Patrick Parschan

About

I am a doctoral researcher in Computational Communication Research at the Department of Media and Communication, LMU Munich. My work connects software and research infrastructure with computational methods for studying communication and society.

My dissertation, supervised by Prof. Mario Haim and Prof. Barbara Plank, compares computational approaches to estimating political positions from text. I am currently developing an experimental framework for comparing 17 text-scaling methods; the dissertation is planned for submission in March 2027.

Alongside my research, I design and teach courses in programming, data analysis and AI. I enjoy helping students move from understanding code to building, testing and evaluating complete systems.

Research

My research focuses on how computational and AI-based methods can be compared, reproduced and used responsibly. It brings together three areas:

• Text analysis and political measurement: how statistical, embedding-based and large language model (LLM) methods represent political positions, and how the validity of those measurements can be assessed.
• Reliable AI and software evaluation: how models, prompts, versions and system configurations affect quality, reproducibility, errors and computational cost.
• Computational communication research: how search, generative AI and algorithmic systems shape the information people encounter.

Research software and infrastructure

At Munich Media Monitoring (M3), I help develop and operate research infrastructure for large-scale collection and computational analysis of media content. My work includes database architecture, API design, CI/CD and AI-based analysis pipelines, and GPU computing infrastructure. The production stack includes SQL, Docker, Argo CD, Kafka and RabbitMQ.

For my dissertation, I built an experiment framework that brings 17 text-scaling methods together across different languages and runtime environments. It records configurations and supports systematic comparisons of model outputs, errors and compute requirements. I am also developing a comparison tool for AI and machine-learning algorithms.

Teaching and software development

I teach programming and data analysis alongside machine learning, NLP and AI. My teaching connects core concepts to practical development work: students build and explain solutions, use version control and tests, and learn to inspect AI-generated code for correctness, maintainability and security. I have taught in German and English, designed asynchronous online teaching, and organized contributions from practitioners in media and technology.

My technical work spans Python, R, JavaScript, C++, SQL, APIs, databases, containers and CI/CD. My social-science training helps me connect engineering decisions to questions of measurement, validity and real-world use.

Selected teaching

• 2026/27: Seminar on political communication and AI; planned sessions on AI-assisted programming and statistical programming with Positron and Posit Assistant.
• 2026: Data analysis with R (LMU Munich).
• 2025/26: AI meets political communication: How humans and machines interpret political positions (designed and taught at LMU Munich).
• 2024/25: Scientific working and an English-language seminar on AI and machine learning for analysing financial and economic news.
• 2023/24: Artificial intelligence – a look under the hood and applications in practice, with guest contributions from media and technology organisations.
• 2018–2021: Introductory statistics and data-analysis seminars in R and SPSS at the University of Regensburg; an R course independently redesigned for asynchronous online teaching during 2020/21.

I have also designed workshops in machine learning and AI-assisted programming for doctoral researchers and postdocs, and supervised bachelor’s and master’s theses on computational text analysis and generative AI. The complete teaching history appears below.

Selected software projects

• Text-scaling experiment framework: shared infrastructure for running and comparing 17 statistical, embedding-based and LLM-based approaches to estimating political positions from text.
• AI and machine-learning comparison tool: a tool in development for comparing algorithms and their configurations.
• Munich Media Monitoring: research infrastructure for collecting and analysing media at scale, developed and operated in a multidisciplinary team.

Recent work

• 2026 · KONVENS: Making Political Text Scaling Comparable: Infrastructure and Hyperparameter Sensitivity for 17 Algorithms, forthcoming in the KONVENS workshop proceedings. Preprint
• 2026 · SEASON: Using generative AI instead of search engines? Comparison and implications in the context of five state elections in Germany. Conference paper
• 2026 · ECREA: Measuring party positions from campaign tweets: What 17 text-scaling methods (do not) agree on, presented at the conference pre-conference on the evolution of election campaigning on social media.
• 2026 · AANZCA (accepted): Co-author on Children’s Information Environments and Right to Information in the Wake of Social Media Age Restrictions and Ok Google, Who Should I Vote for in the Next Election?, to be presented in Perth in November.
• 2025 · EMNLP Findings: Co-author of PolBiX: Detecting LLMs’ Political Bias in Fact-Checking through X-phemisms.

Selected presentations

• 2026 · DACH-CSS, Vienna: The Ideological Machine: From Statistical Language Models to Large Language Models in Text-Based Political Position Estimation.
• 2026 · DGPuK, Dortmund: Munich Media Monitoring (M3), with Mario Haim.
• 2025 · COMPTEXT, Vienna: Computational ideal point estimation using textual data: a systematic review of existing algorithms, with Charlott Jakob.
• 2023 · ICA, Toronto: From auditing to testing: Adequate experimental designs to study algorithmic content curation, with Mario Haim and Julian Unkel.

