Events

Two-thirds of all American adults access the news through social media. But social networks and social media recommendations lead to information bubbles and personalization and recommendations by maximizing the click-through rate lead to ideological polarization. Consequently rumors false news conspiracy theories and now even fake news sites are an increasingly worrisome phenomena. While media organizations (Snopes.com PolitiFact FactCheck.org et al.) have stepped up their efforts to verify news political scientists tell us that fact-checking efforts may be ineffective or even counterproductive. To address some of these challenges researchers at Indiana University are working on an open platform for the automatic tracking of both online fake news and fact-checking on social media. The goal of the platform named Hoaxy is to reconstruct the diffusion networks induced by hoaxes and their corrections as they are shared online and spread from person to person.

Social relations are the foundation of human daily life. Developing techniques to analyze such relations in visual data such as photos bears great potential to build machines that better understand people at a social level. Additionally through better understanding about such hidden information in exposed photos we would like to inform people about potential privacy risks. Social domain-based theory from social psychology is a great starting point to systematically approach social relation recognition. The theory provides a coverage of all aspects of social relations and equally is concrete and predictive about the visual attributes and behaviors defining the relations in each domain. Our work provides the first photo dataset built on this holistic conceptualization of social life that is composed of a hierarchical label space of social domains and social relations and contributes the first models to recognize such domains and relations and find superior performance for attribute based features. Beyond the encouraging performances we have some findings of interpretable features that are in accordance with the predictions from social psychology literature. Our work mainly contributes to interleave visual recognition and social psychology theory that has the potential to complement the theoretical work in the area with empirical and data-driven models of social life.

Data use policies govern how data retrieval systems process data items coming from many different sources each subject to its own integrity and confidentiality requirements. Ensuring compliance with these requirements despite bugs misconfigurations and operator errors in a large complex and fast evolving system is a major challenge.

In this thesis we present comprehensive and practical compliance systems to address this challenge. To be comprehensive compliance systems must be able to enforce policies specific to individual data items or to a particular client's data items the service provider's own policies and policies that capture legal requirements. To be practical compliance systems need to meet stringent requirements: runtime overhead must be low; existing applications can run with little modifications; and bugs misconfigurations compromises in application code or actions by unprivileged operators cannot violate policies.

We present the design and implementation of two comprehensive and practical compliance systems: Thoth and Shai. At a high-level data use policies are stated in a declarative language separate from application code and a small reference monitor ensures compliance with these policies. Thoth and Shai differ in enforcement techniques. Thoth tracks data flows through the system at runtime by intercepting I/O at processes' boundaries and enforces the associated policies. Shai on the other hand combines static flow analysis and light-weight runtime monitoring (sandboxes and capabilities) to ensure compliance of data flows. We demonstrate the practicality of these systems using a prototype search engine based on the popular Apache Lucene.

Computing resources are fundamentally limited and sometimes an exact solution may not even exist. Thus when implementing real-world systems approximations are inevitable as are the errors introduced by them. The magnitude of errors is problem-dependent but higher accuracy generally comes at a cost in terms of memory energy or runtime effectively creating an accuracy-efficiency tradeoff. To take advantage of this tradeoff we need to ensure that the computed results are sufficiently accurate otherwise we risk disastrously incorrect results or system failures. Unfortunately the current way of programming with approximations is mostly manual and consequently costly error prone and often produces suboptimal results.

In this talk I will present our vision and efforts so far towards an approximating compiler for numerical computations. Such a compiler would take as input exact high-level code with an accuracy specification and automatically synthesize an approximated implementation which is as efficient as possible but verifiably computes accurate enough results.

Social media has brought a revolution on how people get exposed to information and how they are consuming news. Beyond the undoubtedly large number of advantages and capabilities brought by social-media platforms a point of criticism has been the creation of filter bubbles or echo chambers caused by social homophily as well as by algorithmic personalisation and recommendation in content delivery. In this talk I will present the methods we developed to (i) detect and quantify the existence of polarization on social media (ii) monitor the evolution of polarisation over time and finally (iii) devise methods to overcome the effects caused by increased polarization. We build on top of existing studies and ideas from social science with principles from graph theory to design algorithms which are language independent domain agnostic and scalable to large number of users.