Artificial Intelligence (AI)
Artificial Intelligence (AI)
Embedded Files
Contextual perception of AI agents in VR
Contextual perception of AI agents in VR
My master’s program also included a research internship. I decided to complete the internship outside of my own university. My main goal was to experience the research environment at another university.
This led to a collaboration with Misha Sra at the University of California, Santa Barbara. The Human-AI Integration Lab, led by Misha Sra, focuses its research particularly on Artificial Intelligence (AI) and Extended Reality (XR).
Given my previous experience with media bias and the development of VR applications, we decided together to conduct a study on the perception of AI agents. During my 3-month internship, I therefore conducted a laboratory study with the support of Misha Sra and other lab staff. The results are subsequently documented in a scientific paper: https://doi.org/10.70401/ec.2025.0013
The study aims to address questions such as:
How is an AI agent that discusses controversial topics perceived?
Does perception vary depending on the method of interaction?
Does the AI agent’s opinion influence perception?
Participants held a five-minute conversation with an AI agent in a VR environment. The topic was current US laws on gun ownership and use. Afterwards, they answered a series of questions about how they perceived the agent.
The study used two between-subjects factors. The first was the input method, with two levels — text input and voice input. The second was the agent's stance: for each participant, the agent argued either congruently or incongruently with the participant's own view, advocating for either looser or stricter gun laws. The agent's gender was assigned randomly to balance out any related effects.
Text input
Text input
Voice input
Voice input
The AI agent was built on the ChatGPT API, with all input processed live during the study using OpenAI's text-to-speech, speech-to-text, and GPT-4o.
Machine Learning
Machine Learning
Machine_Learning_Team_02.mp4The Machine Learning course introduced the fundamentals of the field. Machine learning is a subfield of AI in which a model learns and improves from data and experience rather than being explicitly programmed for its task.
For the project, we implemented a neural network in Python and NumPy for gesture classification. Once trained, the network recognized different user gestures in real time - for example, to control a slideshow with hand gestures. The project video shows a demo.
Page updated
Google Sites
Report abuse