Body and Code: A Distributed Cognition Exploration Into Dance and Computing Learning (2024)

Dance education can support learners’ engagement in body exploration and emotional expression as outlets for learners’ physical and creative energy [26, 30]. Researchers, such as Koff [26] and Vogelstein [44], have advocated for the learning community to recognize dance's potential to build self-knowledge and ground conceptual exploration as part of the embodied cognitive processes of learning [26, 44]. Explorations into STEM and dance education have demonstrated how attention to embodied processes can provide new ways for students to develop their spatial [41], physical [39], mathematical [35], and computational [28] reasoning. Champion's embodied sensemaking analysis [5], for instance, demonstrated how learners were able to “think differently about their bodies as representational tools” as they engaged in translation of concepts such as neurons passing a message into multimodal representations with their bodies, movement, and technology. Computing and dance, more specifically, have grown as technology has expanded the types of computational artifacts learners can construct and the opportunities for conceptual learning across disciplinary boundaries [5, 33, 40]. Researchers have demonstrated its potential to leverage embodied cultural knowledge, instill interest, foster creativity, and evoke a sense of playfulness within computer science education contexts [1, 9, 12, 32]. Fairlie, for example, demonstrated the link between dance activities and the development of computational thinking skills [11]. However, we still do not have sufficient insight into how embodied learning experiences are mediated by technology.

In dance and computing learning experiences, the tools are central to how learners reason across code, their bodies, and their physical environments. Daily etal. [6], for example, observed that elementary school students moved their bodies to think about the commands they needed to code in Alice in order to move a digital avatar in similar or complementary to their own dance movements. Their findings demonstrated that parallels could be made between dance concepts and the code, such as sequencing steps (sequences), repeating steps (looping), and unison (parallelism), which dancers learned as they explored with their bodies. In a slightly different design paradigm for dance and computing technology, the data from dancers’ bodies can be used as a direct input to computational artifacts [8, 33]. These kinds of technologies provide a real-time feedback loop between the user and the computing system, similar to social media applications (e.g., TikTok, SnapChat, etc.) that capture real-time video of users on which effects such as graphics can be superimposed [25]. Instead of learners using code to map between their body and a computational artifact, the learner now has to consider how the computer senses and represents their physical body and movements. The dancer is then invited to think about what it means to develop choreography and code in terms of the digital representations of their body and the physical space. We believe that these two types of dance technology require learners to engage cognitively in different ways, yet we have few studies exploring the cognitive processes shaped across the technology, users, and environment. Our work engages this gap by applying a distributed cognition lens as we analyze how instructors work and learn with one dance computing technology—danceON.

Distributed cognition (DCog) examines a coordinated effort of cognition across individuals, the tools they are using, and the environment. The theory attends to how information propagation and transformation occurs within and across human and non-human agents, centering the role of various externalized representations and how social and cultural contexts mediate interactions within the system [21, 22]. The sociocultural system, “functions by bringing representational media into coordination with one another” [17]. Researchers, such as Hollan etal. [20] and Halverson [17], have explored the importance of this theory for human-computer interaction (HCI) researchers because of the role computing technology plays in processing, representing, and facilitating interaction with information. Hollan etal. [20] highlights how, as a theory for HCI, it orients researchers towards cognition as embodied, socially distributed, and grounded in cultural contexts.

danceON is an open-access, browser-based creative coding environment that enables users to create code while engaging authentically with body motion and dance [4, 33]. Within danceON, users can code virtual animations over videos that are either captured live using a webcam or pre-recorded and then uploaded. Users can create animations bound to coordinates on the screen statically, such as drawing a sun in the corner of the screen, or dynamically, such as drawing birds flying around the screen. Through danceON, users can learn about declaring and manipulating objects, variables, and conditionals; calling on functions; leveraging a pose detection model; and mathematical reasoning in a digital space. Importantly, their bodies are encoded through the system to enable them to integrate body parts in their code and learn about how digital systems approximate the physical world as they use their bodies to create, test, and iterate on their code. danceON uses a pose detection model that consists of body points (e.g., left/right wrist, nose) that are all accessible within the code, enabling users to draw animations following a moving body part, such as drawing lasers shooting from the wrists (Figure 4), and to trigger animations based on conditionals, such as making it rain when both hands are above the shoulders. danceON also enables users to import models trained in Google Teachable Machine [15], providing them access to a probabilistic classifier that can be used to trigger animations. Figure 1 shows the danceON coding environment (Figure 1a) and a sample video still that shows text for a probabilistic classifier from Teachable Machine (Figure 1b). The white dots on the dancer show danceON's skeleton feature, which indicates body keypoints.

