Blog

Computing Resources for WI

Roland Toth, Jochen Knaus

As digital methods advance rapidly, quantitative empirical research requires greater computing power. This includes complex statistical analyses, model training, and operating generative AI models. The necessary hardware is expensive and challenging to maintain, particularly at the institutional level (e.g., due to high temperatures in HPC clusters). Since not every institution can (or should) set up such hardware independently, resources are shared through collaborations with other institutions.

Against this background, the Weizenbaum Institute is exploring options to enable both low-demand computing tasks (e.g., a virtual machine for background web scraping) and high-demand tasks (e.g., running current large language models) for its researchers. Recently, the following options have been identified (some of which are only available to Berlin scientists):

Datenzentrum Berlin (Berlin University Alliance)

  • Not established yet, but planned

de.NBI Cloud (Deutsches Kompetenzzentrum Clouddienste)

  • free to use with scientific project proposal: https://datenkompetenz.cloud/
  • “Self service” virtual machines
  • currently trying to support digital humanities and social science

FUB-IT (Freie Universität Berlin)

  • Free to use (FU account required; guest accounts are available)
  • High-performance cluster (HPC)
  • Software partially included, some can be requested

Gauss Computing Center

  • Up to very large scale projects (both HPC and AI) on proposal in different calls: https://www.gauss-centre.eu/for-users/hpc-access
  • Several equipment up to supercomputers SuperMUC I and II usable

Gesellschaft für wissenschaftliche Datenverarbeitung mbH Göttingen (GWDG):

Galaxy Server

  • Community-driven web-based analysis platform for life science research
  • Free to use batch system with many tools usable by browser, huge data sizes are no problem
  • Although coming from Bioinformatics, there are many tools for text processing, audio and image analyses, as well as statistics can be put together in pipelines.

HPI (Hasso Plattner Institut) UP Potsdam

Zentraleinrichtung Campusmanagement (TU Berlin)

  • Free to use (TUB account required)
  • High-performance cluster
  • Software partially included, some can be requested

Zuse Institute Berlin (ZIB)

  • Support from the state of Berlin, linked to BUA, e.g. BI-FOLD also calculates there
  • Provides tailor-made solutions, e.g. can make own instances of HAWKI, LibreChat or similar available with WI licenses
  • Resources available only by collaboration.

Apart from these external offers, the Weizenbaum Institute itself already provides some services to its researchers:

  • Virtual machines
  • Jupyterhub
  • Gitlab

Researchers who are interested in using any of these services can follow the instructions on the according websites for external services. For internal services, instructions are provided in the internal WI Wiki.

Disclaimer: This list will be updated regularly. If you know of any other resources that are available to members of the Weizenbaum Institute, please let us know.

Mobile methods panel at ICA 2025

Roland Toth

Every year, the International Communication Association (ICA) organizes a conference where communication scholars from around the globe present and discuss their current research. In June 2025, the conference took place in Denver, USA, and many Weizenbaum Institute researchers contributed to it. Among others, Methods Lab member Roland Toth, together with colleague Jakob Ohme, former fellow Joseph Bayer (Ohio State University), and his team members Razieh Pourafshari and Yifei Lu, organized a conference panel called “Miss the Mobile Methods Turn? Reflecting on the Challenges and Potentials of Mobile Communication Science”.

Roland Toth outlines the issue with mobile methods

Many social scientists use mobile data collection methods, such as the Ecological Momentary Assessment, data donations, or event log data, but there are hardly any standards or guidelines to work with. The panel addressed this lack of standardization and accessibility of mobile methods within and outside of communication science. In the introduction to the panel, the issue was outlined and attendees were asked to complete a short survey about their expertise regarding different mobile methods, problems they recognized, and whether and to what extent they are willing to contribute to future initiatives to improve the situation.

