Umeå University's logo

umu.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Biased large language models for debating robots
Umeå University, Faculty of Science and Technology, Department of Computing Science. (Intelligent Robotics)ORCID iD: 0000-0001-7242-2200
Umeå University, Faculty of Science and Technology, Department of Computing Science. (Intelligent Robotics)ORCID iD: 0000-0001-7242-2200
2024 (English)Conference paper, Oral presentation only (Other academic)
Abstract [en]

The recent development of large language models (LLMs) and improvements in speech recognition have made it realistic to envision AI-driven robots replace humans in debates, or even debating with each other. One application area is politics, where debating robots could support human politicians and, in principle, they could also debate on their own. However, political debating are highly complex and is often guided by values and ideologies, rather than rational decisions based on explicit facts.

In this paper we discuss how introducing appropriate bias into an LLM can be a way to accomplish this. As a simple proof of concept, we present a novel system of three robots conducting verbal debates on selectable topics. The robots are driven by the LLM GPT-3.5, and the desired political view, level of knowledge, and speaking style of each robot are configurable. The results demonstrate how LLMs may be used to argue both for and against different standpoints in debates, and how the output arguments depend on a programmed bias reflecting desired values and ideological principles.

Place, publisher, year, edition, pages
2024.
Keywords [en]
AI, politics, robots, cognition
National Category
Other Computer and Information Science
Research subject
Computer Science
Identifiers
URN: urn:nbn:se:umu:diva-231703OAI: oai:DiVA.org:umu-231703DiVA, id: diva2:1912237
Conference
ICSR'24, 16th International Conference on Social Robotics, Odense, Denmark, October 23-26, 2024
Funder
Swedish Research Council, 2022-04674Available from: 2024-11-11 Created: 2024-11-11 Last updated: 2024-11-12Bibliographically approved

Open Access in DiVA

fulltext(1599 kB)79 downloads
File information
File name FULLTEXT01.pdfFile size 1599 kBChecksum SHA-512
6f893fcc659cd956af22306f371f84586d030e7ff1be1e8aac3256b60ebccd3bfabcafcdb52dbd037b1aa4df7d828b50928c74c26ad096dbd8c1c9a3ee276dca
Type fulltextMimetype application/pdf

Authority records

Bensch, SunaHellström, Thomas

Search in DiVA

By author/editor
Bensch, SunaHellström, Thomas
By organisation
Department of Computing Science
Other Computer and Information Science

Search outside of DiVA

GoogleGoogle Scholar
Total: 79 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 334 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf