We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Threshold Designer Adaptation: Improved Adaptation for Designers in Co-creative Systems

Abstract: To best assist human designers with different styles, Machine Learning (ML) systems need to be able to adapt to them. However, there has been relatively little prior work on how and when to best adapt an ML system to a co-designer. In this paper we present threshold designer adaptation: a novel method for adapting a creative ML model to an individual designer. We evaluate our approach with a human subject study using a co-creative rhythm game design tool. We find that designers prefer our proposed method and produce higher quality content in comparison to an existing baseline.
Comments: 6 pages, 2 figures, International Joint Conference on Artificial Intelligence (IJCAI), Special Topic on AI, the Arts, and Creativity
Subjects: Machine Learning (cs.LG); Human-Computer Interaction (cs.HC)
Cite as: arXiv:2205.09269 [cs.LG]
  (or arXiv:2205.09269v1 [cs.LG] for this version)

Submission history

From: Matthew Guzdial [view email]
[v1] Thu, 19 May 2022 01:13:22 GMT (2895kb,D)

Link back to: arXiv, form interface, contact.