Gabriella Harari presents issues with capturing the context of smartphone use

After the introduction, experienced mobile communication scholars presented examples of empirical research that highlight current practices, challenges, and future perspectives associated with mobile methods. Following these short talks, the presenters, attendees, and organizers discussed possible standards, priorities, and endeavors to strengthen this growing research area, as well as how the community can organize to support the use of mobile methods to the Communication field at-large.

Jakob Ohme moderates the general discussion with the presenters, attendees, and organizers

The panel was received well and paved the way for further steps to gradually tackle these issues. The discussion shed light on the wide interdisciplinary use of mobile methods and their relevance, both in and beyond communication science, that goes along with that. It showed that there currently is potential for all panel presenters, attendees, and organizers, but also other communication researchers, to contribute to the advancement of mobile methods across fields, which the Weizenbaum Institute aims to participate in. Accordingly, based on the results of the survey, future initiatives such as white papers, methods reviews, or workshops addressing the current state and standardization of mobile methods will be planned. Stay tuned for these in the future!

Spotlight: Dein Feed, deine Wahl

Diana Ignatovich

Whether it’s unwinding after a long day or killing time in line for coffee, social media has become a major source of entertainment and connection. The immediate satisfaction it often brings can boost the mood, but it’s not just entertainment we’re absorbing. Amidst the lighthearted content, is exposure to a wide range of information, much of which we may not fully process. Research suggests a link between high social media usage and lower self-control, which may also perpetuate processes of dissociation that many refer to as mindless scrolling. As we passively absorb media from smiling koalas to political protesting, how are we actually understanding the more consequential information and even shaping our opinions?

Lion Wedel and Jakob Ohme consider these influences in collaboration with Bayerischer Rundfunk, Stuttgarter Zeitung, and the University of Zürich in their project, Dein feed, deine Wahl, (Your Feed, Your Choice/Election). In this ongoing initiative, TikTok users are encouraged to donate their data in an anonymous manner and in turn receive a direct analysis of the political content and parties appearing in their feed.

The findings depict the daily number of TikTok videos posted by political parties

Jakob and Lion shared their perspectives in an interview, offering insight backed by their expertise in political communication and digital media research.

To begin, Lion and Jakob discuss the broader implications of media on shaping one’s ideological framework. Specifically, which types of content shape opinions the most, and how do they attract attention?

To this, Lion responds,

“The more time you spend preparing a video, the less popular it gets. Like quick and dirty typically works better.”

This was in reference to a podcast titled, Was tun? Die Strategien hinter dem Comeback der Linkspartei (1/3): Wie Heidi Reichinnek die AfD auf TikTok überholte, in which Felix Schulz, social media manager in Heidi Reichinnek’s office, explores the strategic use of TikTok to engage young voters. From this, Lion highlights the influential processes through which opinions are likely to form online. He quotes the manager’s technique, asserting that his videos did so well because they managed to make a compelling statement in the first one to three seconds of the clip. Even in regard to political news, Lion describes, “It does not matter if it’s true or false, or if it’s catchy or misleading, you just have to get that attention grabbing moment.” He concludes by suggesting this content that keeps its audience engaged to the video’s end, is more likely to shape opinions.

But what are the implications to the broader democratic landscape in Germany if people form political opinions based on whatever content is most stimulating in their feed, regardless of its credibility? Can the use of traditional media repertoires play a role in fact-checking political discourse?

Statistics of the 2025 Weizenbaum panel report exploring last year’s political participation in Germany may suggest otherwise. Since 2021, there has been a decline in the use of traditional media sources such as newspapers and radio for news consumption, while internet usage has remained consistently stable. Moreover, over 60% of voters up to the age of 30 receive their political information from social networking (Schöffel et al., 2025), suggesting platforms such as TikTok are inevitably shaping how many engage in political debate.

Given recent civic tensions in Germany, including shifts in numerous elections and widespread protests, there appears to be a similar polarization in the spread of social media information. As many are more likely to express their opinions than change them, and online discussions often dominate in one direction of opinion (Xiong & Liu, 2014), raising awareness of these influences not only to our personal perspective but also the broader political environment becomes critical.

