Webbläsaren som du använder stöds inte av denna webbplats. Alla versioner av Internet Explorer stöds inte längre, av oss eller Microsoft (läs mer här: * https://www.microsoft.com/en-us/microsoft-365/windows/end-of-ie-support).

Var god och använd en modern webbläsare för att ta del av denna webbplats, som t.ex. nyaste versioner av Edge, Chrome, Firefox eller Safari osv.

Porträtt på Isak Engdahl. Foto: Jakob Roséen.

Isak Engdahl

Doktorand

Porträtt på Isak Engdahl. Foto: Jakob Roséen.

Wuthering Weights—Localisation Trajectories of Machine Learning Models for Local Ends

Författare

  • Isak Engdahl

Redaktör

  • Martin Berg
  • Vaike Fors
  • Meike Brodersen

Summary, in English

This chapter analyses trajectories of contextual reconstruction of machine learning models with calibrated weights, sometimes known as pre-trained or foundation models, as they are made operational in specific settings. Through ethnographic vignettes on car detection, bird tracking, and medical technology, it underscores the importance of cooperation and negotiation among actors facing contingencies during trajectories of contextual reconstruction. These stories help us see the behind-the-scenes work of bringing machine learning systems into existence, including interactions between machine learning scientists and domain members. Trajectories are shown to be non-linear, negotiated, and contingent on technical, infrastructural, and regulatory conditions. The process of localising such models results from the collective effort of actors, their negotiations, and resource combinations. Actors use their resources to gather expertise and pool work efforts to address unforeseen events during contextual reconstruction trajectories. The chapters thus considers the wielding of an implicit social license, whereby actors leverage their accreditations to successfully navigate contingencies during work with models that otherwise are open-source and retrievable for many actors.

Avdelning/ar

  • Sociologi

Publiceringsår

2024-09-09

Språk

Engelska

Sidor

321-336

Publikation/Tidskrift/Serie

De Gruyter Handbooks of Digital Transformation

Volym

2

Dokumenttyp

Del av eller Kapitel i bok

Förlag

De Gruyter

Ämne

  • Sociology (excluding Social Work, Social Anthropology, Demography and Criminology)

Nyckelord

  • Science and Technology Studies (STS)
  • machine learning
  • computer vision
  • ethnography
  • foundation models
  • object detection

Aktiv

Published

Projekt

  • Machine Pedagogy
  • Show & Tell: Scientific representation, algorithmically generated visualizations, and evidence across epistemic cultures

ISBN/ISSN/Övrigt

  • ISBN: 9783110792249
  • ISBN: 9783110792256