Rule(s) of Recommendation: What the making of a recommender system can tell us about the difficult relation between social order and machine learning

Pöchhacker, Nikolaus Helge

TUM School of Social Sciences and Technology

Back
Back to start of result list
Permanent link for displayed object

If you experience problems opening the document, please try this link.

Original title:: Rule(s) of Recommendation
Original subtitle:: What the making of a recommender system can tell us about the difficult relation between social order and machine learning
Translated title:: Rule(s) of Recommendation
Translated subtitle:: Was uns die Entstehung eines Empfehlungssystems über das schwierige Verhältnis zwischen sozialer Ordnung und Machine Learning sagen kann.
Author:: Pöchhacker, Nikolaus Helge
Year:: 2021
Document type:: Dissertation
Faculty/School:: TUM School of Social Sciences and Technology
Advisor:: Passoth, Jan-Hendrik (Priv.-Doz. Dr.)
Referee:: Passoth, Jan-Hendrik (Priv.-Doz. Dr.); Ziewitz, Malte (Prof. Dr.); Maasen, Sabine (Prof. Dr.)
Language:: en
Subject group:: SOZ Soziologie
Keywords:: Algorithms, Machine Learning, Social Order, Power, ANT, Software Studies
Translated keywords:: Algorithmen, Machine Learning, Soziale Ordnung, Macht, ANT, Software Studies
TUM classification:: TEC 006
Abstract:: Algorithms and AI systems are becoming increasingly important actors in contemporary societies. Drawing from Science and Technology Studies, media theory, and computer science and based on ethnographic research I describe the development of a machine learning powered recommender system in a public broadcaster. The study shows that the projects for the practical implementation of AI negotiates different social orderings with each other, enabling and limiting algorithmic power at the same time.
Translated abstract:: Algorithmen und KI-Systeme werden zu immer wichtigeren gesellschaftlichen Akteuren. Ausgehend von STS, Medientheorie und Informatik und basierend auf ethnographischer Forschung beschreibe ich die Entwicklung eines auf maschinellem Lernen basierenden Empfehlungssystems in einer ÖR Rundfunkanstalt. Die Studie zeigt, dass Projekte zur praktischen Umsetzung von KI verschiedene soziale Ordnungen miteinander verhandeln, die die Macht von Algorithmen gleichzeitig ermöglichen und begrenzen.
WWW:: https://mediatum.ub.tum.de/?id=1608953
Date of submission:: 19.04.2021
Oral examination:: 28.07.2021
File size:: 2334098 bytes
Pages:: 246
Urn (citeable URL):: https://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:bvb:91-diss-20210728-1608953-1-9
Last change:: 28.02.2022
BibTeX