According to Jakob, the nature in which algorithmic selection processes function, “affect political landscapes to an extent that we probably ten years ago did not think was possible.” He outlines his interpretation of how algorithmic platforms contribute to political shifts, attributing the cause to the often passive behavior of individuals online, where their feeds reflect the content they prefer. As seen in previous research, social networking individuals may feel a reduced sense of self awareness and volition (Baughan et al., 2022), often consuming the content that is most interesting and aligns with their worldview. Jakob concurs, stating, “We can all function like this from time to time.”

He continues,

“There is content that works better with the algorithm and especially on TikTok. We can see that as soon as something works, it will capture a lot of attention, but as soon as something doesn’t work, it will completely drown. There are certain kinds of political content that work better and that is the emotional appeal, opinionated, negative and extreme information.”

Jakob asserts that certain political parties, especially those on the right-wing, are more adept at taking advantage of this dynamic, as their content tends to perform better online. Algorithms may as a result appear to favor them, not because of any inherent political bias, but because these parties successfully leverage a system that operates based on audience engagement patterns to maximize their impact.

Consequently, the Dein Feed, deine Wahl initiative establishes an objective foundation for identifying usage patterns within TikTok’s algorithms that ultimately contribute to the broader political climate. Jakob emphasizes the significance of these usage trajectories, aiming to explore their association with voting decisions and provide a descriptive overview of the extent to which individuals encounter political party-related content. In addition to this, he expresses the ambition to reverse-engineer algorithms to mitigate their effects. By examining how usage patterns influence algorithmic decisions and lead individuals to encounter more of the same content over time, researchers are better equipped to address these dynamics. This approach is especially essential given that analyzing video content and user interaction data has been largely unexplored due to its methodological complexity.

Overall, this project paves the way for regulating social media platforms in the long term, with the hope that it remains accessible for prospective political debates and elections. However, to move forward, Lion identifies the collection of data donations as one of the first priorities. The quality of the analysis significantly depends on the number of public contributions volunteered to their data donation lab.

Therefore, if you’d like to support the project or are curious of the political makeup in your own feed, please donate your usage here!

New Publication: Ethics of Data Work

Diana Ignatovich

Machine learning is becoming increasingly central to academic research, yet it often depends on data workers in exploitative conditions whose contributions are largely overlooked in ethical guidelines and unacknowledged within the academic community.

Last year, the Methods Lab outlined the aims of a project to target this issue in a short blog post. We’re now excited to announce the resulting published discussion paper: “Ethics of Data Work: Principles for Academic Data Work Requesters.”

This paper builds on the insights of an interdisciplinary group of scholars, practitioners, and data workers, alongside expert workshops held at the Weizenbaum Institute in 2024. It organizes practical principles for engaging more ethically with platform-based data workers, including how to define data work to then address important gaps in current ethical guidelines. The paper therefore offers concrete recommendations and regulations based on the most pressing challenges faced by these contributors. As the rapid development of AI continues to rely on the insight and labor of real people, it’s crucial to reflect on how research is conducted to ensure those workers receive proper acknowledgment for their role. This discussion paper calls for commitment to fair treatment, transparency, and meaningful support to make ethical data work a consistent part of the machine learning research process.

If you would like to learn more about the experiences and working conditions of these data workers, check out our blog post featuring creative projects from the Data Workers’ Inquiry!

New Publication: Extracting smartphone use from Android event log data

Diana Ignatovich

Back in October 2024, the Methods lab shared a preprint of a study by Methods Lab member and data scientist, Roland Toth, and former research fellow, Douglas Parry, exploring how to isolate meaningful measures of smartphone use from Android event log data. We’re now pleased to announce that this work has been peer-reviewed and published in the journal Computational Communication Research.

