Texte remanié de: PhD--Cambridge university, 1992. Titre de soutenance : The formation of opinion and the communication network in London 1637 to c. 1645. / Bibliogr. p. 307-326. Index.

The literature on authoritarian politics emphasises the threat unexpected shocks such as economic downturns or political and security challenges pose for regime stability. However, we know relatively little about how incumbents can influence the process by which citizens evaluate government performance and attribute blame in non-democratic regimes. To gain insights into these questions, I study how government responses to collective shocks to citizen income and security influence support for electoral authoritarian regimes, i.e. those that combine authoritarian practices with multiparty elections. I propose that when shocks make information about government performance publicly observable, illiberal governments can take action that moderates or ameliorates their effect on the levels of support they enjoy. Little constrained by constitutional rules, critical media or coalition partners, electoral authoritarians can use tactical redistribution to appease the discontented electorate on the one hand and propaganda to manipulate attributions of responsibility for the shock on the other. Repression against opposition parties and activists in this context is used rarely, and only after targeted transfers and propaganda have failed to prevent support from eroding and crowds from taking to the streets. The thesis illustrates arguments with the case of contemporary Russia - an electoral authoritarian regime with high levels of personalist rule - and leverages evidence from government and citizen responses to natural disasters, economic downturns, terror attacks and electoral protests. Empirical analysis combines original datasets on the framing of economic news in Kremlin-controlled media, the forced dismissals of government actors, the provision of tactical redistribution, and the use of repression against opposition parties and activists with over 60,000 responses from nationally and regionally representative public opinion surveys. Bringing new data and evidence from individual-level surveys to bear on the debate of how non-democratic governments manage public opinion, the thesis makes an original contribution to scholarship on authoritarian vulnerability and resilience.

This paper explores the environmental policies and practices of the U. S. Department of Defense (DoD) from the late 1950s until the mid 1990s in relation to the views and attitudes of the American people. While by no means a comprehensive examination of the military and each of the branches, this paper covers the general attitudes and rationales of the DoD as a whole. The time frame covered deviates from more obvious choices, such as "since World War II" or a specific decade, intentionally. Cutting off 15 or so years from each end could situate this work as the middle of a three -volume set for a number of reasons. For the starting time, nearly a generation passed since World War II. This allowed adequate time for the younger populace to play a role in the decisionmaking process. This also allowed for a beginning just prior to Sikes Act and America's renewed interest in the environment. At the other end, the terminus follows the end of the Cold War and two rounds of military downsizing. This set up a time of not only preventing environmental harm in the future, but also digging in and cleaning past harm for the future. As such, this leaves ample room for future work to concentrate on the time periods omitted here. The chapters cover the topic by decade with extra emphasis given to weather modification, which spanned multiple decades in such a manner thatcould not be effectively covered broken apart. For the latter decades, the focus follows more closely to Presidential Administrations. This method could not be fully utilized during the preceding discussion, in part due to the unique and revolutionary nature of the overall upheaval of the 1960s. The information included in the appendices provides a more detailed look into the ideas of the American people, which could not be included in the body of the text without detracting from the flow and readability. The appendices include a chapter on the public's responses to the General Social Surveys administered from 1972 to 1991. The questions examined have been grouped biennially in sequential even and odd years.

Religion and gender have been found to play significant roles in shaping po- litical attitudes such as party identification and ideology. While much of the focus has been on the "religion gap" and the "gender gap," little empirical research has explored how religion affects the political attitudes of men and women differently. Using a 2004 Pew survey, this study examines how religion and gender interact to affect four different areas of President Bush's approval in 2004: general approval, economic policy approval, foreign policy approval, and social policy approval. The results support a "religious gender gap" theory, where the effect of gender on presi- dential approval varies across levels of religious commitment. For general, economic policy, and foreign policy approval, secular men and women are more similar on average than highly religious men and women. For social policy approval, highly religious men and women are more similar on average than secular men and women.