In Workshop 1, the facilitators introduced participants to danceON's coordinate (e.g., where:{x:0,y:50}) and body keypoints (e.g., where:{x:pose.rightWrist.x,y:pose.rightWrist.y}) system to enable them to bind and trigger animations based on how they declared objects and conditionals. Figure 1a shows the binding of a purple-colored circle to the dancer's right wrist). The topics covered included illustrative examples, which participants remixed through short exercises as they designed body movements and coded animations. For example, one activity showed participants how to draw wings (i.e., triangles) bound to keypoints on the arms, which they remixed to draw shapes on other parts of their body. Participants were also prompted to think about how they could remix the code from the examples (e.g., Figure 4) to design a movement around their new code.

Workshop 2 focused on connecting body movement with coding concepts with an emphasis on conditionals. Participants began the session by brainstorming concepts they considered important from their disciplines (e.g., conditionals, rhythm, body awareness). Participants were then led through activities to engage them with movement and code on conditionals (Figure 2). They remixed code to experiment with the relationships between the conditional code and when animations would appear. For example, in Figure 2a, the text “I think” appears when the x-coordinate of the nose is less than 200. Participants then brainstormed two different movements; their peers then developed the conditional logic to trigger different visualizations for each.

In Workshop 3, participants explored how to work with distances in danceON. They started with an exercise where they moved their bodies according to given cues. Participants then remixed code that used distances in danceON, such as code that increased or decreased the diameter of a circle bound to their nose depending on the distance between their wrists (Figure 3). Next, participants created movements that could be used to play with distances in code, and taught the movements to their peers. Finally, the participants explored distances and spatial orientation in danceON through a visual analysis of John Baldessari's Umbrella (Orange): With Figure and Ball (Blue, Green) [2]. They then recreated the image by drawing shapes with danceON and physically enacting parts of the image1. Participants discussed how they used visual analysis to think about drawing animations and simulating depth in danceON.

We recruited STEM and dance instructors from various avenues to participate in our workshops: a community organization partner, a dance education program, and an interactive media program. We recruited from our community partner, STEM From Dance (SFD), a community organization dedicated to empowering girls of color through programs that combine dance with tech. The dance education program prepares graduate students for careers as dance educators and the interactive media program trains graduate students in applying interactive media for creative expression and critical engagement; both programs are at New York University (NYU), a research university in the northeastern United States. In total, we recruited six educators who self-selected into the workshops that they wanted to participate in based on their time availability. The participants also filled out a survey before the workshops that collected information about their experience with dance and coding. Table 2 describes the participants in our study, including the workshops that they participated in. All participants consented to participate in the workshops through an IRB-approved consent form and were compensated $55 USD per hour (in the form of gift cards) for their participation.

We video and audio recorded all workshops through the video conferencing platform, Zoom [47]. We used a laptop camera and a separate mounted camera to video capture the sessions and a microphone (either through the built-in laptop mic or a separate mic connected to a laptop) for audio capturing. The audio from the recordings were transcribed for analysis. We collected the participants’ outputs from the workshop activities such as their code and writings/sketches. In some activities, their outputs were dances or body movements; in those cases, we took video stills or clips from the video recordings for analysis.

We created cases of each of the individuals or groups working through each of the activities to explore how information was interacted with, shared, and transformed. We open-coded [42] the participants’ transcripts and video recordings, guided by the research questions described in Section 1. In our analyses, we attended to the concepts (e.g., body shape, visual analysis) that the instructors focused on; their descriptions of relationships and tensions between their disciplinary knowledge, practices, and computing concepts; and how they demonstrated connections between dance and computing. To understand how participants engaged with the computational representations of their bodies and the physical space with danceON, we coded for instances that pointed to participants negotiating between the digital and physical representations of bodies, space, and movement and how they engaged in translations across physical and digital representations. The visual and auditory parts of the data provided us with speech, gaze, and body positioning but we could not have certainty on where instructors were directing their attention or what they were thinking. The researchers, therefore, went through iterative rounds of watching videos and discussing interpretations to code the cases collectively, refining our definitions of them and coming to consensus on any disagreements. The cases were built from triangulated [31] data across our sources to ground our coding and interpretations in the participants’ activities and artifacts. We identified four themes: (1) the instructors’ use of their bodies as a shared point of negotiation, (2) the participatory role of danceON within the sensemaking process, (3) the instructors’ use of representations to translate across physical and digital spaces, and (4) the instructors’ collaborative sensemaking. We present four cases which variably illustrate examples of these themes in the context of the instructors’ work.

