Big data is transforming all aspects of the human experience, be it everyday life, scientific exploration and discovery, medicine, business, law, journalism, and decision-making at all levels of government.

On the one hand, big data is primarily driven by computing technology becoming better and cheaper. As expected, these advances in computing technology lead to exponential increases in the size of data generated and processed. However, they also translate to exponential increases in data collected by scientific instruments, be it centralized massive instruments (such as the Large Hadron Collider and the Large Synoptic Survey Telescope), or by great numbers of small-but-now-more-affordable instruments (such as those used for next-generation sequencing), or by even greater numbers of personal mobile devices and tiny sensors.

On the other hand, the thirst for data is becoming the norm both from a consumer point of view (e.g., businesses want to collect as much data as possible for their customers) and also from a producer point of view (e.g., people increasingly feel the urge to share more and more details of their lives on social networks), leading to an exponential increase in user-contributed content.

Despite the increases in computing technology and availability/demand for data in the last few decades, the performance of one critical component in the data processing pipeline has remained roughly the same. Namely, the ability of humans to process data has not changed significantly in the last few decades! 

We refer to this disparity as: the big data - same humans problem. This means that taking into account the user point-of-view is extremely crucial. Imagine a very efficient data management system that can process 1,000,000 data inputs per second and generate a mere one alert per second (i.e., a reduction of 1 in 1 million). Such a system will immediately overwhelm any potential user; prioritization/ranking of results, classification of important/not-important results, and other similar techniques are absolutely essential to make such a system usable/useful.

This project aimed to solve this problem and make data management more user-centric, by considering user preferences in different aspects of data management. 

In particular, the project investigated both ``flavors'' of data: (i) data at rest, i.e., traditional database systems where data is stored in a database and query results are computed based on the current contents of the database and (ii) data in motion, i.e., data stream management systems, that process a never-ending input stream of data, searching for patterns/conditions of interest to the users. 

Towards this, the project examined three different dimensions. First, it considered user preferences for different aspects of quality. For example, when a user prefers fast answers, but is willing to tolerate a small degree of staleness in the results, or when a user prefers to always get ``fresh'' results, but is willing to tolerate some small delay. Such preferences can be used by the system to determine resource allocation. Our proposed framework (named Quality Contracts) enables the system to satisfy the (potentially conflicting) preferences of multiple users instead of having a single global quality metric. Multiple research questions where addressed within this space, including how to determine the execution order of different operations in the presence of priorities, how to decide which data to drop if the system is overwhelmed, etc.

Secondly, the project examined user preferences for query result personalization (e.g., for ranking). In particular, a novel framework was proposed that combines quantitative preferences (e.g., I give this movie four stars out of five) and...

Please report errors in award information by writing to: awardsearch@nsf.gov.