Chen, K. (2019). When People' Voices Matter: Examining Mini-Public Deliberation and Digital Crowdsourcing with Machine Learning Tools. Stanford University, Stanford, CA. PDF
[winner of the 2019 National Communication Association (NCA) Political Communication Division's Lynda Lee Kaid Outstanding Dissertation Award]
Governments across the world have been increasingly using a variety of forms of public consultation to inform governance and strengthen legitimacy. In some public consultation, people’s voices are thoughtful and consequential for policymaking while in other forms of public consultation, people’s voices are more likely to be distorted by political interests. This dissertation aims to answer a question (challenge) posed by the latest development in the field deliberative democracy: what institutional designs can make a deliberative system function well or badly?
To answer this question, I examine in depth, cases of mini-public deliberation and digital crowdsourcing across nations in Ghana, United States and China. In chapter 2, I studied a well-organized mini-public consultation -- Deliberative Poll -- to challenge the prevailing skepticism that ordinary citizens cannot reason about complex policy issues and their voices are unlikely to impact policymaking. I used various machine learning tools to measure people’s use of reasoning from thousands of speech acts in a deliberative poll in Tamale, Ghana. I found that Tamale participants utilized a comparable or even higher level of reasoning when they expressed their opinions and when they responded to their peers, compared to their peers in the most advanced nation.
In chapter 3 and chapter 4, I studied another important component of a deliberative system -- digital crowdsourcing -- to challenge the premise that digital media and technology enhance government responsiveness by amplifying citizen voices and lowering the cost of communication. I argue, instead, that politicians have a hard time handling the massive amount and uneven quality of citizen demands posted in digital platforms. Faced with this data challenge, they often resort to the strategy of delegation. Collecting evidence from various countries including China and the United States, I show that this coping strategy can distort citizen voices -- dampening some citizens’ influence, while amplifying others. To demonstrate these consequences, I employ a range of data sources and research designs, including administrative records (public and leaked), web-scraped citizen posts and government responses, and field interviews.
In the conclusion chapter, I summarized five key lessons on designing an effective deliberative system for facilitating healthy and impactful public dialogues in both democracies and authoritarianisms. Theoretically, my dissertation aims to reconcile the literature on deliberative systems with distortions in everyday political talk by showing that political talk will be less likely to suffer distortions under good designs such as the deliberative microcosm or digital crowdsourcing under carefully-designed website technologies. In addition to contributing to the relative new development of the field --deliberative system -- by providing specific design lessons, this dissertation also contributes to the measurement and methodology of studying the quality and impact of a deliberative system. To my knowledge, this dissertation is the first one to apply machine learning to the Discourse Quality Index to systematically analyze the quality of people’s reasoning in political talk. Besides applying a variety of machine learning tools across the chapters, this dissertation also contributes to the measurement of quality in political talk by enhancing how we evaluate people’s reasoning beyond the standard DQI.
[winner of the 2019 National Communication Association (NCA) Political Communication Division's Lynda Lee Kaid Outstanding Dissertation Award]
Governments across the world have been increasingly using a variety of forms of public consultation to inform governance and strengthen legitimacy. In some public consultation, people’s voices are thoughtful and consequential for policymaking while in other forms of public consultation, people’s voices are more likely to be distorted by political interests. This dissertation aims to answer a question (challenge) posed by the latest development in the field deliberative democracy: what institutional designs can make a deliberative system function well or badly?
To answer this question, I examine in depth, cases of mini-public deliberation and digital crowdsourcing across nations in Ghana, United States and China. In chapter 2, I studied a well-organized mini-public consultation -- Deliberative Poll -- to challenge the prevailing skepticism that ordinary citizens cannot reason about complex policy issues and their voices are unlikely to impact policymaking. I used various machine learning tools to measure people’s use of reasoning from thousands of speech acts in a deliberative poll in Tamale, Ghana. I found that Tamale participants utilized a comparable or even higher level of reasoning when they expressed their opinions and when they responded to their peers, compared to their peers in the most advanced nation.
In chapter 3 and chapter 4, I studied another important component of a deliberative system -- digital crowdsourcing -- to challenge the premise that digital media and technology enhance government responsiveness by amplifying citizen voices and lowering the cost of communication. I argue, instead, that politicians have a hard time handling the massive amount and uneven quality of citizen demands posted in digital platforms. Faced with this data challenge, they often resort to the strategy of delegation. Collecting evidence from various countries including China and the United States, I show that this coping strategy can distort citizen voices -- dampening some citizens’ influence, while amplifying others. To demonstrate these consequences, I employ a range of data sources and research designs, including administrative records (public and leaked), web-scraped citizen posts and government responses, and field interviews.
In the conclusion chapter, I summarized five key lessons on designing an effective deliberative system for facilitating healthy and impactful public dialogues in both democracies and authoritarianisms. Theoretically, my dissertation aims to reconcile the literature on deliberative systems with distortions in everyday political talk by showing that political talk will be less likely to suffer distortions under good designs such as the deliberative microcosm or digital crowdsourcing under carefully-designed website technologies. In addition to contributing to the relative new development of the field --deliberative system -- by providing specific design lessons, this dissertation also contributes to the measurement and methodology of studying the quality and impact of a deliberative system. To my knowledge, this dissertation is the first one to apply machine learning to the Discourse Quality Index to systematically analyze the quality of people’s reasoning in political talk. Besides applying a variety of machine learning tools across the chapters, this dissertation also contributes to the measurement of quality in political talk by enhancing how we evaluate people’s reasoning beyond the standard DQI.