summary: With AI-enabled chat tools, people converse more effectively, perceive each other more positively, and use more positive language.
sauce: Cornell University
Researchers at Cornell University have found that people converse more efficiently, use more positive language, and perceive each other more positively when using artificial intelligence-enabled chat tools.
Research published in scientific reportexamined how the use of AI in conversations affects the way people express themselves and see each other.
Malte Jung, Associate Professor of Information Science, said:
“We don’t live and work in isolation. The systems we use affect how we interact with others.”
However, in addition to being more efficient and proactive, the group also perceived its partner as less cooperative when participants thought that the partner used more of the responses suggested by the AI. I discovered that I no longer felt attached to my partner.
Lead author and postdoctoral researcher Jess Hohenstein said: “This shows the persistent overall skepticism people seem to have about AI.”
As a first experiment, the researchers tested the first smart reply platform, “Moshi” (“Hello” in Japanese), modeled after Google’s now-defunct “Allo” (“Hello” in French). We have developed a smart reply platform called , announced in 2016. Smart Reply is generated from LLMs (Large Language Models) to predict plausible next responses in chat-based interactions.
Participants were asked to speak about policy issues and were assigned to one of three conditions: Both participants can use Smart Her Reply. She is the only participant who can use Smart Reply. Or neither participant can use Smart Reply.
Researchers have found that using Smart Reply improves communication efficiency, positive emotional language, and positive evaluations by communication partners. On average, Smart Reply made up 14.3% of his sent messages (1 in 7 hers).
However, participants who suspected their partners responded with Smart Reply were rated more negatively than participants who were thought to have typed their responses themselves. This is consistent with common assumptions about the negative effects of AI.
“AI could help you write,” said Hohenstein. This suggests that you are sacrificing some of your own voice by using text-generating AI. ”
Mr Jung said: This suggests that anyone who can control an algorithm can influence people’s interactions, their language, and their perception of each other. ”
Funding: This work was supported by the National Science Foundation.
About this Artificial Intelligence Research News
author: Becca Bowyer
sauce: Cornell University
contact: Becca Bowyer – Cornell University
image: image is public domain
Original research: open access.
“Artificial Intelligence in Communication Affects Language and Social Relations” Malte Jung et al. scientific report
overview
Artificial Intelligence in Communication Affects Language and Social Relationships
Artificial intelligence (AI) is already widely used in everyday communication, but despite concerns about the negative impact of AI on society, the social impact of using AI to communicate has remained largely unexplored. Not.
We investigate the social impact of one of the most popular AI applications, algorithmic response suggestions (“Smart Reply”). It is used to send billions of messages every day.
Two randomized experiments provide evidence that these types of algorithmic recommendation systems change the way people interact and perceive each other in social and antisocial ways.
We found that using algorithmic responses changed language and social relationships. More specifically, they communicate faster, use positive emotional language, and conversation partners rate each other as more intimate and supportive.
However, consistent with popular assumptions about the adverse effects of AI, people rate it more negatively when they suspect it’s using algorithmic responses.
Thus, while AI can speed up communication and improve interpersonal awareness, AI’s general anti-social connotations undermine these potential benefits when used overtly.
