Joel Ross

Crowdsourcing is rapidly emerging as a new paradigm for solving problems: rather than having a single expert individual or group find a solution to a problem, the task is instead given to a crowd--a large group of amateur and expert users alike. Indeed, crowdsourcing methods are currently allowing humans to solve difficult computation problems--such as image identification and natural language processing--in place of computers. I seek to explore the potential social, technological, and environmental impacts of using crowdsourcing to perform computational tasks, as well as to develop technological innovations that can support this process. By deepening the understanding of this mode of interaction, my research can enable novel and more effective ways of solving significant social and computational problems.

Amazon.com's Mechanical Turk (http://www.mturk.com) is an online system in which workers are paid small sums of money to work on projects that, while quick and easy for a human, are very difficult for computers to perform correctly and efficiently--for example, identifying and object in an image or categorizing data. Furthermore, workers have little or not context for the work they perform through this highly mediated system. In this project, we explore the question of who are the Mechanical Turkers, looking to identify the kinds of people who work on MTurk and understand the reasons (both intrinsic and extrinsic) for their participation. We are also considering how the ubiquity of monetary rewards may shape the interactions with and potential exploitations within this crowdsourcing system.

A popular way for people to understand their environmental impact is by using an online carbon footprint calculator. Although there are a variety of such calculators available, the majority share the same form of user interaction. We analyze how, with this mode of interaction, most calculators focus on the environmental impact of individual actions without drawing attention to the broader impacts of those actions on the surrounding community and world. To address these problems, we present the Better Carbon calculator, which uses collaborative filtering and location- based calculation to provide an individual footprint estimate while simultaneously affecting and improving the estimates for other people in a user's community. This method also allows Better Carbon to be more extendable, as additional forms of impact can be considered without requiring additional user effort. Better Carbon can thus provide quicker and easier footprint estimates, and help the process of calculating a carbon footprint create stronger linkages within the communities of its users.
Better Carbon can be found at http://www.bettercarbon.com

Mobile devices such as mobile phones are becoming more ubiquitous and gaining more capabilities, leading to their increasing use as portable media players. However, the great majority of produced visual media is targeted towards devices with larger displays and greater hardware capabilities, making it difficult for these media to be presented on mobile devices. These differences will only increase as multimedia becomes more and more interactive, such as with increasingly popular mobile games. In this project, we are exploring methods for retargeting visual media to mobile devices, considering how such methods have been applied to text, still images, and video content, and considering implications for the retargeting of games and virtual worlds.

As mobile devices become more ubiquitous, we are likely to see the spread of computational systems for urban sensing and other forms of context identification. However, there has not yet been sufficient consideration of how to reduce the potential environmental impacts of such systems. In an attempt to combat these foreseen impacts, we apply the concept of sustainable sharing to urban sensing, envisioning a distributed mode of interaction in which users can automatically share locally sensed information with other collocated devices. We present simulations and survey findings that demonstrate the viability of such a peer-to-peer interaction, as well as indicate key challenges for deployment such as privacy concerns. Due to energy requirements, sensor sharing using current technology may not yet make urban sensing more sustainable. However, our findings suggest that sharing capabilities and functionality can eventually be a viable interaction mode for environmentally sustainable urban sensing.

More and more, people desire information that enables them to reduce their environmental impact during their daily lives. Although numerous services that offer this information have appeared, current "expert"-driven methods are unable to deal with the vast scope of determining the environment impact of every consumer product. Furthermore, few of these services are able to provide information whenever a user my desire it, such as at the point of purchase. In this paper, we suggest that participatory culture may allow for large-scale environmental impact information via distributed data gathering and analysis. We developed a vision of ubiquitous environmental impact information enabled by participatory culture, in which information for all products is provided by all people and is available at all times.
The prototype can be found at http://www.greenscanner.net