Research
15556
page-template-default,page,page-id-15556,bridge-core-2.3.2,qode-page-transition-enabled,ajax_fade,page_not_loaded,,vertical_menu_enabled,qode-title-hidden,qode_grid_1300,side_area_uncovered_from_content,qode-content-sidebar-responsive,qode-theme-ver-29.3,qode-theme-bridge,disabled_footer_top,disabled_footer_bottom,qode_header_in_grid,wpb-js-composer js-comp-ver-6.10.0,vc_responsive,elementor-default,elementor-kit-15264

Objectives

Applying qualitative and quantitative methods to assess large-scale text data

Conceptualising and measuring social concepts

Applying automated text and social network analysis

Reviews

EPJ Data Science

ACM Web Science Conference

EDML Evaluation and Experimental Design in Data Mining and Machine Learning

ASONAM Advances in Social Network Analysis and Mining

Digital Journalism

Academic Service

Local Chair Organizing Team, Virtual Sunbelt Conference, International Network for Social Network Analysis, June 2020.

Technical Committee Member, The twelfth International Conference on Social Media Technologies, Communication, and Informatics, SOTICS 2022, October 16-20 2022, Lisbon,

Projects

KI CP@TUM: TUM KI Competence Program – Professional Learning Communities for the Systematization of AI-Specific Study Programs at TUM

The KI CP@TUM program is designed to give students the opportunity to acquire a fundamental understanding of how artificial intelligence (AI) works and its implications on a social and ethical level – in the field of education also with a view to the use of AI in schools and classrooms, for example in the area of learning analytics and intelligent agents. To this end, students should be enabled to interact with AI, to develop an awareness of the possibilities and limitations of AI and to use it responsibly and reflectively. This is especially true for students of non-technical courses of study, some of whom have only implicitly come into contact with AI so far. The project therefore aims to adapt new and existing teaching/learning modules on AI also for primarily non-technical courses of study and to work out transferable structures for the packaging of offerings. The new teaching/learning modules created within the framework of the project are to be consolidated throughout TUM in the form of an AI certificate. At the TUM School of Social Sciences and Technology, Department Educational Sciences, students are to implement (prepared) AI prototypes themselves. An example of this for the grammar school teaching profession is the evaluation of drone images for the classification of vegetation in biology. Equivalent fields of application can be determined for physics and chemistry as well as for the vocational teaching profession. In all domains, learners should always be provided with AI scenarios that can contribute to the achievement of the UN Sustainable Development Goals (United Nations 2015). Students should be able to develop their own school-based teaching-learning concept on AI based on the content of the topic area. Students should be able to recognize and critically question potentials for the use of AI in their future professional fields of action.

Co-designing a Risk-Assessment Dashboard for AI Ethics Literacy in EdTech

Various AI-based technologies are entering the world of education (“AI EdTech”) at an accelerating pace. The introduction of AI EdTech promises to improve learning. However, initial case studies reveal significant ethical and legal risks ranging from bias to privacy threats. Decision-makers struggle to manage these risks systematically and holistically, given the lack of best practice guidance and specific legislation. To fill this gap and foster AI ethics literacy, this project’s principal focus is to develop practical guidance for decision-makers to help them navigate through the process of ethical decision-making in applying AI EdTech. Guided by a set of interdisciplinary research questions, the project works with iterative and agile design-based methods. It will work with current decision-makers to evaluate outcomes through qualitative and quantitive approaches. The expected results are

(1) a tailored risk management framework for assessing risks in AI EdTech,

(2) an AI EdTech Dashboard for qualified engagement with AI EdTEch education scenarios (“vignettes”), functionalities, and ethical/legal risks, and

(3) a design guide for AI ethics literacy.

The project’s impact will contribute to AI ethics literacy relevant to intersectional audiences.

Online Firestorms and Resentment Propagation on Social Media: Dynamics, Predictability and Mitigation

Social media serves as a place to gather information, interact, and form opinions. More recently, online firestorms, fake news and hate speech have shaken our beliefs and hopes about the positive power of social media to their very foundations. While negative emotions are in the core of human behavior, algorithms on social media, enhanced by Artificial Intelligence (AI) can produce and reinforce new dynamics.

In this project, we will address the mathematical modeling of the formation and dynamics of opinions in large groups of interacting people on social media. Our primary objective is to understand the driving factors of social media group level phenomena that lead to negative dynamics and to offer approaches on how to detect, react to, and possibly mitigate these dynamics early on. The fundamental goal is to reveal the possible relationship between the simple “social forces” acting at individual level, being the “first principles” of social interaction or the game rules, and the potential emergence of a global behavior.

The results of our study will provide insights of ethical relevance by discussing responsibility, delegation and control mechanisms in human-AI interacting systems.

Online-Offline Spillovers – Potential real-world implications of online manipulation

This project analyzes the previously unexplored questions of whether people’s online behavior spills over to their behavior in the offline world and what mediates the respective effects.

Employing a two-stage experimental setup, we first use field experiments on social media for online manipulations of our study participants. Second, we study the potential spillovers to our participants’ offline behavior in a laboratory setting. Specifically, we investigate whether attention from others on social media leads to a polarization of people’s political opinions and erodes their commitment to truth. We hypothesize that the treatment group receiving a relatively high levels of attention on social media will show more polarized profiles of political opinions.

Qualiservice – Conceptual tool development for anonymizing text data

The anonymization of qualitative interview data is of high importance. For the purpose of secondary use of data, anonymized data is essential. While automated processes in anonymization tasks are becoming more and more common, we provide a tool that keeps researcher in control of their data. Automated decisions give all-in-one solutions, but studying qualitative interview data depends on the needs of every single researcher. We provide a tool that enables researcher to make individual decisions with the information needed, on the level required. In this report, we propose a solution to anonymize qualitative interview data with the purpose to create own coding schemes and individual abstraction levels. We built a tool that assists in working with textual interview data. By using the tool, processes can be optimized and important information can be obtained at the same time.

Pluralisation and Hierarchisation of Models for Lyric Poetry in the Italian Renaissance

The project seeks to extend its diachronic perspective towards the seventeenth century and, at the same time, to shift its focus onto a dialectics of pluralisation and the positioning of new authorities that is observable not only in the field of poetic theory, but especially in the interaction of theory and poetic practice. This is of particular importance as in the course of the sixteenth century, love poetry is affected by new pluralizing tendencies brought about by the discovery of new models, both practical (the genre of the ode) and theoretical (Aristotle, Longinus).

It is the aim of this project to demonstrate how in the situation described, on the one hand, order is re-imposed by introducing new hierarchical structures and more and more complex systems and, on the other hand, conflicts are avoided by ascribing new meanings to canonical texts and by obscuring contradictions both in theory and lyrical practice. New efforts at establishing order often create an apparent but superficial order which, thus, harbours the potential for further pluralisation. In the wake of Aristotelianism, there is a new tendency towards the elaboration of a poetological system that no longer uses a single author as its sole model but integrates model authors such as Petrarch as exemplifications of a set of rules existing independently (Tasso) or even has to discard them as inappropriate (Tassoni, Marino).

It will be a long-term aim of the project to integrate its findings into the larger attempt to draw up an archaeology of the seemingly unified, but highly heterogeneous concept of ‘lyric poetry’ which can first be discerned in certain publications just before and after 1600 (e.g. Marino, La Lira) and which remains problematic even today.