The next research project I hope to work on looks at how people make sense of robots. I am particularly interested in how people apply gendered assumptions and attributions to robots or artificial agents, and how apply those assumptions then impact how people engage with robots. In a sense, robots have become another medium as they are another conduit through which we communicate our social and cultural values while we communicate to them and through them to each other. In this regard, understanding how we perceive them helps us to understand how we are already using them as a conduit for the communication of these values.

Growing out of work my partner and I have done on BMO from Adventure Time and Vocaloid Hatsune Miku, I have started talking to my students about this topic to see how they react to robots. I’ve been showing my students images of different robots, asking them to ascribe a gender to the robot, and then following up with what leads them to make such attributions.

For example, this past week I showed 15 images of robots from pop culture and asked my students about how they reacted to the robots as gendered entities. This discussion was part of our day covering gender identity as a social identity. After my students did their presentation defining gender compared to sex, and discussing gender fluidity and the different conceptualizations of gender, I walked them through these images, and had them write down their quick reactions. I gave them only about 30 seconds to write down what gender they thought the robot had and what led them to assign that gender. I told them I wanted what first came to mind when they saw the image, assuring them that there were no right or wrong answers.

Their answers, and our discussion of their answers, has helped me get a better sense on what leads to these gendered assumptions and attributions. Because I am hoping to expand on this idea for my next research project, I thought I would share the results here as well as my thoughts on what it all means. The following are the images I showed my students, in the order that I showed them (starting with Ava from Ex Machina), plus some findings about how they responded to those images.

This slideshow requires JavaScript.

The students unanimously labeled Ava and Rosie as women, and Baymax, Chappie and Data as men. Their reasoning for Ava focused primarily on her body features, which they referred to as “her” breasts, soft faces, curves and little hands; some also suggested that the way she was lying down in the image was feminine. For Rosie, their focus was primarily on her attire, given the stereotypical maid’s outfit she is depicted in; although others said it was due to her perceived attitude given her body posture, or her perceived function given that she was cleaning.

For Baymax, they also focused on the robot’s body, associating it with an older man’s body because they thought “he” had a beer belly. Others ascribed masculinity to the robot given the context of the image; that is, because the robot is surrounded by elements of sports and technology in the image. When it came to Chappie, they reasoned that the image depicted his attitude as being strong, without fear, and ready to fight. They also apparently saw aspects of the robot’s body as supporting that reason, saying “he” was muscular with a broad chest. The argument for Data also focused on the robot’s body, but primarily on his short hair or chiselled face. However, two students did recognize the character and referred to Data’s position in the television show as the basis of their judgment.

This last reasoning is something to explore for a moment. All of these characters come from different time periods in pop culture, and my students did not recognize many of them. Even when they did recognize the character, they did not always reason that the character had a specific gender that aligned with the character’s depiction in that text. For example, with Eve, two people assigned a feminine identity because of seeing the movie Wall-E, but one person said the robot had no specific gender for this same reason. A similar result happened for R2-D2, where some argued for their reasoning based on seeing Star Wars, but one person did not assign a masculine identity for the same reason. This discrepancy of reasoning from the text but not arriving at the same interpretation suggests a need to further consider how much these pop cultural texts impact gendered assumptions and attributions.

Overall, Ava, Maria, Rosie and Eve were all seen primarily as feminine, while the rest, except one, were seen primarily as masculine. The one holdout was Twiki, as the class was split between assigning a masculine or feminine gender identity to the robot. Both sides argued their reasons based on the robot’s body features. The difference apparently came down to the following: if they focused on the robot’s “hair,” they ascribed feminine; but if they focused on the robot’s “broad shoulders,” they ascribed masculine.

Also, across all the reasons given by the students, several key themes recurred. First and foremost was their focus on the robot’s body features, suggesting that they were looking for analogies to human biology to make their determination. This reason suggests the primacy of biological determinism linking sexual characteristics to gender identity.

Other reasons suggest the activation of social and culturally constructed values and stereotypes about gender. Some argued for the robot’s appearance as signaled by its attire, suggesting that certain clothes or appearances are meant for a certain gender (e.g. skirts for women, dirtiness for men). Others looked at the behavior of the robot in the image and inferred its attitude, which was then categorized based on gender stereotypes (e.g. sexually suggestive for women, strong and rugged for men). Another reasoning focused on the aesthetic appearance of the robot, chiefly in its coloring, with certain colors being seen as more masculine than feminine. Others assumed that the robot had a specific function or job, and these functions aligned with a certain gender (e.g. maid for women, army for men).

Finally, some argued that robots, by default, were masculine. For some robots, the students’ reason came down to simply that the entity in the image was a robot. This default suggests the stereotype labeling science and technology as a masculine hobby or professional. This association, then, could lead people to assume that, in the absence of strong visual signifiers for biological determinism or sociocultural stereotypes (based on attire, attitude, aesthetics or function), robots are masculine. This reasoning could also reflect a Western tendency in language to default to masculine, thereby othering the feminine. This reasoning could also be why the majority of robots in pop culture (at least in Western cultures) appear to be assumed as masculine.

This class discussion was very illuminating for me, especially in uncovering these themes in their reasons for gendering robots. And these reasons seem to link what other researchers have found, such as in this article, indicating that I need to look for more such work to see if more overlaps exist. More than that, this discussion has also helped invigorate my interest in moving forward on a formal research study for this project.

My theory is that the main reason to develop gendered robots is same reason for anthropomorphizing them: to ease human-robot interactions. It appears that people may perceive a robot as gendered given the robot’s appearance, performance or function/role as aligning with specific socially and culturally constructed norms (as even the biological determinism argue relies on the sociocultural construction linking biology and gender). Overall, the idea of gendering robots is to make people more comfortable when interacting with robots by drawing on common social and cultural values about gender as a means to structure the interaction.

I believe I could test this theory through an IAT (implicit association test) or a card sorting task (similar to one I did during grad school with a mentor and some peers). Similar to the generation of this class discussion, I could show people images of robots and ask them to categorize or label the robots based on gendered assumptions and attributions. I could then have ask the people to explain their categorization scheme, which would test to see if their explanations contain those three reasons of appearance, performance, and function/role, or to see if other reasons emerge. I could then ask them how their gendered perception impacts how they would engage with the robots of each category. This could also be done with just the audio of robots or AI entities speaking.

I think this could be very interesting, and I would love to collaborate on this project. I just have a book to write first on my fractured fandom project.