Aktuelle Publikationen

This PhD thesis investigates patterns of nonviolent activism in civil war contexts with the example of the Nepalese Civil War. The thesis is divided into three major parts. The first part offers a broad theoretical classification of nonviolent action research, while introducing the motivation and research contribution of the thesis. In the second part the three different contributions (papers) of this thesis are presented. In the third part, a summary and discussion of the findings concludes the thesis. The thesis contributes to civil war research by focusing on civilian actors and investigates how they can resist their civil war environment, being more than refugees or recruitment pools for armed factions. The thesis advances the study of nonviolent resistance in civil wars by introducing a novel dataset of nonviolent activism. The thesis further presents results of an empirical field research project which deals with questions regarding organization of nonviolent activism and third-party support of activists also prior to their nonviolent action events. In doing so the thesis combines different state of the art quantitative and qualitative methods and statistical tools to investigate novel research questions. For the first and second project of the thesis a new and unique disaggregated dataset on nonviolent activism on the event level was constructed to investigate patterns of nonviolent activism during the Nepalese Civil War in unprecedented detail. Utilizing this dataset, the first paper investigates a link between direct civil war violence as grievances and nonviolent activism in a spatial panel regression analysis. In the second quantitative contribution, a multilevel model tests different forms and kinds of nonviolent activism and activist group patterns during the civil war to predict a violent state reaction during the nonviolent event. An additional geographically weighted regression outlines spatial variation of the tested variables throughout the country and substantiates the results from the multilevel model. The third contribution uses qualitative expert interviews to investigate the involvement of nongovernmental organizations (NGOs) to support and train activist groups during the Nepalese Civil War. This contribution utilizes data from an own field research study conducted in Nepal in 2018. The first paper presents a robust, significant relationship between civil war violence and nonviolent action across the Nepalese districts in the spatial panel regression on a yearly as well as monthly basis during the civil war. The paper therefore strongly argues for a linkage between direct civil war violence as grievances with subsequent nonviolent action by civilians. This relationship was already found for battle related violence between conflict factions, but according to this paper now also seems to true for direct violence against civilians. It might explain why we regularly find nonviolent action also far away from the current battlefields in civil wars. The findings of the second contribution show that a high likelihood of disturbance of the public order by nonviolent action events predicts a violent state reaction. Political orientation of activists or number of activist groups during an event also significantly predicted the likelihood of a violent state reaction, but to a lesser extent. The presence of journalists for example to document violence was not related to the likelihood of a violent state reaction. Results of the expert interviews in the third contribution showed in seven examples how nongovernmental organizations supported activist groups, which received counseling in goal formation and selection, and illuminate how conflict-affected parts of the population were supported to become activists step by step, using nonviolent tactics to receive compensation for war crimes and/or demand an end of the war. The findings of this thesis contribute fundamentally to theoretical motivations for nonviolent action during civil wars (direct civil war violence as grievances), and subsequent possible violent state reaction towards nonviolent action events. It extends our understanding of how the decision towards nonviolence and nonviolent action is made, what obstacles are to overcome, but it also outlines in examples how activist groups can receive a helping hand from third parties. In doing so the thesis greatly enhances our understanding of how nonviolent action is facilitated during civil wars, what motivates civilians to do it, what state reaction is to be expected, and what kind of support channels may exist.

What are the social legacies of civil war and how do they differ for men and women? Despite a growing body of research on civil war consequences, the social legacies of conflict remain among the least understood impacts of war. Furthermore, empirical evidence on the distinct effects wartime violence has on men and women is scarce. Quantitative research mostly overlooks the gendered experiences, consequences, and potential benefits of conflict. The aim of my dissertation is to contribute to research on gendered civil war legacies by combining observational and experimental micro-level evidence in different post-conflict settings. The overarching question of my dissertation is addressed in three self-contained essays which test theories of the gendered impacts of war.

Chapter 2 asks whether civil war (dis)empowers women and explores the causal relationship between civil war and women’s labor force participation by leveraging the arbitrary Côte d'Ivoire--Burkina Faso border as setting for a natural experiment. Using a regression discontinuity design, I find that Ivoirian women are 25 percentage points less likely to work outside the home post-war. I investigate three potential mechanisms and argue that the decline in female labor force participation might be explained by a shift towards more traditional gender norms and a reduction in women's bargaining power.

