First impressions are pivotal. While reading another person’s cues, an abridged version of them forms as we draw on complex social inferences in merely seconds of interaction. That is, if they are human. What if they only resemble a human, but are incapable of inner experience or independent thought? Is it possible to truly form an emotional connection with a robot?
Tinkering inside the human brain
Previous research has shown that the two brain networks that spark to life during the extraordinary feat of understanding intention are the person perception network (PPN) and the mentalizing network (MTN). When interacting with a robot, the PPN springs into action: the occipital, extrastriatal and fusiform body as well as posterior superior temporal sulcus help us scan the face for nanoscale alterations in emotion and the body for subtle changes in movement. When interacting with a human, the MTN is active: the ventral and dorsal medial prefrontal cortex, anterior temporal lobe, temporal-parietal junction and precuneus help us speculate a person’s beliefs, desires, feelings and motives that may explain their visible behavior. When interacting with a robot, however, activity in the MTN is reduced, rendering this network essentially silent.
Sometimes when we interact with a robot that seems “too human” or see an image that is “too realistic,” we experience the uncanny valley phenomenon. An eerie feeling arises, the hair on the back of your neck stands up, and you experience a sense of repulsion and disgust. But what is happening in the brain when we get this feeling? Researchers used a humanoid robot named Nao to find out.
Perceiving a Robot-Human Interaction
Human participants were placed in an MRI scanner, where they completed three tasks:
- A Mentalizing task – reading a short story meant to activate the MTN
- A Person Perception task – gazing at pictures of objects and people meant to activate the PPN
- An Interaction Categorization task – observing interactions between either two humans or a human and a robot, and deciding whether one agent was helping another
This last task with the helping judgment is particularly interesting because it not only ensures that the participant holds an identical processing goal in mind regardless of interaction type, but it also compels the participant to establish a relationship between the two agents.
Why the Eerie Feeling?
At the end of each session, a survey was administered in which the participant rated how eerie vs. believable each observed interaction was and how intelligent and capable of emotion the robot seemed. Participants of this study strongly objectified the robot, rating the interaction less believable than similar ones with human counterparts.
Interestingly, fMRI revealed increases in the activity of brain regions known to implement emotional states, such as the amygdala and insula. It seems that the human’s initial reaction was to probe into the robot’s emotional capability. Reading the intentions behind Nao’s steel gaze activated the MTN. Remember that previous research has shown that MTN activity decreases when a human interacts with a robot; the results of this study suggest that the MTN serves a distinct role in the impression forming process when observing a human interact with a robot. In particular, the activation of the ventral medial prefrontal cortex suggests that humans use more abstract social reasoning to decipher a robot’s intention.
So what does PPN and MTN activation do? Coupled with the high activity in the emotion network, the observer is forced to give a mind to the mindless and emotion to the emotionless. In this way, a first impression with a humanoid robot is made, and communication is established. Nevertheless, it is awkward and stilted, so an uneasy feeling ensues.
Given the rapid and inevitable advances in the field of robotics, where humanoid robots are envisioned to help with domestic, educational and medical support tasks, certain design changes are essential in order to avoid the uncanny valley phenomenon. With future technological and artistic improvements, subconscious conflict could be avoided when the brain’s expectations of human realism are met. Only then will we be able to escape the uncanny valley’s eerie feelings.
Written by Teodora Stoica.
Wang Y & Quadflieg S. (2015) In our own image? Emotional and neural processing differences when observing human-human vs human-robot interactions. Soc Cogn Affect Neurosci. http://doi.org/10.1093/scan/nsv043
Images made by Jooyeun Lee.