The effective resistance between two nodes of a weighted graph is the electrical seen resistor network with branch conductances given by edge weights. comes up in many applications and fields addition to analysis, including, for example, Markov chains continuous-time averaging networks. In this paper we study problem allocating weights on order minimize total resistance, i.e., sum resistances all pairs nodes. We show that convex optimization can be solved efficiently either numerically or,...

The algebraic connectivity of a graph is the second smallest eigenvalue Laplacian, and measure how well-connected is. We study problem adding edges (from set candidate edges) to so as maximize its connectivity. This difficult combinatorial optimization, we seek heuristic for approximately solving problem. standard convex relaxation can be expressed semidefinite program (SDP); modest sized problems, this yields cheaply computable upper bound on optimal value, well choosing added. describe new...

We initiate the study of markets for private data, through lens differential privacy. Although purchase and sale data has already begun on a large scale, theory privacy as commodity is missing. In this paper, we propose to build such theory. Specifically, consider setting in which analyst wishes buy information from population he can estimate some statistic. The obtain an accurate cheaply, while owners experience cost their loss privacy, must be compensated loss. Agents are selfish, wish...

A mechanism for releasing information about a statistical database with sensitive data must resolve trade-off between utility and privacy. Publishing fully accurate maximizes while minimizing privacy, publishing random noise accomplishes the opposite. Privacy can be rigorously quantified using framework of differential which requires that mechanism's output distribution is nearly same whether given row included. The goal this paper to formulate provide strong general guarantees, subject We...

A mechanism for releasing information about a statistical database with sensitive data must resolve trade-off between utility and privacy. Publishing fully accurate maximizes while minimizing privacy, publishing random noise accomplishes the opposite. Privacy can be rigorously quantified using framework of differential which requires that mechanism's output distribution is nearly same whether or not given row included excluded. The goal this paper strong general guarantees, subject to We...

A large fraction of user-generated content on the Web, such as posts or comments popular online forums, consists abuse spam. Due to volume contributions sites, a few trusted moderators cannot identify all abusive content, so viewer ratings must be used for moderation. But not viewers who rate are trustworthy and accurate. What is principled approach assigning trust aggregating user ratings, in order accurately content? In this paper, we introduce framework address problem moderating using...

Discussion forums, employed by MOOC providers as the primary mode of interaction among instructors and students, have emerged one important components online courses. We empirically study contribution behavior in these collaborative learning forums using data from 44 MOOCs hosted on Coursera, focusing primarily highest-volume contributors---"superposters"---in a forum. explore who superposters are their engagement patterns across platform, with focus following question---to what extent is...

Crowdsourcing is now widely used to replace judgement or evaluation by an expert authority with aggregate from a number of non-experts, in applications ranging rating and categorizing online content all the way student assignments massively open courses (MOOCs) via peer grading. A key issue these settings, where direct monitoring both effort accuracy infeasible, incentivizing agents 'crowd' put make good evaluations, as well truthfully report their evaluations. We study design mechanisms for...

We model the economics of incentivizing high-quality user generated content (UGC), motivated by settings such as online review forums, question-answer sites, and comments on news articles blogs. provide a game-theoretic within which to study problem high quality UGC, in contributors are strategic exposure. Our has feature that both contributions well extent participation is determined endogenously free-entry Nash equilibrium.

Software development is among the first demonstrations of using Large Language Models (LLMs) to enhance human productivity. Such a co-pilot paradigm envisions LLM working side-by-side with developers assist in programming tasks. Ensuring security software products critical factor for success such paradigm. There have been various anecdotal reports on LLMs detect vulnerabilities programs. This paper set experiments applying four well-known two widely referenced public datasets evaluate...

Motivated by the emergence of auction-based marketplaces for display ads such as Right Media Exchange, we study design a bidding agent that implements advertising campaign in marketplace. The must acquire given number impressions with target spend, when highest external bid marketplace is drawn from an unknown distribution P. quantity and spend constraints arise fact are usually sold on CPM basis. We consider both full information setting, where winning price each auction announced publicly,...

Most models for online advertising assume that an advertiser's value from winning ad auction, which depends on the clickthrough rate or conversion of advertisement, is independent other advertisements served alongside it in same session. This ignores important 'externality effect': as audience has a limited attention span, high-quality page can detract ads page. That is, utility to winner such auction also set winners.

Gamification is growing increasingly prevalent as a means to incentivize user engagement of social media sites that rely on contributions. Badges, or equivalent rewards such top-contributor lists are used recognize user's contributions site, clearly appear be valued by users who actively pursue and compete for them. However, different use badge designs, varying how, what, badges awarded--- some StackOverflow award meeting fixed levels contribution, while others like Amazon Y! Answers reward...

We consider the problem of designing a survey to aggregate non-verifiable information from privacy-sensitive population: an analyst wants compute some statistic private bits held by each member population, but cannot verify correctness reported participants in his survey. Individuals population are strategic agents with cost for privacy, ie, they not only account payments expect receive mechanism, also their privacy costs any revealed about them mechanism's outcome---the computed as well...

Gamification is growing increasingly prevalent as a means to incentivize user engagement of social media sites that rely on contributions. Badges , or equivalent rewards, such top-contributor lists are used recognize user's contributions site, clearly appear be valued by users who actively pursue and compete for them. However, different use badge designs varying how, what, badges awarded. Some sites, StackOverflow, award meeting fixed levels contribution. Other Amazon Y! Answers, reward...

Previous studies have highlighted the high arrival rate of new contenton web. We study extent to which this content can beefficiently discovered by a crawler. Our has two parts. First,we inherent difficulty discovery problem using amaximum cover formulation, under an assumption perfect estimates oflikely sources links content. Second, we relax thisassumption and more realistic setting in algorithms mustuse historical statistics estimate pages are most likely toyield recommend simple...

We investigate the design of mechanisms to incentivize high quality outcomes in crowdsourcing environments with strategic agents, when entry is an endogenous, choice. Modeling endogenous markets important because there a nonzero cost making contribution any which can be avoided by not participating, and indeed many sites based on crowdsourced content do have adequate participation. use mechanism monotone, rank-based, rewards model where agents strategically make participation choices capture...

When online ads are shown together, they compete for user attention and conversions, imposing negative externalities on each other. While the competition in sponsored search can be captured via models of clickthrough rates, post-click conversions cannot: since value-per-click an advertiser is proportional to conversion probability conditional a click, which depends other displayed, private value no longer one-dimensional, GSP mechanism not adequately expressive. We study design expressive...