
School of Information, U-M
Social identity is commonly defined as a person's sense of self derived from perceived membership in social groups. While standard economic analysis focuses on individual-level incentives in decision-making, group identity has been shown to be a central concept in understanding such phenomena as ethnic and racial conflicts, discrimination, political campaigns, and education in social psychology, sociology, anthropology, and political science.
We propose to use the experimental approach to systematically study the effects of social identity on individual preferences and how the mechanism designer can utilize group identity as one of the design variables in designing the optimal contract in a diverse work force.
We also explore the effects of social identity on equilibrium selection. Our studies draw on methods from experimental economics and social psychology and aim to capture causal effects of identity on economic decision making.
One to two REU students will participate in the research project gaining experience in all phases of the process from conducting literature reviews, to the design and implementation of the laboratory experiments, data analysis, and writing research papers. Professor Chen and her doctoral student mentor will closely supervise the REU students. REU students will participate in their weekly Behavioral and Experimental Lab meetings throughout the ten-week period.
REU students should (currently) be a sophomore or junior, with some prior programming experience. Students should have taken at least one programming course and one course in statistics. They will be expected to learn z-tree (a programming language for economic experiments), to program for experiments and to conduct data analysis.
Rearchitecting Online Conversation
Professor Paul Resnick
School of Information, U-M
One or two undergraduate students will work on the research project "Pivots, Trackers, and Recommenders: Making Online Conversations Simultaneously Serve Multiple Audiences."
The goal of the project is to create an alternative architecture for online conversation that improves on some of the features of online forums. In particular, the new architecture will enable a better integration of conversation with information objects, and enable a single space to appear on-topic to newcomers yet still permit social interaction among regulars.
The student(s) will help develop the software implementing the new architecture and prepare it for field trials in online communities. The software to be developed with be an add-on module or modules for Drupal, a popular open-source CMS.
The student(s) can expect to learn about recommender systems, social computing, information retrieval, and Drupal development, as well as practice programming skills. Requirements: PHP, Javascript.
For more information about the project, please contact Daniel Zhou at mrzhou@umich.edu.
Or for background about the project, see www.si.umich.edu/~presnick/papers/chi08/ and portal.acm.org/citation.cfm?id=1057126.
Trading Agent Competition
Professor Michael Wellman
Computer Science and Engineering, U-M
Professor Daniel Grosu
Computer Science, Wayne State University
The Trading Agent Competition (TAC) is an international forum designed to promote and encourage high quality research into the trading agent problem. TAC tournaments have been held annually since 2000.
TAC provides a forum for researchers to evaluate programmed trading techniques in a market scenario by competing with agents from other design groups. Over the past decade, it has attracted participants from many institutions from dozens of countries around the world.
Two to four REU students will participate in the TAC teams each summer, with half of them joining the University of Michigan team supervised by Professor Wellman, and the other half joining the WSU team supervised by Professor Grosu. The students will contribute to the design and implementation of trading agents to enter the TAC-09 tournament, or to the development and analysis of the new 2009 TAC Ad Auction game (aa.tradingagents.org). Faculty and doctoral student advisors will closely supervise the student research activities.
As in the past, our undergraduate students will be expected to participate as co-authors in the writing of the academic papers to be published in peer-reviewed journals. Some programming experience preferred (particular language not essential).
What is Incentive-Centered Design? Field Studies of Human Computing on the Internet
Professor Jeffrey MacKie-Mason
School of Information, U-M
The Internet has unleashed an explosion of new forms of human interaction and collaboration. We see this in social computing, information sharing, software production, artistic creation, democratic engagement and many other domains. These applications are built on computing and networking technology. However, they represent a departure from past computing and communications: humans are now an integral part of the system, and indeed, the technology increasingly fades into the background while the human interactions they support come to the fore.
With technology-focused computing and networking, most important problems either required building hardware or programming software. With human-focused computing and networking, a new issue is central: human motivation. Humans are crucial components of modern information systems, and they are very smart devices, but they are not programmable. People are autonomous, and to induce them to behave in ways that make systems succeed requires that they be motivated. To design human-focused information systems, we have been developing an emerging science of incentive-centered design.
