Study on virtual babies shows gender stereotypes can influence emotional perceptions

A recent study provides evidence that the name given to a virtual baby can influence how people perceive the baby’s emotions, even if the baby does not display any specific behaviors associated with gender. This suggests that our biases and preconceived notions about gender can affect how we interpret and understand the emotions of others, even in virtual settings where there are no explicit gender-related behaviors.

Computer graphics (CG) have advanced greatly in recent years, allowing for the creation of realistic virtual characters. However, according to the Uncanny Valley theory, the more realistic an artificial human looks, the more discomfort it can cause to the observer. This discomfort arises from the perception of human likeness and identification issues. Gender characteristics, in particular, are influenced by social and cultural constructs about appropriate roles for women and men.

Previous studies have shown that women feel more comfortable with realistic female CG characters, while men feel comfortable with both female and male characters. The researchers wanted to investigate whether these results would still hold true if the virtual humans had no gender identification.

“In my PhD, I work on human perception about virtual humans. So, as virtual humans try to simulate real humans (both visually and behaviorally), I need to take into account characteristics of real humans,” said study author Victor Flávio de Andrade Araujo, a PhD student in computer science at the Pontifical Catholic University of Rio Grande do Sul and member of the VHLab.

“In this regard, I have two scientific reasons for answering this question. The first is that, with regard to virtual humans, evidence in the literature shows us that there is a gender bias both in the perception of those who observe them and those who develop them. When I talk about developing, it is in relation to design, both visual and behavioral (animation). When I talk about who watches them, I’m talking about the audience in general, that is, people who watch movies, series, games, and all media that use virtual humans (virtual characters).”

“The second reason is that, as there is evidence of gender in relation to virtual humans, I (together with my supervisor, Prof. Soraia Musse, and my co-supervisor, Prof. Angelo Costa, who is a specialist and researcher on gender issues) decided to look for scientific studies in psychology that evaluated gender bias, Araujo said. “In this specific case, we replicated a psychology gender study and wanted to see if the gender effects that happened in real life would also happen in the virtual world.”

On a personal level, the researcher also expressed a desire to address issues of diversity in virtual humans. “I prefer to work on this for a sense of fairness, more diversity, more equality, towards a better world,” Araujo explained. “Using gender studies as an example, more diversity would involve more representativeness. For example, the representativeness data of virtual characters in games show that most of the main characters are male, few are female, and almost none of characters are not in this ‘binary spectrum.'”

To conduct their study, the researchers replicated an experiment by Condry and Condry from 1976. In the original experiment, participants watched a video of a real 9-month-old baby sitting in a baby chair facing a mirror. The researchers chose babies as subjects because they are commonly perceived as genderless.

The baby had neutral clothes and no accessories to avoid gender stereotypes. The participants were divided into two groups, with one group receiving the baby with a female name and the other group receiving the baby with a male name. Participants were then asked to assess the baby’s emotional levels using predefined scales.

“We replicated this experiment but translated it into a virtual world,” Araujo told PsyPost. “That is, participants watched videos of a virtual baby reacting to virtual toys.

In the new study, 148 volunteers watched short videos of a 3D model of a baby. In the first scenario, the baby played with a ball, which was meant to represent a positive emotional reaction. In the second scenario, the researchers used a virtual jack-in-the-box, where the baby’s animation and facial expression expressed surprise, aiming to represent a negative reaction. Finally, they included a 3D model of a colored unicorn as an object that was expected to have a neutral effect on the baby.

The participants were instructed to rate the baby’s pleasure, anger, and fear levels for each video, similar to the original study. Three versions of the questionnaire were created, each containing the baby with a female name, a male name, or no name (unnamed baby). Participants were randomly assigned one of the three questionnaires.

In line with the previous study, the results showed that participants’ perception of emotions differed based on the baby’s assigned gender. Women perceived the baby with a female name as more emotional and feeling more pleasure, while men perceived the baby with a male name and the unnamed baby as more emotional and feeling more fear.

“What I was most impressed with was the similarity of the results,” Araujo said. “Both Condry and Condry’s work, and ours, had indications that women rated the baby with a female name as more emotional and men rated the baby with a male name as more emotional (an in-group preference). And, remember, the virtual baby has always been the same, the videos have always been the same.”

The researchers also found that participants who received the unnamed baby mostly perceived it as male, indicating a gender bias in their responses.

“In this case, what was surprising was that most people assigned the male gender to the unnamed virtual baby, even though the virtual baby had no visual attributes that would indicate that the baby was a boy. Our hypothesis is that, generally, in baby girls, people add accessories such as dresses, pink, earrings, etc., while the virtual baby had none of that. So people might have assumed it was a boy for that reason.”

The findings suggested that even without gender-specific behaviors, simply assigning a gendered name to a virtual baby can influence how people perceive emotions. This indicates that our biases from real life can carry over into virtual environments.

The study highlights that “real world gender issues are reflected in the virtual world,” Araujo told PsyPost. “Especially because we have to take human issues into account when working on virtual humans. For developers, designers, researchers, and other professionals, we don’t need to use ‘gender stereotypes’ (social standards, like wearing pink for a female character) when creating female characters. Nor stereotyped behavioral animations, for example, differentiating walking animations between male and female characters.”

“I say this for two reasons: first, we in computing need to understand that we create products with the public in mind. And by creating products with these characteristics, we are creating a culture that does not match reality. Second, trying to eliminate stereotypes, we can debinary the design of virtual characters, which helps with representation issues (we can have female, male, non-binary characters, etc.). One solution, which is what was done in this work, is to present characters through narratives, where we can inform the name, gender, etc, that is, without stereotypes in visual features.”

But the study, like all research, includes some limitations. The study only focused on the perception of emotions in virtual babies and did not explore other aspects of gender bias or stereotypes. The findings may not generalize to other virtual characters or scenarios.

“Thinking about virtual humans in general and not just thinking about virtual babies, we still need to measure these gender issues, especially in relation to our results, when we talk about adult virtual humans,” Araujo said. “This leads me to explain that, according to the literature (Condry and Condry’s own study talks about this), babies are a case that allows us to study genderless effects. So, our results show that they need to be extended to adult virtual humans, as the complexity involving genderless effects is greater. On top of that, we are already conducting these tests with adult virtual humans.”

“I hope that my work leaves a positive message, which inspires people to think about creating a better world,” Araujo added.

The study, “Towards Virtual Humans without Gender Stereotyped Visual Features“, was authored by Victor Araujo, Diogo Schaffer, Angelo C. Brandelli, and Soraia R. Musse.

© PsyPost