We identified multiple instances of the instructors using the coded animations bound to their bodies as a way to explore the machine learning pose models that danceON uses. In Workshop 1 (Section 3.2), Linda[S] explored an example called “laser arms” (Figure 4); the DCog system consisted of two participants Linda[S] and Mary[S], two researchers Shuang and Willie, and the computer running danceON in front of Linda[S] and Shuang. danceON is reading Linda[S]’s bodily states and outputs the animation of lasers shooting from her wrists through the video panel (Figure 4b shows an example of the animation). Linda[S] is looking at this animation and moving her body as she builds her mental model of how it works. She explores the breakdowns between what she believes should be happening and what she sees as she moves her body.

[1:34] Linda[S]: “I'm shooting laser arms” [moves hands together (Figure 5a) and apart (Figure 5b)][1:41] Linda[S]: “Occasionally it like...” [moves her hands up (Figure 5c) and down (Figure 5d)] [1:44] Linda[S]: “My, my hands don't have to necessarily be visible for it to still shoot I guess... um, which I guess, um, probably the camera can see more than it is showing me, which is why. Or that's my guess.”

As Linda[S] makes her first hypothesis aloud to the others in the room, Mary[S] and Willie (one of the researchers) begin looking at Linda[S]’s movement to observe the behavior that Linda[S] is describing.

[1:55] Mary[S]: “It's tracking you... [looking at Linda[S]’s screen] It looks like when your fingers pop up it...” [1:57] Linda[S]: “Yeah cause right now, I'm shooting lasers and there's definitely no hands.”

Looking at the screen, Mary[S] and Linda[S] begin to use danceON's skeleton feature (the body keypoint indicators) (Figure 6a) as an additional representation to make sense of the model.

[2:01] Mary[S]: “What is the like bottom marks? [pointing at Linda[S]’s screen with the skeleton (Figure 6a)]” [2:05] Linda[S]: “Like the skeleton? There. This is the shoulder [moves the cursor over the body keypoints]. I'm trying to keep my hands out [of the camera screen].” [2:19] Linda[S]: (Responding to Mary[S]) “It is right shoulder.” [lifting her shoulder up] [2:25] Linda[S]: [Moves her whole body up in the seat to raise both shoulders] “Ehhhhh. I still don't have hands but I'm shootin’ lasers. Maybe it's guessing. [looking at Mary[S] ] It's just like ‘we assume you're not an amputee.’ ”

Linda[S] makes a second hypothesis about the underlying model and its biases around the type of body it is expecting.

[2:32] Shuang: “That is weird” [looking at her own screen and then turning on the skeleton feature on the projected danceON video (Figure 6b)] [2:39] Linda[S]: “The dots, the last dots it gets are my shoulder. It'll still” [moving hands from keyboard and the laser animation stops] [2:43] Shuang: “It just doesn't [she gets up and walks over to see Linda[S]’s screen]... let me see”

Shuang is now sitting next to Linda[S] to watch the behavior. The researchers make a hypothesis as Linda[S] continues to explore with Shuang.

[2:46] Linda[S]: “I kinda want to figure out what's going on.” [sits up straight and the lasers start and stop]. [2:49] Linda[S]: [Lasers start up again while her hands are out of the video canvas] “See there are lasers.” [2:53] Shuang: “Interesting” [Linda[S] moves her hands down and the lasers disappear completely] [2:54] Linda[S]: “But...” [she moves hands up and the lasers come back in full] [2:56] Shuang: [Looking at the screen] “I know why this is.” [2:59] Willie: “What's happening?” [3:00] Shuang: “It's the bottom the grey bar it's actually part of the...” [talking about the video control bar that is overlaid on part of the video canvas]. [3:03] Willie: “It's still part of the space.” [3:05] Shuang: “Yeah” [3:06] Linda[S]: [Shuang moves the laptop screen back to lift up the camera angle. Linda[S] moves her hands below the camera view (Figure 6c)]. Okay but now my hands are definitely not there. Lasers.” [3:12] Shuang: “I can't figure out why.”

