Algorithms are playing an increasingly prominent role in many public service domains, ranging from healthcare to criminal justice. It is important that the use of algorithms is conducted in a responsible and trustworthy manner to mitigate concerns (e.g., biased, unfair or untransparent decision making) and make use of the value (e.g., efficiency, productivity or safety) that algorithms can offer to the public sector. Transparency is frequently mentioned as an important factor in building, maintaining and strengthening citizen trust in the use of algorithms by governments. Still, there is ongoing debate about what transparency should entail in practice and how it actually influences trust. This dissertation seeks to clarify how transparency about algorithm use by public sector organizations influences citizen trust. This dissertation answers the question “How does transparency affect citizen trust in algorithm use by government organizations?”, using a literature review, document analysis, interviews and survey experiments. Throughout the dissertation, the police serve as a recurring context for exploring broader phenomena related to algorithm use in the public sector. At the same time, I highlight the broader relevance of my findings for public sector organizations across all studies. The dissertation begins by developing a framework to empirically examine the relationship between transparency and trust in government algorithm use. This multi-level framework highlights that transparency is required not only at the level of the algorithms themselves (micro), but also in how government organizations use them (meso) and in the regulations that govern their use (macro). The findings of the empirical studies reveal two mechanisms through which algorithmic transparency can foster citizen trust: a communicative (direct) route and a disciplining (indirect) route. The first route concerns the communicative function of transparency. This involves what organizations communicate externally about their algorithms: explaining how they work and produce their outputs, and how they are embedded within organizational and institutional frameworks. This could directly impact citizen trust. The second route involves the disciplining function of transparency. The requirement to provide transparency encourages public organizations to engage in critical reflections, which can lead to improved algorithmic decision-making processes (a “disciplining effect”) and to better substantiated algorithmic decisions. This could indirectly impact citizen trust. Furthermore, the findings in this dissertation demonstrate that the relationship between transparency and trust occurs not only within individual levels but also through an interconnectedness across the micro, meso and macro levels. Distinguishing between micro, meso and macro levels is relevant, as it offers both theoretical and practical insights into how transparency can be operationalized in theory and implemented in practice. This dissertation end with a call for transparency-by-design: I hope that the insights will encourage public organizations to think about how they can be transparent about the functioning and use of their algorithms before they develop and deploy an algorithm. If an organization cannot explain why an algorithm comes to a certain decision, or how employees use it in decision-making processes, the question arises as to whether it should be used at all.