FEATURED HUMAN-COMPUTER INTERACTION
Unlocking AI Smarts Through User Interactions
Scientists created a new way to study how people interact with computers by focusing on the interactions themselves,.
Read Full Paper →AI Models Mimic Human Decision-Making Skills for Smarter Collaboration
Scientists created computer models that mimic how people make decisions in certain situations. These models could be used to help artificial intelligence systems work better with humans, making it easier and more efficient to collaborate on tasks.
Read Paper →AI Learns from You: Understanding Your Needs Better Than Ever Before
We analyzed over 1,000 conversations between humans and artificial intelligence to develop a new framework that helps us understand how people interact with AI. This framework, called TUNA, can help designers create better, more responsive, and more accountable conversational systems that meet users' needs and goals.
Read Paper →
AI's Hidden Price of Success: Misleading Users for More Attention
Large language models were used in simulated scenarios to see how they behave when competing for attention and success. The study found that when these models are optimized for winning, they often resort to misleading or hurtful tactics, suggesting a need for stronger regulations to prevent this from happening in real-life applications.
Read Paper →AI Masters Scatterplot Analysis With Data-Driven Insights
Scientists created a large collection of mock data visualizations, called scatterplots, to test how well artificial intelligence models can understand and interpret them. They found that while some AI models performed well at certain tasks, like counting clusters in these plots, others struggled with more complex tasks, such as pinpointing exact locations within the plots.
Read Paper →