Cette thèse a pour objectif la détection du langage figuratif dans les réseaux sociaux. Nous nous focalisons en particulier sur l'ironie et le sarcasme dans Twitter et proposons une approche basée sur l'apprentissage supervisée afin de prédire si le message véhiculé dans un tweet est ironique ou non. Pour ce faire, trois étapes ont été réalisées : (1) Analyse des phénomènes pragmatiques de l'ironie et annotation multi-niveaux d'un corpus de référence, (2) Développement d'un modèle de détection automatique pour les tweets en français qui exploite à la fois des traits sémantiques et le contexte extralinguistique, (3) Etude de la portabilité du modèle pour la détection de l'ironie dans un cadre multilingue (italien, anglais et arabe). Les résultats obtenus pour cette tâche extrêmement complexe sont très encourageants et permettrons d'améliorer sensiblement la détection de polarité lors de l'analyse de sentiments. / This thesis aims to detect figurative language devices in social networks. We focus in particular on irony and sarcasm in Twitter and propose an approach based on supervised learning to predict if a tweet is ironic or not. This required three steps: (1) Analyze the pragmatic phenomena of irony and a multi-level annotation of a corpus of reference, (2) Development of an automatic detection system for French tweets that exploits both semantic features and the extra-linguistic context, (3) Study of the portability of the model to detect irony in a multilingual setting (Italian, English and Arabic). The obtained results for this extremely complex task are very encouraging and will allow a significant improvement of polarity detection in sentiments analysis.

abstract: Studies on what shapes public perceptions of ex-prisoners are abundant. One omission is the detailed investigation of how perceptions of former inmates might vary by the amount of time since their last incarceration term. More specifically, it remains unknown whether increased length since an ex-prisoner’s last incarceration spell is positively linked to higher levels of trust. This study (N = 448) uses a factorial vignette design to test the perceived trustworthiness of former inmates across two hypothetical scenarios. Time since last incarceration spell is used as the independent variables in a series of ordered logistic regression models. The role of gender is also explored. Results show that trust perceptions of ex-prisoners minimally vary by time since last incarceration spell when personal victimization is at risk, but the magnitude is small and shows no clear pattern of declining risk over time. Less support is observed in situations where property victimization is at risk. These findings illustrate the complexity of how people perceive and feel about ex-inmates in situations of trust. / Dissertation/Thesis / Masters Thesis Criminology and Criminal Justice 2018

Cette thèse porte sur l’opinion publique et la démocratie. Elle s’intéresse particulièrement aux Guatémaltèques et à leur faible propension à appuyer la démocratie et certains des principes associés à celle-ci. Entre des explications qui associent ce comportement à un artéfact et à une culture autoritaire, des éléments contextuels et d’importantes nuances prévalent. Ce travail montre qu’il est nécessaire de traiter l’appui à la démocratie comme un objet multiface et malléable. En puisant dans les théories des perspectives et des bifurcations, un instrument d’analyse est élaboré sous le libellé de « temps court de l’appui à la démocratie ». Ce dernier nous permet de comprendre pourquoi, en période de crise politique, les points de vue sur la démocratie sont susceptibles d’évoluer. Sensibles aux circonstances extraordinaires, les opinions peuvent aussi, dans des périodes plus calmes, connaître une certaine stabilité. Les théories de l’ambivalence permettent d’élaborer un deuxième instrument, le « temps long de l’appui à la démocratie », qui aide à expliquer pourquoi les sondés offrent souvent des réponses « contradictoires » en matière de droits et de libertés. Aussi ambivalents que les autres Centraméricains, les Guatémaltèques se démarquent cependant par une forte tendance à esquiver les questions de sondage, à s’opposer aux droits des gens qui critiquent le gouvernement et à évaluer négativement le travail des institutions. Ces caractéristiques mettent en lumière des tensions entre les demandes et l’évaluation de l’offre en démocratie que font les citoyens. Cette thèse débute en étudiant les dimensions de la culture politique guatémaltèque et en proposant une typologie des indicateurs employés dans les grandes enquêtes internationales. / This thesis deals with public opinion and democracy. It mainly focuses on Guatemalans and on their weak propensity to support democracy and some of its dimensions. Between explanations that tie this behaviour to an artifact and to a culture of authoritarianism, contextual factors and important nuances can be found. We thus argue that it is advisable to deal with support for democracy as a multifaceted and malleable object. Drawing from perspectives and bifurcation theories, we develop an analytical instrument that we refer to as the "short-run of support for democracy". The latter allows us to understand why and how, in times of crisis, citizens’ viewpoints on democracy can easily become volatile. The fact that opinions are sensitive to extraordinary circumstances does not prevent them, under usual circumstances, from being relatively stable. Building on theories on ambivalence, we elaborate another instrument, the "long-run of support for democracy", which allows us to explain why survey respondents often give "contradictory" answers to questions dealing with rights and freedoms. While as ambivalent as other Central Americans, Guatemalans stand apart for their likelihood to avoid certain survey questions, to prefer limiting the rights of people who criticize the government, and to be particularly distrustful of state institutions. As background to those opinions and other paradoxical attitudes, we highlight existing tensions between citizens’ demands and their evaluation of the offer in democracy. This thesis first examines the various dimensions of the Guatemalan political culture and proposes a typology of indicators used in international barometer surveys.

