This Study Shows How Distinguishable Smiles Are

Having emotions in robotics is advantageous in various ways. Giving a pleasant experience in human-machine interaction will be practicable. For service providers, smiling makes a huge difference in order to create a friendly environment. Incorporating smiles in robots will lead to better performance as virtual agents. This study helps in using smiling to enhance the rating of users to performance of robot or virtual agents.

Borutta et al. (2009) introduced a psychological system-theoretic approach to generate various types of smiles in robots artificially. They also described seven different types of smiles and discussed the hypothesis whether smiles are distinguishable.

A system-theoretic approach was based on the Zurich Model of Social Motivation, which describe the effect of smiling on emotional state of a human. Each of the type of smile among seven types of smile was based on one of the three state dimensions – security, arousal or autonomy.

Security – 1. Trustful Smile 2. Smile of Relief 3. Embarrassed Smile
Arousal – 4. Anxious Smile 5. Surprised Smile
Autonomy – 6. Superior Smile 7. Inferior Smile

Total 20 participants, 10 male and 10 female, took part in experiment in which 126 videos were shown to each participant with 18 videos of each type of smile. The experiment was conducted in two parts. In first part, participants answered this question – “What happens in this situation and how does the person feel?” In second part, participants answered this – “Which kind of smile is shown?” and had to answer among seven types of smile.

The results of first part show that fearful and embarrassed smiles were identified rarely. Inferior smile was mixed with dominance. Best results were obtained for trustful and surprised smiles. The results of second part show that there was a great variation in accuracy of classification by participants. Some of them had accuracy of 19%, while some of them achieved up to 50%.

The overall results show that the classification rate is higher for some types of smiles, especially trustful, surprised and inferior smile. In comparison, superior and embarrassed smiles were classified worst. The frequency response for trustful and anxious smile was similar, so they had been mixed up frequently. 

The perception of smiling as a positive emotion, weaknesses in animation and presenting some situations in fast motion to participants were major challenges before researchers. But, they were successful to generate seven different types of smiles and some of those types of smiles were distinguishable.

This study gave helpful insights about how to design emotions of robots or virtual agent based on ability of people to distinguish smiles.

-----------------------------------------------------------------------------------------------------------------------

Reference - 

Researchers - Isabell Borutta , Stefan Sosnowski , Michael Zehetleitner, Norbert Bischof and Kolja Kuhnlenz

Research paper name - Generating Artificial Smile Variations Based on a Psychological System - Theoretic Approach

Published in: The 18th IEEE International Symposium on Robot and Human Interactive Communication

Toyama, Japan, Sept. 27-Oct. 2, 2009

-----------------------------------------------------------------------------------------------------------------------

Image Courtesy - 

Trustful - Image courtesy of adamr at FreeDigitalPhotos.net

Surprised - Image courtesy of stockimages at FreeDigitalPhotos.net

Embarrassed - Image courtesy of stockimages at FreeDigitalPhotos.net

Anxious - Image courtesy of Gualberto107 at FreeDigitalPhotos.net