Linda[S] and Shuang disprove the hypothesis about the hidden part of the canvas being the problem by positioning her hands further outside of the camera view and still getting the animation to trigger. Willie tries to help her troubleshoot.

[3:31] Willie: “What happens if your entire body is out of the frame? Does anything appear?” [3:41] Linda[S]: [She moves her body out of the screen and then puts just her hand in the view (Figure 6d)] “Nothing” [moves her hand around] [3:42] Mary[S]: [Looking at Linda[S]’s screen and the loss of the skeleton dots and animation] “It's not even tracking you.” [3:43] Linda[S]: “No. [moves hand around some more and the body keypoints phase in and out with the laser animation] Oh. There we go.”

In this case, the laser animations, the skeleton, and body positions of Linda[S] mediated the collaborative exploration of the pose model. The laser animation served two roles in the cognitive processing: (1) providing a representation that motivated exploration with the body and (2) focusing attention to particular points in the model (i.e., the wrists and arms) as a way to understand the model's behavior. The skeleton was then used to drill into the unexpected behavior. Linda[S]’s body positions were able to be seen, referenced, and discussed as Linda[S] tried certain things and others suggested certain movements. Linda[S] moved around the camera's view using her physical distance from the camera as she was working with the limits of what danceON “sees” and represents. The interaction between Linda[S]’s body positions (i.e., the physical space) and the animation behavior (i.e., the digital space) supported Linda[S], Mary[S], and Shuang as they used their coding knowledge to collaboratively create and test technically-informed hypotheses of what was being “seen” and why the model was providing particular predictions. They worked across the various representations from Shuang and Linda[S]’s interactions with danceON.

In Workshop 2 (Section 3.3), the instructors remixed code examples to design conditionals and corresponding movements to trigger animations. For this exercise, Linda[S] and Aria[D] aimed to draw text above their head when their hand is close to their chin (see their final creations in Figure 7). The DCog system formed between Linda[S] and danceON involved interactions in which Linda[S] repeatedly: moved away from the computer to get into danceON's camera view → held up her right fist against her chin (Figure 8a) → observed the animation changes on the screen → then came back to the computer to continue to code (Figure 8b). She repeated this seven times using her body to test and debug her code. Similar to Case 1, the animation (i.e., the text) became a way for Linda[S] to determine how the computer was recognizing her body. When asked about their process, Linda[S] explains:

“We were just playing with, like, the hands in relation to the rest of the self, especially how it affects our gestures [...] and then finding where, like, there was no chin point, so I think it's like the shoulder with some math to make it a little bit [above the shoulder] and then talking ‘bout how to play with the text [...] it's like a range between the nose and the shoulder so that it wouldn't get the head, but if you go above that, it [triggers the “she thinks too much” text]. ”

The process of constructing the code drew their attention to their gestures and body positioning so they could detect it with the code. They then had to use their bodies in conjunction with the pose model outputs to decide what body parts they could use as they navigated the system's limited set of body keypoints (i.e., no chin point). They then mathematically approximated, tested, and iterated to get the location of the chin and the distance of the fist to the chin.

Later, as Linda[S] and Aria[D] continue to work, Linda[S] drinks from a cup with her right hand (Figures 9a and 9b). Aria[D] observes that this triggers the animation they coded earlier (i.e., the right-wrist-to-chin movement, Figure 9b). This serendipitous moment, captured and represented by danceON, drew them back into the code. The interaction resonated with their initial idea of conveying a meme effect by using captions, expressions, and gestures in their activity assignment, so they incorporated it into their output (Figures 7 and 9c).

In this case, the consistently available representations from danceON running the code made it so the DCog system was processing information across the agents even when Linda[S] and Aria[D] were not actively coding at the time. The code provided another layer of insight into the physical environment and embodied interactions of the instructors through its digital representations. Linda[S]’s unplanned body movements impacted the animation, which they fed back into their creative process. This moment was not merely about triggering an animation; it highlighted how incidental actions within the physical environment could reveal and inform the digital interactions designed by the instructors. By providing continuous feedback on the programmed movements, danceON turned a routine action into a creative impetus illustrating danceON's participatory role in the cognitive processes.