Research interests

• Political position estimation, text as data and measurement validity
• Natural language processing, transformers, word embeddings and large language models
• Reproducibility and evaluation of AI systems and software
• Algorithmic curation, search engines and generative AI in public communication
• Computational social science and agent-based testing

Academic background

I hold an M.A. in Digital Humanities from the University of Regensburg (2021), where my thesis evaluated transformer models for classifying information needs in cooking dialogues. I also hold a B.A. in Political Science with a minor in History (2019) and completed 40 ECTS of university IT training, including programming, software development, machine learning and data analysis. Before joining LMU, I worked in political science at the University of Regensburg (2019–2021) and on the SPARTA project at the Bundeswehr University Munich (2021).

My academic training combines empirical social research with hands-on programming and software development. I have also completed university teaching development in course design, assessment, advising and evaluation, and research stays at the University of Vienna and the Hebrew University of Jerusalem.

My earlier publications and conference contributions appear under the name Patrick Schwabl; author names below follow the published records.

Publications

Journal papers

Conference proceedings

• Parschan, P. (2026, forthcoming). Making Political Text Scaling Comparable: Infrastructure and Hyperparameter Sensitivity for 17 Algorithms. Proceedings of the 21st Conference on Natural Language Processing (KONVENS), Volume 2: Workshops. Preprint.

Book chapters

Conference posters and papers

Other

University teaching & other courses

Courses and workshops (2022–2026)

• 2026/27 · LMU Munich: Seminar on political communication in the age of AI, including planned practical sessions on AI-assisted programming and data analysis.
• 2026 · LMU Munich: Data analysis with R.
• 2025/26 · LMU Munich: AI meets political communication: How humans and machines interpret political positions (designed and taught).
• 2025 · ICA pre-conference hackathon: Workshop on Cursor and AI-assisted programming for doctoral researchers and postdocs.
• 2024/25 · LMU Munich: Scientific working; How money talks: Stock market and economic news analysis with artificial intelligence and machine learning in communication research (English, designed and taught).
• 2023/24 · LMU Munich: Artificial intelligence – a look under the hood and applications in practice (English, designed and taught; practitioner contributions from media and technology organisations).
• 2023 · ICA pre-conference hackathon: Introductory machine-learning course for doctoral researchers and postdocs.

Supervision and academic development

I have supervised five bachelor’s theses since 2022 and three master’s theses since 2023. Topics include word-embedding analysis of party manifestos and bias in generative AI. Since October 2025, I have been a mentee in the OTH Regensburg Academic Steps programme, which prepares researchers for professorships at universities of applied sciences. I also completed university teaching certificates covering course design, assessment, student advising and evaluation.

Earlier university teaching

• Crash course introduction to machine learning at the preconference hackathon during ICA 2023
• Introduction to data analysis with SPSS at LMU Munich, 2023
• Political communication during the Ukraine war at LMU Munich, 2022
• Tutor for introduction to Communication Science at LMU Munich, 2022
• 3 Day course on machine learning at Bamberg Graduate School of Social Science, 04-06 April 2022
• Introduction to data analysis with R at University of Regensburg, 2019, 2020, 2021
• Introduction to data analysis with SPSS at University of Regensburg, 2018, 2019, 2020
• Tutor for introduction to Political Science methodology at University of Regensburg, 2018

Services

Honorary offices

• Student and early career representative (2022–2024), Computational Methods Division of the International Communication Association

Conference Reviewer

• International Communication Association conference (ICA): 2022, 2023, 2024
• International Interdisciplinary Conference on the Quantitative and Computational Analysis of Text-, Image- and Video-as-Data (COMPTEXT): 2023, 2024, 2025
• Jahrestagung der Fachgruppe Methoden der DGPuK (DGPuK Methods Conference): 2022
• Quality & Quantity (journal); ICA: 2025

Other

I host and organize the in-person Computational Exchange at LMU, a coding and methods workshop that meets three times per semester. I also co-organized the pre-conference hackathons at ICA in 2024 and 2025 and joined the organizing team in 2026.

I am a member of the International Communication Association, the COMPTEXT Association and the German Nachwuchsgesellschaft für Politik- und Sozialwissenschaft (DNGPS). At LMU, I contribute to the Mittelbau and to a working group on regulations for cumulative dissertations.

Funding and scholarships

• Young Talents programme, Bavarian Research Institute for Digital Transformation (bidt): €12,000 doctoral funding over four years.
• Research travel: €2,000 from the Bavarian Research Alliance for a research stay in Israel and €1,000 through the Erasmus programme for a research stay in Vienna.