Incentive-centered design is a multidisciplinary science that combines the behavioral sciences of motivation (including economics, psychology, game theory and others) with engineering design sciences of computing and networking. It is an emerging field, that is increasingly successful but not yet widely known or understood.
In this project, an REU student will participate in a project to write articles and possibly a book to define and explain the emerging science of incentive-centered design. The main research activity will be to develop a set of real-world case studies of human motivation problems on the Internet, and interesting solutions to them. As a well-known example (but not necessarily one of the case studies we will do), how does Wikipedia convince volunteers to create content, and edit others' contributions to improve their quality? What specific problems of motivation emerged as Wikipedia developed, and how were they handled? Other well-known examples (but we will in fact likely be looking for lesser known cases) include Flickr (why do people tag their photos for the benefit of others?), open source software, and Facebook. We will be particularly interested in finding both unique and common methods used by different Internet projects to address human motivation and incentive problems.
A qualified student will have good information-finding skills, be good at organizing and structuring facts, be diligent about documenting research, and be a clear writer. Experience with data gathering, management and basic statistical analysis is a plus, but is not essential. The student should be familiar with and interested in social computing, Web 2.0, and other human-focused Internet activities.
Testing Non-monetary Incentives for Contributing Internet Content
Professor Jeffrey MacKie-Mason
School of Information, U-M
The Internet enables geographically separate people to share valuable information inexpensively. Many information sharing systems have been successful and socially valuable: Wikipedia (encyclopedia); Amazon(product reviews); Flickr (photos); Yahoo! Answers (questions and answers); Digg (editorial selection of news stories); del.icio.us (recommended Web sites); CiteULike (scholarly citations). However, creating, editing, uploading, and sharing information requires time and effort. Most promising projects fail because not enough people are sufficiently motivated to make the effort.
Monetary compensation is not feasible on many information sharing systems for practical or cultural reasons. Therefore, we are designing and testing non-monetary incentives to motivate contribution effort. For example, one non-monetary mechanism has been used in peer-to-peer communities to motivate file sharing: download speed depends on how much one uploads. We will conduct a human-subject laboratory experiment to test the performance of a few selected non-monetary mechanisms.
One to two REU students will participate in the design and implementation of the laboratory experiments, communications with the subjects, data analysis, and preparation of research articles. REU students should have programming skills. They will be expected to learn z-tree (a programming language for economic experiments) program for experiments and to conduct data analysis.
Bidding Behavior in Government Procurement Auctions
Mark McCabe
Visiting Assistant Professor
School of Information, U-M
This project makes use of a unique dataset assembled during a 2007 federal and state antitrust investigation of a merger between the two largest providers of school bus transportation services in the United States. Our focus will be on auctions for school bus contracts conducted by school districts throughout the state of NJ over the past decade. Using the data, we plan to address a number of questions including:
- How many bidders are required to achieve a competitive outcome?
- If the answer to (1) is "too many!" can we determine whether firms are (tacitly) colluding to keep prices high?
- Have past mergers raised prices?
- Given NJ's approach to auctions, is there room for improvement, and if so, what might alternative auction mechanisms look like?
The student(s) will work closely with Professor McCabe and an SI doctoral student. Depending on their skills and interests, they could assist with data collection/processing, literature reviews, programming and/or econometric analysis. We continue to collect data from the folks in NJ (with the assistance of its Department of Education and Attorney General) and anticipate frequent interaction with officials there.
Ideally the student(s) will have an understanding of basic econometrics. Knowledge of a specific econometric software package is helpful but not required. Much of the data analysis will be conducted using MatLab and Stata. Therefore an ability to write basic code for MatLab/Stata or experience with programming in general is desirable. Good people skills would also be helpful – we expect the student(s) to manage some of the data collection efforts with various school districts.
Home > Research > Research Experience for Undergraduates (REU) > Projects
|
|
|
|