The article titled “Extracting Meaningful Measures of Smartphone Usage from Android Event Log Data: A Methodological Primer” outlines a practical and reproducible step-by-step guide for deriving objective indicators of human usage from raw mobile data, offering valuable insights for research in social science and related disciplines. It details the extraction of key usage metrics through written explanations, visual aids, and pseudo-code. The paper is a vital resource for researchers seeking to understand patterns of mobile phone engagement and its implications in today’s rapidly evolving digital environment.

Workshop: Introduction to MAXQDA

Methods Lab

Join us for the workshop Introduction to MAXQDA, designed for all researchers, students, and professionals interested in qualitative data analysis. On May 28th, 2025, at the Weizenbaum Institute, certified MAXQDA trainer Dr. phil. Aikokul Maksutova will lead a basic yet comprehensive workshop introducing the software’s core features, aligning with the key stages of digital qualitative research.

This event will offer guidance on MAXQDA’s essential tools for documenting, coding, and analyzing qualitative data. Participants will become familiar with navigating the Code System and a range of additional features, such as functions for exporting data, linking memos, and generating visualizations. Each segment will include hands-on activities using various datasets, enabling participants to confidently apply the skills they’ve learned on their own.

To conclude, special guest and representative of MAXQDA, Ms. Tamara Pataki, will inform participants of the software’s latest innovations and host an open Q&A session.

To learn more, please visit our program page. We hope to see you there!

DeZIM Summer School 2025

Diana Ignatovich

For those interested in strengthening their skills in social research methods, we’re pleased to announce that registration is now open for the DeZIM Summer School 2025 (Deutsches Zentrum für Integrations und Migrationsforschung).

Running from August 12 to 14, the program is free and open to all, offering workshops in both qualitative and quantitative methods. Courses are designed for participants ranging from beginners to advanced, and all are welcome to join multiple sessions. However, space is limited, so we encourage early registration by completing this survey before the deadline on June 30, 2025.

As part of the ongoing collaboration between DeZIM and the Weizenbaum Institute, both institutions share access to each other’s workshops. Through this partnership, we aim to create more opportunities for researchers to develop and strengthen their methodological expertise.

To learn more about additional upcoming workshops, check out our Methods Ticker!

Workshop Recap: Introduction to Programming and Data Analysis with R

Diana Ignatovich

A third edition on the Introduction to Programming and Data Analysis with R workshop took place on March 12th and 13th, 2025. Roland Toth with the Methods Lab at the Weizenbaum Institute engaged almost 20 participants with essential methods of data analysis via comprehensive coverage of fundamental R programming concepts and techniques.

Roland asks participants about their former experience with programming

On the first day, Roland guided participants through the basics of R syntax and its integration with Markdown/Quarto in an interactive environment. This included the very basics of programming like functions, objects, and indexing, but also data-related practices like data wrangling, sanity checks, and simple statistical analyses. Among others, participants also gained insight on managing warnings and errors that might stunt the process of coding throughout projects.

On day two, after an introduction to data visualization techniques, participants put their learning into practice: They explored provided survey data and developed a research question, so they could prepare and statistically analyze the data accordingly in R. The result was a reproducible HTML report on the reasoning behind the research question, all data wrangling steps, an exploration of the data set, the analysis, and the results including an interpretation. Attendees also supported each other’s progress whenever possible, while Roland offered personalized guidance.

The workshop alternated between lecture-like and interactive formats

The workshop concluded with a thorough review of useful functions and packages in R. Throughout the event, participants were encouraged to ask questions freely and frequently, and they took the opportunity. The Methods Lab would like to give a great thanks to all guests for their attendance and lively participation!