Chapter 3 investigates social (dis)trust in post-war Sri Lanka and analyzes a list experiment to explore the causal relationship between war-related sexual violence and intra- and inter-ethnic group trust. Combining the list experiment with survey data of the Tamil population, I find evidence that war-related sexual violence affects trust decisions of men and women differently. Although both Tamil men and women lose trust in fellow Tamils, female victims are more trusting towards their ethnic out-group. Possible explanations might be that both context of sexual violence and coping strategies differ by gender.

Chapter 4 focuses on social (dis)integration in the Democratic Republic of Congo and explores how forced recruitment shapes male ex-combatants' violent behavior. Based on survey data, I present evidence that former forced recruits commit significantly more violence against their intimate partners and their children compared to voluntary recruits. Using structural equation modeling, I scrutinize the combatant socialization mechanism to show that more intense exposure to violence as part of armed groups and ensuing mental health problems mediate this relationship.

Taken together, my dissertation highlights the importance of systematic micro-level evidence as well as the significance of applying a gender perspective to conflict research. I argue that it is necessary to account for gendered war experiences and consequences, both in theory-building and methodology, as well as in policy-making. This thesis contributes to different literatures on the consequences of armed conflict, the social implications of violence for individuals, and the risks of recurring violence. In chapter 5, I derive several implications and suggest avenues forward for academic research and policy-making.

dc.title:

This dissertation addresses one of the most serious risks associated with automated decision-making: bias. This is not a new phenomenon, and decisions have always been biased, but automated decision-making multiplies the risks in many ways. The main challenges are: How can we detect biases? Who should be held accountable for biased predictions? And how can biases be mitigated or corrected? The three studies within this dissertation help answer these questions by emphasizing the importance of monitoring our own machine learning (ML) pipelines, auditing third party prediction systems, and exposing the potential abuse of predictive algorithms when given sensitive data.

The first paper (section 2) addresses the question of how to direct ML users to high-performing, robust, and fair models. ML systems have been shown to harm human lives via discrimination, distortion, exploitation, or misjudgment. Although bias is often associated with malicious behavior, this is not always the case. Inductive biases, for example, such as knowledge about parameter ranges or priors can help to stabilize a model optimization process. Furthermore, decomposing into statistical bias and variance, allows for model selection with minimum future risk. Since "all models are wrong, but some are useful", we should analyze as many biases in ML as feasible before putting faith in our predictions.

The second paper (section 3) addresses the question about how to audit recommender bias on social media. The goal of this experiment is to quantify the causes of algorithmic filter bubbles by analyzing amplification bias in the recommender system of Twitter. Using simulation of human behavior with bots we can show that 'filter bubbles' exist and that they add an additional layer of bias to 'echo chambers'. More precisely, the algorithm responded far more strongly to bots that actually engage with content than to bots that just follow human accounts. This demonstrates that the Twitter algorithm significantly depends on human interactions to adapt to preferences of its users. This has serious consequences since users may be unaware of the large personalization bias that happens when they like or share content.

The third paper (section 4) addresses the question whether online communication is predictive of offline political behavior. We can predict the party affiliation and turnout likelihood of a person with fair accuracy using a unique dataset consisting of thousands of ordinary citizens, including their Twitter statuses, integrated with public US voter registration files. Our results show social media communication is sufficiently biased to provide information about attitudes and political behavior of an average person in the real world. We demonstrate how, in addition to us, political, commercial, or bad faith actors may acquire this sensitive data to build prediction models, for example, to influence a customers retail journey or perhaps worse discourage them from voting on scale.

Biases can limit the potential of ML for business and society by cultivating distrust and delivering distorting or discriminating results. However, if our societies can (1) implement effective data privacy regulations (2) require internal debaising steps and encourage external independent auditing (3) educate the broader public of biases and ways to report them (4) and invest in training interdisciplinary computational scientists, we may be better prepared for negative consequences of the next industrial revolution.