UMD Theses and Dissertations
Permanent URI for this collectionhttp://hdl.handle.net/1903/3
New submissions to the thesis/dissertation collections are added automatically as they are received from the Graduate School. Currently, the Graduate School deposits all theses and dissertations from a given semester after the official graduation date. This means that there may be up to a 4 month delay in the appearance of a given thesis/dissertation in DRUM.
More information is available at Theses and Dissertations at University of Maryland Libraries.
Browse
3 results
Search Results
Item Designing for the Human in the Loop: Transparency and Control in Interactive Machine Learning(2020) Renner, Alison Marie; Boyd-Graber, Jordan; Findlater, Leah; Computer Science; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Interactive machine learning techniques inject domain expertise to improve or adapt models. Prior research has focused on adapting underlying algorithms and optimizing system performance, which comes at the expense of user experience. This dissertation advances our understanding of how to design for human-machine collaboration--improving both user experience and system performance--through four studies of end users' experience, perceptions, and behaviors with interactive machine learning systems. In particular, we focus on two critical aspects of interactive machine learning: how systems explain themselves to users (transparency) and how users provide feedback or guide systems (control). We first explored how explanations shape users' experience of a simple text classifier with or without the ability to provide feedback to it. Users were frustrated when given explanations without means for feedback and expected model improvement over time even in the absence of feedback. To explore transparency and control in the context of more complex models and subjective tasks, we chose an unsupervised machine learning case, topic modeling. First, we developed a novel topic visualization technique and compared it against common topic representations (e.g., word lists) for interpretability. While users quickly understood topics with simple word lists, our visualization exposed phrases that other representations obscured. Next, we developed a novel, ``human-centered'' interactive topic modeling system supporting users' desired control mechanisms. A formative user study with this system identified two aspects of control exposed by transparency: adherence, or whether models incorporate user feedback as expected, and stability, or whether other unexpected model updates occur. Finally, we further studied adherence and stability by comparing user experience across three interactive topic modeling approaches. These approaches incorporate input differently, resulting in varied adherence, stability, and update speeds. Participants disliked slow updates most, followed by lack of adherence. Instability was polarizing: some participants liked it when it surfaced interesting information, while others did not. Across modeling approaches, participants differed only in whether they noticed adherence. This dissertation contributes to our understanding of how end users comprehend and interact with machine learning models and provides guidelines for designing systems for the ``human in the loop.''Item TRANSPARENCY AND TRUST IN JOURNALISM: AN EXAMINATION OF VALUES, PRACTICES AND EFFECTS(2015) Koliska, Michael; Steiner, Linda; Journalism; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Journalism scholars and practitioners have repeatedly argued that transparency is crucial to generate trust in the news media, which, over the years, has faced continues decline in public trust. As news organizations have been encouraged to implement transparency in their daily work, transparency has increasingly gained the status of a professional norm in journalism. However, very little is actually known about how journalists think and apply transparency in their everyday practices or how news organizations in the United States implement transparency. Similarly, normative assumptions about the trust-generating effects of transparency have not been consistently shown to exist. This dissertation examined to what extent journalists at 12 national news outlets embraced transparency on a day-to-day basis and how these news organizations implement transparency online at the news item level. Moreover, this dissertation tested whether existing features of transparency (hyperlinks, editorial explanations, corrections, staff biographies etc.) impact audiences’ trust perception of a news story. The results of the mixed method approach showed that transparency in journalism is far from being a professional norm, which guides journalists’ news production processes. An analysis of 27 in-depth interviews found that journalists rarely consider transparency in their work. Journalists agreed that the notion of transparency has value. They repeatedly suggested that the news outlets they work for utilize transparency as a promotional tool to engage audiences and to appear transparent, rather than significantly disclosing information about the inner workings of their news organization. The results of the content analysis supported this claim as the findings show that the transparency features news organizations currently use provide little information for audiences to learn about how journalism is done. Meanwhile, the results of two experiments indicate that participants may not recognize the intended meanings of the varied transparency features, as participants’ trust perception did not vary across different transparency conditions. The findings of this dissertation suggest that transparency in journalism is still a goal rather than reality. News organizations have not opened up to the extent that they may be understood as transparent organizations; instead their efforts to pull back the curtain so that audiences may see the inner workings of newsrooms can be considered translucent at best.Item Official Secrecy: Self, State and Society(2004-07-12) Ellington, Thomas Coke; Barber, Benjamin R; Government and Politics; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)In meeting the threat posed by terrorism, the democratic state also faces a paradox: Those practices best suited to defending the state are often least suited to democracy. Such is the case with official secrecy, which has received renewed attention. Military and intelligence operations frequently depend on secrecy for their success. At the same time, democracy depends on openness, a fact too often neglected by democratic theory. Democratic theory presumes that citizens are at least minimally capable of making decisions to steer the ship of state, a presumption that requires citizens not only to have the skills necessary to make political decisions but to have the information necessary to make those decisions competently. However, in many areas of the most vital public interest (e.g. foreign policy, nuclear weapons, decisions regarding war and peace), the state intentionally conceals information from citizens. While other factors, such as high information costs, may work against an informed citizenry, official secrecy is qualitatively different and uniquely damaging to democratic governance, even granting that in some instances it may be a necessary evil. Official secrecy subverts the very democratic values it is frequently designed to protect, denying citizen competence, reducing accountability and diminishing the legitimacy of the state, as well as distorting the historical record and creating fertile ground for paranoid-style thinking. Democratic theorists have not been unaware of the importance of information to democratic citizenship. Indeed awareness has promoted the defense of the institutions of free expression as the best means for ensuring that necessary political information is accessible. However, that is no longer enough, as the last century has seen states become producers and repositories of information on a never-before-seen scale. The task for democratic theory now is to recognize this change in the information environment and recognize the importance of this new locus of political information. Understanding and minimizing the impact of official secrecy is a necessary part of this process.