This study aimed to examine previous empirical literature indicating that death penalty support contains a divide among Blacks and Whites and a gap among males and females. Previous literature has indicated that there has been a persistent racial divide and gender gap in death penalty support that has spanned over 60 years of research. Attempts to attenuate these divides have failed to fully explain why Whites are more likely than Blacks to support the death penalty and men are more likely than women to support the death penalty. This study proposes the use of empathy to control for these divides because research has indicated that those who are more empathic tend to be less punitive. Using data collected from a survey conducted on Amazon’s Mechanical Turk, a paid task website, this study attempted to attenuate the racial divide and gender gap by controlling for empathy. The sample consisted of 403 usable surveys that contained questions that measured sociodemographic characteristics, three measurements of empathy (cognitive, affective and ethnocultural), death penalty support, and attribution styles. The results indicated that there was not a racial divide or gender gap in death penalty support despite over 60 years of research indicating otherwise. Furthermore, this study failed to find a significant relationship between cognitive and affective empathy with death penalty support. This study did find a relationship between attribution styles and death penalty support as well as ethnocultural empathy with death penalty support. Individuals who scored higher on the situational attribution style were less likely to support the death penalty. Those who scored higher on the ethnocultural empathy scale were also less likely to support the death penalty. Future research should refrain from testing with Amazon’s Mechanical Turk as was not generalizable to the United States population. Research should be continued on different samples that have been shown to be more reliable than online surveys. Finally, research should be continued beyond empathy to examine what effects other controls have on the racial divide and gender gap in death penalty support.

This study develops and tests a group conflict model as an explanation for international immigration beliefs in the United States and Canada. Group conflict is structured by evaluations concerning group relationships and group members. At a conceptual level group conflict explains a broad range of policy beliefs among a large number of actors in multiple settings. Group conflict embodies attitudes relating to objective-based conditions and subjective-based beliefs.

Our society is overwhelmed with an ever growing amount of information. Effective management of this information requires novel ways to filter and select the most relevant pieces of information. Some of this information can be associated with the source or sources expressing it. Sources and their relation to what they express affect information and whether we perceive it as relevant, biased or truthful. In news texts in particular, it is common practice to report third-party statements and opinions. Recognizing relations of attribution is therefore a necessary step toward detecting statements and opinions of specific sources and selecting and evaluating information on the basis of its source. The automatic identification of Attribution Relations has applications in numerous research areas. Quotation and opinion extraction, discourse and factuality have all partly addressed the annotation and identification of Attribution Relations. However, disjoint efforts have provided a partial and partly inaccurate picture of attribution. Moreover, these research efforts have generated small or incomplete resources, thus limiting the applicability of machine learning approaches. Existing approaches to extract Attribution Relations have focused on rule-based models, which are limited both in coverage and precision. This thesis presents a computational approach to attribution that recasts attribution extraction as the identification of the attributed text, its source and the lexical cue linking them in a relation. Drawing on preliminary data-driven investigation, I present a comprehensive lexicalised approach to attribution and further refine and test a previously defined annotation scheme. The scheme has been used to create a corpus annotated with Attribution Relations, with the goal of contributing a large and complete resource than can lay the foundations for future attribution studies. Based on this resource, I developed a system for the automatic extraction of attribution relations that surpasses traditional syntactic pattern-based approaches. The system is a pipeline of classification and sequence labelling models that identify and link each of the components of an attribution relation. The results show concrete opportunities for attribution-based applications.