HIGH GEAR

[Film] Robot Gendering Nao…and Then?
Artist: Alexandra Bejarano

This video intends to provoke reflection on the gendering practices in social robotics, bringing to question the biases that may be perceived and demonstrated through a robot’s presentation. In this video, a Nao robot’s presentation is altered through changes in appearance and voice, two design cues that are often used in creating distinct identities for social robots. The perception/understanding of these design cues may change depending on the viewer or interactant of a robot and their own biases thus the robot in this video asks, “…then how do YOU perceive me?” Meanwhile, by manipulating these design cues, robot designers may intentionally or unintentionally reflect their own biases and gender robots possibly with stereotypical ideas of femininity and masculinity thus the robot in this video asks, “…then whose and what biases are at play?” From this video, one may conclude that when gendering robots, one must consider multiple perspectives and the implications of those perspectives on robot design and interaction.

The video can be watched here.

[Photography] [0,1]
Artists: Terran Mott, Alex Leto (DU), Autumn McMillan

Our team conducted an open-ended exploration of the Aldebaran gender estimation function. We were interested in a user’s experience interacting with this function. How does it feel to interact with the robot in this way? Can the function be “fooled?” Our goal in this project is to call attention to both the premise that such technological capabilities can or should exist, as well as to the fact that, from a user perspective, Aldebaran’s gender estimation seems inconsistent at best.

We completed our exploration using a NAO robot. Its gender estimation function is part of the ALFaceCharacteristics API. The documentation for this capability includes the sparse explanation that “Gender estimation of a person in the form [gender, confidence]. Gender is either 0 for female or 1 for male and the confidence is in [0, 1].”

Our testing team included three women, one man, and one non-binary person (including this project’s three main contributors). Each tester spent time in front of a NAO robot and were prompted to experiment with making small changes to their appearance, including different facial expressions, putting hair up/down, using cloth as prop hair, removing or putting on glasses, etc. This activity became an unusually personal experience for several participants. Our third contributor reflected that the experience caused them to over analyze themselves and their gender presentation; they found themselves placing value on making faces that would be categorized as a different sex than what they were assigned at birth. Others reflected that it felt unsettling when the robot “guessed very wrong” when a participant meant to look very masculine or feminine, contributing to the users’ perception that the algorithm sometimes seems to just be guessing. Many of us noted that the robot seemed to notice smiles as a clear sign of femininity. This pattern was likely the result of the function’s key-point detection, yet it evoked social norms, stereotypes, and gendered expectations around politeness.

We recorded the video feed from the NAO robot as well as the binary gender and confidence interval that the Aldebaran function outputs at different points. The included series of art pieces, or “reels” are compiled from photos from these sessions. For each tester, the photos are ordered from most confidently labeled “male” by the gender estimation function to the most confidently labeled “female”. In ordering the photos in this way, we invite the viewer to consider how changes in appearance affected the gender predictions. The pieces are envisioned as being most impactful when created for a large, diverse group of participants. They are also designed to be formatted as a full collection complete with context about how the photos and labels were obtained, as well as how the photos in each reel were ordered. 

These reels provide a more intimate look into the user interaction with the function. Intended to represent what a robot “sees”, the photos reflect the sterile feeling of being reduced to a binary gender label. Through these pieces, we aim to provide a springboard for further discussion about the premise and design of such algorithms. There are obvious flaws in attempting to predict binary gender from an image of a person. For those who are frequently asked to defend their gender and bodily autonomy, the belief that artificial intelligence can visually identify sex or denger reinforces harmful binaries that neglect the expansive nature of humanity. To foolishly trust a machine to categorize people based on their facial features reminds us of the transphobic idea that “you can always tell” when someone is trans.

We invite viewers to consider the following questions:

• Is it actually possible to create a gender prediction algorithm that doesn’t rely on or impose harmful ideas in some way?

• Should this type of algorithm exist? In what settings, if any, would it be necessary?

• Imagine that you reordered our reel photos according to what you felt looked most masculine or feminine. How different is your response? How might this exercise make you feel?

[Music] Taiko
Artists: Rena Zhu and Ruojia Sun (CU)

This art hack project explores the norm-critical design of robots in two ways: (1) Designing robots around the theme of slow living, and (2) Applying gender-neutral design in a robot for mindfulness and relaxation. Robots have been associated with utility and efficiency from their conception and persisting to today. This is reflective of societal values of productivity, which leads to stress and burnout in excess amounts. We critique the connection between robots and productivity, instead explore robotics that centers slowness and mindfulness. We are inspired by robot technologies for well-being which leverage touch-based calming interaction, along with simulated breathing rate or heart-rate, to improve mood and relieve anxiety. We hypothesize that these non-anthropomorphic robots, mostly based on soft pet-like designs, are still unintentionally designed with gendered implications. This is demonstrated through difficulties recruiting a diverse gender pool in research studies for all of the aforementioned robots, with one of the studies having 38 women as participants and no other genders. We push back on how robots are currently designed in the space of well-being by proposing a design concept of a novel robotic object for mindfulness and relaxation, Taiko, driven by a gender-neutral design. Inspired by Tibetan singing bowls used for meditation, Taiko is a design concept for an interactive bowl, serving as a decorative piece for the home while not in use, and when tapped, providing calming interaction through touch, sound, and light. Taiko is made of wood, bringing warmth to a space, inviting touch, and producing a pleasant sound. Extending on other technologies to bring awareness to the body for mindfulness and self-regulation, Taiko would listen to the user’s tapping patterns and express itself through music and light, playing alongside the user, then gradually slowing down. We present Taiko in the format of a product flyer. Through long-term use, Taiko would aim to help establish meditative habits and increase mindfulness. We hope that Taiko can be a valuable symbol for users of all genders to embrace slow living and it will inspire robot designers to challenge social norms through intentional robot design and reimagine human-robot relationships.