Investigating Conversational Dynamics in Human-Robot Interaction with fMRI
How do our brains 🧠respond when we converse with a robot versus a human (🤖 vs 👱)?
Today, I read an interesting paper on neuroscience and robotics.
Authors:
Ekaterina Torubarova
Caroline Arvidsson
Julia Uddén
André Pereira
I wrote a summary:
Researchers have been interested in the similarities and differences people experience when having conversations with robots versus humans.
Measuring brain activation alongside behavioral data provides unique insights into the psychological states of individuals interacting with robots that cannot be observed through behavior alone.
The current study utilized a publicly available fMRI dataset of 25 participants. While inside the fMRI scanner, participants had a conversation about an image on a screen with either a human partner or a robot partner (using the Wizard of Oz procedure).
During the periods when participants were speaking, there were no significant brain activation differences between the human and robot partner conditions.
However, during the listening periods, understanding the robot’s speech was associated with greater activation in the bilateral auditory cortex compared to understanding the human’s speech.
Lastly, during the speech preparation phase, preparing speech with a human partner was linked to greater activation in both the bilateral auditory cortex and the visual cortex compared to preparing speech with a robot.
The results suggest that more effort is required to process the robot’s speech, particularly since, in this study, the robot lacked social elements like facial expressions and prosody.
Conversely, the human partner was more expressive, potentially leading participants to recruit more of the auditory and visual cortex when preparing their conversation with a human partner.
💡 I think given that AI-generated speech is increasingly similar to human speech in terms of prosody and expressions, it would be interesting to replicate this study and examine brain activations in response to human versus AI speech.