Career Tutorial: LLMs for all Expertise Levels (March 7, 2025)

Methods Lab

In a joint effort, the Career Development and the Methods Lab are excited to announce the hybrid “Career Tutorial on LLMs for all Expertise Levels”. In this tutorial, beginning with fundamental concepts of LLMs and in-context learning, we’ll address the “Needle in the Haystack Problem” and compare ultra-long context models with RAG approaches. Through practical demonstrations, participants will gain hands-on experience with RAG’s core functionalities and understand its objectives. The session delves into scaling solutions using vector databases and advanced implementations, including chunking strategies, hybrid RAG, and graph-based RAG architectures. We conclude with an overview of emerging trends, examining agentic RAG and the integration of reasoning models in deep research applications. This comprehensive exploration equips attendees with both theoretical knowledge and practical insights into the latest developments in AI language models.


For more information, visit our program page. We are looking forward to your participation!

Workshop Recap: Introduction to Git

Diana Ignatovich

On February 6th, 2025, LK Seiling facilitated a workshop for an Introduction to Git, with support from Sascha Kostadinoski and Quentin Bukold. This was co-organized by the Methods Lab and took place at the Weizenbaum institute. The hybrid event provided a thorough overview on the foundation of Git and its relative platforms for about 30 participants. 

Firstly, Git was introduced for its general relevance. Seiling explored the qualities of its version control system and the advantages of efficiently managing changes to files. Its widespread use and accessibility were also highlighted by the software’s free and open source application. At its core, Git enables collaborative work by allowing concurrent adjustments to files by multiple participants and offers a system to track the changes made without requiring alterations to the original file.

Next, participants were invited to open the Terminal and guided through some basic commands. To this end, commands for traversing directories, creating, moving, organizing, and deleting files were explained and demonstrated in detail.

LK Seiling explains how to stage and commit changes

In the second hour of the workshop, Seiling encouraged participants to implement these basics by imagining the context of a classic Python project, one that might require collaborative engagement. Here, Python scripts were saved, renamed and staged accordingly to git messages and configurations. The principle git practices were emphasized to remind the audience of when and how to commit changes to the previously specified local repository. Furthermore, Seiling prepared guests to make requests when merging work, added description templates for joint projects and generally taught the features of use for group collaboration. 

This was followed by instructions on the key functionality of Git, such as the Git repository, Git commands, branches, and conflict resolution. For instance, the branches gave insight into how to leverage simultaneous work done separately from the overall code base. This is especially beneficial for feature development while also helping to streamline the process of reviewing changes before merging. Throughout this instruction, commands were given to switch branches and merge scripts in the terminal, which was operationalized with a quickly constructed example. Seiling also provided necessary information on managing repositories, including visuals of the basic workflow and linkage between local and remote repositories, either for individual or collaborative effort.

For those curious when to use which Git platform, Bukold jumped in to detail the major differences between Github, Gitlab and Git.

Terminal commands are used to perform actions with Git

Later, Seiling explored some advantageous elements of the GitLab platform, accessible free of charge to Weizenbaum researchers, by describing the repository graph, issue tracking and project management tools. To elaborate, the repository graph structures insight into how a participant makes a contribution or change by arranging branches to show merges or commits, particularly relevant for collaborative code projects. In case of software malfunctioning, the issue tracking feature allows one to see who is working on what branch for an update on the progress of the problem. Finally, Gitlab’s management tool was outlined for instances of assigning work, applying tags to notify when projects are finished and to open or close potential issues. 

To close, Kostadinoski briefly summarized the basic elements of Git, along with its implications in data work, such as for software development and research. He simplified key terms and embraced questions in a Q&A. Seiling joined in, encouraging participants to “learn by doing” and stay connected with each other via Weizenbaum associated Github accounts for future internal coordination. 

Throughout this workshop, participants were presented with various tasks and benefited from frequent recaps that highlighted key points, ensuring a solid understanding of the material. Attendees both online and in person freely asked questions and received support from instructors. Therefore the Methods Lab would like to give a huge thank you to LK Seiling, Sascha Kostadinoski and Quentin Bukold for their clear instruction on the foundations of Git and for facilitating such an engaging environment for all participants.