What is the epistemological structure underlying data collection, selection and processing in AI systems?

Hey there! As part of my master’s thesis, I am seeking AI developers, ethics officers and others for an interview within the timeframe of May 19th to June 29th, 2025. Let me treat you for a coffee, tea or a classic German Döner!

Abstract of the Master’s Thesis

The starting point of this master’s thesis is the increasing availability and everyday integration of AI-based technologies such as Large Language Models (LLMs) and image generators, which raise central questions about the underlying power-specific knowledge systems. As publicly available technologies, do LLMs and image generators bear a moral responsibility for the way they represent reality? If so, what knowledge should be presented and in what form to ensure responsible representation – understood as sensitive, just, inclusive, critical, diversity-oriented, and power-sensitive?

At the core of the issue lie technological actors, such as companies like OpenAI, Google, X, and Meta. These entities not only influence the interpretation and presentation of information through the development of text and image-generating systems, but also actively impact knowledge production through the structural design of their models. Consequently, they frequently reproduce ideological models such as techno-solutionism, techno-realism, or capitalist realism. Can such a responsible representation of reality be achieved solely through a data-driven supplementary strategy, or is the epistemological framework within which data is collected, selected, and interpreted itself an integral component of the problem?

Within the interdisciplinary field of Artificial Intelligence and Visual Anthropology, this master’s thesis delves into the epistemological framework underpinning data collection, selection, and processing within AI systems. The study draws upon generative AI systems, such as language models and AI-powered film software, as illustrative examples. In this context, AI developers are qualitatively and empirically interviewed to assess their comprehension of ethical principles within the organisation and their adherence to principles of justice as integral components of knowledge generation.The objective of this research is to critically and normatively analyse the epistemological foundations of machine knowledge development. The primary focus is on the ethical and cultural contextualisation of AI-driven representations, particularly within the domain of computer vision. The work serves as a valuable contribution to a reflective examination of the material, physical, and local conditions that underpin knowledge production in the era of artificial intelligence.

Requesting for interviews: Target Audience – Who am I looking for?

I am seeking AI developers, ethics officers, executives, or data scientists for a 45-minute interview. Ideal candidates are those interested in discussing corporate ethical guidelines, training data for generative AI development, and the structural and strategic outlook of their organization. While not a requirement, I would especially welcome the opportunity to speak with individuals from social or cultural science backgrounds and/or those who identify as part of a minority group in terms of sex, gender, class, ethnicity, or nationality.

Interviews can be conducted in German, English, or French.

All findings are subject to a confidentiality agreement and will not be published without the explicit consent of the interviewee and the company. It is possible to omit specific research questions and/or anonymize results for both the individual and the organization. Throughout the research process, I will maintain close communication with all participants and their respective organizations.

---

Felix Keilhack

fkeilhack@yahoo.com

---