University of Lincoln: To Be a True Human Companion, Social Robots Need Imperfections

How Perfect Is Truly Perfect? A Recent UK Study Reveals That Certain Flaws in Robots Are Key to Fostering Friendly Human-Robot Interaction.

New research suggests that if robots are inherently too perfect, humans are less likely to establish successful working relationships with them.

Studies indicate that an increasing number of “companion” robots are being deployed to assist in the care of older adults and children with autism, Asperger’s syndrome, or emotional disorders. While programming aims to enhance robotic intelligence, it may inadvertently create barriers to the development of long-term human-robot relationships.

The experiment was led by robotics experts at the University of Lincoln in the United Kingdom, who found that people prefer frequent interactions with robots that exhibit behavioral cognitive biases similar to their own—namely, judgmental biases that give rise to individual traits and personalities but also lead to errors and flawed judgments.

The investigation was conducted by Dr. Mriganka Biswas, with necessary assistance provided by Dr. John Murray from the School of Computer Science at the University of Lincoln.

In October 2015, they first presented their research findings at the International Conference on Intelligent Robots and Systems held in Hamburg.

“By introducing certain flaws, such as lapses in judgment, erroneous assumptions, signs of fatigue or boredom, or even exaggerated behaviors, we aim to understand how to make robots’ interactive behaviors resemble those of ordinary people,” said Dr. Mriganka. “By developing these cognitive biases, enabling them to exhibit human-like imperfections, we found that such flaws make their actions more comprehensible to the general public, thereby reducing feelings of alienation and making people more willing to engage with these interactive robots.”

Currently, most human-computer interactions are based on a set of ordered and structured rules and behaviors, but scholars at the University of Lincoln have taken a novel approach by introducing traditionally

Two key factors that play a significant role in human communication—namely, the cognitive biases of “misclassified memories” and the “empathy gap”—were assigned to two different robots to observe differences in their performance.

One of these two robots is Dr. Murray’s ERWAN (Emotional Robot With Affective Network), capable of expressing five basic emotions. The other is Keepon, a beige-colored robot designed to study social development through interaction with children.

Researchers examined multiple interactions between robots and human participants. They found that in more than half of the exchanges, the robots were not influenced by cognitive biases; however, during the remaining interactions, ERWIN made errors in recalling simple facts (involving language use and expression), while Keepon exhibited excessive joy or sorrow (frequently employing various bodily movements and producing noises).

Subsequently, all participants shared their impressions, and the results indicated that the vast majority found it particularly amusing when the robot made mistakes.

“The cognitive biases we introduced made this interaction more akin to casual small talk between ordinary people,” Mriganka explained. “By monitoring participants’ varied responses to the robots, we found overwhelming evidence that people actually prefer robots that make common mistakes, forget things, and occasionally exhibit exaggerated emotions—just like humans do.”

“Our understanding of robots is often influenced by science fiction; however, there is a real conflict between these idealized notions and the reality of distant robotic systems as well as the specialized researchers studying humanoid robots. A companion robot needs to be sufficiently friendly, while also possessing the ability to recognize users’ emotions and needs and respond accordingly. Nevertheless, previous studies have predominantly employed robots lacking human-like characteristics, which can evoke a sense of alienation in users—how are we to interact with companions that appear more perfect than ourselves?”

“As long as a robot can display human-specific flaws in its interactions with humans, we are confident in establishing long-term, positive human-robot relationships.”

The findings of this study provide a reference for Dr. Mriganka’s next phase of research in determining whether to continue using robots endowed with human-specific deficits. Dr. Mriganka’s current research primarily focuses on MARC (Multi-Actuated Robotic Companion), a 3D-printed humanoid robot whose design is supported by the open-source platform projectInMoov.

Previous studies have shown that the appearance of humanoid robots can help users intuitively understand their behavior. Hand gestures, body language, and speech are familiar to our human sensory systems and easy to interpret, as we have been trained from childhood to understand these cues.

Mriganka will continue to investigate whether behaviors characterized by cognitive biases and human flaws elicit more positive responses from users.

Compiled from:http://www.lincoln.ac.uk/news/2015/10/1157.asp