As we explored conditionals in Workshop 2 (Section 3.3), Morgan[D], one of the dance instructors, proposed using danceON to detect different shapes of movement (a classification of movements in dance [27]). She envisioned creating a short clip that used conditionals to animate a response when the dancer moves into various shapes. Morgan[D] initially proposed doing a series of round shapes each corresponding to its own conditional case (Figure 10). As she worked, the DCog system consisted of Morgan[D] moving into round-shapes in front of danceON while physically and verbally explaining to Shuang (one of the researchers) the positions of the body parts that compose each movement (Figure 10):

Morgan[D]: [Pointing to her left wrist with her right hand (Figure 10a)] “... and this one, this is going to be this.” [continues to further bend down (Figure 10c)] “Or we could do it to knees.”

Morgan[D] pointed to the different body parts that correspond to danceON's body keypoints (e.g., wrist, nose, and knees) as she created the “shapes” with her body. She used her bodily representations in the physical space to reason through what can be sensed based on the skeleton representation in the digital space. In doing so, she came up with the conditional logic to differentiate the shapes and communicate these to Shuang who helped her construct the code.

During this process, Morgan[D] stood sideways relative to the screen demonstrating her recognition of the camera's limits around perceiving her body's physical dimensions—i.e., the “flattening” of her body as the danceON video reduces the 3D world into a 2D digital representation. In the DCog system, the flattening of her rounded shapes would be represented by both the video where she would see herself being projected in a straight line if facing the camera head on (versus rounded if sideways); and by the body keypoints on the screen, which would cluster together if facing the camera head on (versus spread out if sideways). Similar to a dance performance, where the dancers and choreographers need to be cognizant of the direction the body is facing, Morgan[D] considered how danceON was viewing her movement to strategically orient her body for better recognition. Despite not having prior coding experience, Morgan[D] was able to use her knowledge as a dancer surrounding body positioning, body orientation, and shapes of movement to navigate and translate between the physical and digital representations. Using the feedback from the body keypoints, she was able to communicate verbally and physically how the conditionals could be set up in code to sense round shapes.

In Workshop 3 (Section 3.4), participants conducted a visual analysis and recreation of an artwork using danceON. The artwork analyzed was Baldessari's Umbrella (Orange): With Figure and Ball (Blue, Green) [2]—the artwork shows a graphical orange umbrella in the background with a person balancing on a graphical green circle in the foreground with their face blocked by a blue circle. When the instructors were prompted to discuss their initial observations, they explain:

Morgan[D]: “It could be hard balancing on the ball or jumping... Maybe it's either fear or excitement.” Shuang: “What tells you about the sentiments?” Morgan[D]: “Well, the arms in the air, that's something... I'm also going off the background...it almost looks like a thunderstorm and fear. Or maybe it's like a beautiful day and she's happy.” Linda[S]: “She has no face... I hope she's not [being hit] by a ball?!” Amy[S]: “I think one thing is we like to assume that she's standing on the ball, but potentially it could be affixed to the foot and then the person could just be jumping and the ball is following.”

This initial discussion sets the stage for a DCog system where each participant's observations and interpretations contribute to a collective understanding of the artwork. Morgan[D] provided descriptions of embodiment through reasoning around the motion of the bodies and their potential emotions. Both Linda[S] and Amy[S] explored the character's perspective: Linda[S] created a humorous scenario where the character was hit by a ball, hence waving their arms in the air and Amy[S] refactors the relationship between the ball and the character by imagining a physical connection between the ball and foot tying it together with a narrative. Having to think through the “sentiments” and how the emotions were being expressed within the image expanded the creative possibilities around their interpretations of the bodily states of the characters. These explorations played a crucial role as they moved on to recreate the piece with their bodies in danceON.

Two of the instructors, Linda[S] (in light orange) and Amy[S] (in black), collaborated to recreate the artwork (Figure 11). Linda[S] positioned herself close to the computer to observe the composition (Figure 11a) in danceON while trying to avoid being captured by the camera (Figure 11b). Amy[S] stood behind Linda[S] while lifting her left leg as the performer (Figure 11a). In the video, it looks as if she is “balancing” herself on top of Linda[S] to complete the recreation.

