Copyright (c) 2008 All rights reserved. http://works.bepress.com/luca_de_alfaro Recent documents in Luca de Alfaro en-us Sun, 17 Aug 2008 16:44:50 PDT 3600 http://works.bepress.com/luca_de_alfaro/5 http://works.bepress.com/luca_de_alfaro/5 Tue, 15 Jan 2008 12:03:20 PST he Wikipedia is a collaborative encyclopedia: anyone can contribute to its articles simply by clicking on an ``edit'' button. The open nature of the Wikipedia has been key to its success, but has also created a challenge: how can readers form an informed opinion on its reliability? We propose a system that computes quantitative values of trust for the text in Wikipedia articles; these trust values provide an indication of text reliability.The system uses as input the revision history of each article, as well as information about the reputation of the contributing authors, as provided by a reputation system. The trust of a word in an article is computed on the basis of the reputation of the original author of the word, as well as the reputation of all authors who edited the text in proximity of the word. The algorithm computes word trust values that vary smoothly across the text; the trust values can be visualized using varying text-background colors. The algorithm ensures that all changes to an article text are reflected in the trust values, preventing surreptitious content changes.We have implemented the proposed system, and we have used it to compute and display the trust of the text of thousands of articles of the English Wikipedia. To validate our trust-computation algorithms, we show that text labeled as low-trust has a significantly higher probability of being edited in the future than text labeled as high-trust. Anecdotal evidence seems to corroborate this validation: in practice, readers find the trust information valuable. B. Thomas Adler Internet Commerce, Reputation Systems http://works.bepress.com/luca_de_alfaro/4 http://works.bepress.com/luca_de_alfaro/4 Wed, 21 Feb 2007 12:47:11 PST We present a content-driven reputation system for Wikipedia authors. In our system, authors gain reputation when the edits they perform to Wikipedia articles are preserved by subsequent authors, and they lose reputation when their edits are rolled back or undone in short order. Thus, author reputation is computed solely on the basis of content evolution; user-to-user comments or ratings are not used. The author reputation we compute could be used to flag new contributions from low-reputation authors, or it could be used to allow only authors with high reputation to contribute to controversial or critical pages. A reputation system for the Wikipedia could also provide an incentive for high-quality contributions. We have implemented the proposed system, and we have used it to analyze the entire Italian and French Wikipedias, consisting of a total of 691,551 pages and 5,587,523 revisions. Our results show that our notion of reputation has good predictive value: changes performed by low-reputation authors have a significantly larger than average probability of having poor quality, as judged by human observers, and of being later undone, as measured by our algorithms. B. Thomas Adler Internet Commerce, Reputation Systems http://works.bepress.com/luca_de_alfaro/3 http://works.bepress.com/luca_de_alfaro/3 Tue, 21 Nov 2006 11:08:55 PST On-line forums for the collaborative creation of bodies of information are a phenomenon of rising importance; the Wikipedia is one of the best-known examples. The open nature of such forums could benefit from a notion of reputation for its authors. Author reputation could be used to flag new contributions from low-reputation authors, and it could be used to allow only authors with good reputation to contribute to controversial or critical pages. A reputation system for the Wikipedia would also provide an incentive to give high-quality contributions.We present in this paper a novel type of content-driven reputation system for Wikipedia authors. In our system, authors gain reputation when the edits and text additions they perform to Wikipedia articles are long-lived, and they lose reputation when their changes are undone in short order. We have implemented the proposed system, and we have used it to analyze the entire Italian and French Wikipedias,consisting of a total of 691,551 pages and 5,587,523 revisions. Our results show that our notion of reputation has good predictive value: changes performed by low-reputation authors have a significantly larger than average probability of having poor quality, and of being undone. B. Thomas Adler Internet Commerce, Reputation Systems http://works.bepress.com/luca_de_alfaro/2 http://works.bepress.com/luca_de_alfaro/2 Tue, 31 Oct 2006 00:16:07 PST A concurrent reachability game is a two-player game played on a graph: at each state, the players simultaneously and independently select moves; the two moves determine jointly a probability distribution over the successor states. The objective for player 1 consists in reaching a set of target states; the objective for player 2 is to prevent this, so that the game is zero-sum. Our contributions are two-fold. First, we present a simple proof of the fact that in concurrent reachability games, for all " > 0, memoryless "-optimal strategies exist. A memoryless strategy is independent of the history of plays, and an "-optimal strategy achieves the objective with probability within " of the value of the game. In contrast to previous proofs of this fact, which rely on the limit behavior of discounted games using advanced Puisieux series analysis, our proof is elementary and combinatorial. Second, we present a strategy-improvement (a.k.a. policy-iteration) algorithm for concurrent games with reachability objectives. Krishnendu Chatterjee Game Theory http://works.bepress.com/luca_de_alfaro/1 http://works.bepress.com/luca_de_alfaro/1 Mon, 30 Oct 2006 23:37:09 PST We present a novel abstraction technique which allows the analysis of reachability and safety properties of Markov decision processes with very large state spaces. The technique, called magnifying-lens abstraction, copes with the state-explosion problem by partitioning the state-space into regions, and by computing upper and lower bounds for reachability and safety properties on the regions, rather than on the states. To compute these bounds, magnifying-lens abstraction iterates over the regions, considering the concrete states of each region in turn, as if one were sliding across the abstraction a magnifying lens which allowed viewing the concrete states. The algorithm adaptively refines the regions, using smaller regions where more detail is needed, until the difference between upper and lower bounds is smaller than a specified accuracy. We provide experimental results illustrating that magnifying-lens abstractions can provide accurate answers, with drastic savings in memory requirements, in many cases where previous abstraction techniques yield no benefit. Luca de Alfaro Probabilistic Systems