Amy[S]: [Posing and speaking to Linda[S] ] “Your body is also kind of a semicircle... You got promoted into an umbrella” (Figure 11b)

The process of choreographing this image prompted the instructors to negotiate their use of the physical space with its digital representation as captured by danceON. The cognitive process was not confined to individuals but extended through how the instructors were moving their bodies in relation to one another, the environment, the camera view, and how they interacted with the static coded objects. While Linda[S] positioned herself close to the camera to “become” the umbrella, Amy[S] still had to consider her physical distance both from the camera and from Linda[S] to mimic the original composition. Amy[S] suggested adjusting the digital elements to better fit the physical space: “and then we can make it displaced and like kind of huge [...] scoot it over and make it bigger.” The instructors collaboratively mapped physical locations and their bodies to danceON's “flattened” representation of space.

Throughout the activity, there was an ongoing re-interpretation of digital representations of body, graphical elements, and physical spaces across participants to achieve a collective goal. Participants’ interpretations, physical interactions with danceON, and adaptations to its constraints all shaped the cognitive processes.

A key takeaway from our work is the value of having the body as “part” of the code and the computational artifacts produced within the dance computing learning environment. Co-creating dance performances with danceON necessitated a complex interplay of resources from the various actors participating within the distributed cognition system. The STEM and dance instructors each brought a range of disciplinary and individual expertise, skills, and experiences to creating and testing movements with their bodies, exchanging ideas through discussion, and writing code to draw interactive visualizations. Within these interactions, the participants’ bodily states were a shared point of negotiation as the DCog system coordinated and transformed information throughout their activities. From a social constructivist perspective [45], the physical and the digital representations facilitated the social interaction central to learning in ways that enabled participants to draw on their cultural schema for more than just communicating semantically; they conveyed ideas to the other participants and the computing devices with their movements, gestures, and postures. Work by Amber Solomon has begun to bring attention to the need for computing education to consider gestures and embodied language within learning design. She and her team documented how instructors use gestures to support student learning and how students use gestures to communicate their knowledge [37, 38]. Our work extends this call, advocating for educational computing technology that integrates gestures and embodied ways of knowing as part of the interaction design and code learners create.

Further, such a design requires learners to be present and bring themselves into the learning process as part of their computational artifacts. Learners communicated with and brought their knowledge into how they moved and talked about their bodies in relation to the computing system and their computational designs. The literature has shown a range of ways in which the materiality of computational objects can support learners to engage with their identities and cultures to support meaningful experiences [23, 24, 34]. In our work, the body became part of the computational material demonstrating the various ways the instructors’ dance cultures, knowledge, and ways of being were present within the experience. This invites critical investigation into the broader challenges of designing learning environments that champion this kind of presence. For example, Mathayas etal. [29] questioned assumptions that embodied forms of participation are always good and desired by participants, calling scholars to explore how we can create embodied learning spaces that also invite meaningful and dignity-affirming participation by learners.

danceON served a participatory role during the instructors’ creation and learning process because of the various representations the system could show the participants and the opportunities it provides to the instructors to observe, manipulate, and work with those representations—e.g., from the skeleton that the participants could toggle on and off, to the diversity of coded visual animations that the participants could remix and create. Instructors navigated these representations as they made intentional choices about when to attend to the different representations, create different representations, and change their movements in response. Paulo Blikstein's [3] idea of selective exposure or the intentional foregrounding (or backgrounding) of specific aspects of a technology to manage cognitive load, can help designers focus learners’ attention to specific concepts or skills. We saw how danceON at times focused the participants’ attention through visual outputs that simplified what danceON saw—such as the laser animation that highlights the wrists without necessitating attention to the entire skeleton model—while also engaging the participants to move their bodies and playfully explore—such as the shapes of movement and “she thinks” activities (Figures 2, 7, and 10). The ways the system exposed the participants to parts of the machine learning model enabled them to want to explore the “seams” of the technology—i.e., when it was and was not triggering—impacting their processes. What is unique in this case, is that the selective exposure does not just have power because it can simplify the cognitive load but because the artistic output creates a motivational affordance based on how the model is being exposed. Importantly, in a collaborative environment, we saw how participants collectively discussed the danceON representations (i.e., what was exposed) in relation to the computational concepts they engaged with. Our work asks us to consider how computational art can be viewed as part of the participatory role of the technology. Specifically, how it can push the boundaries of designing how we selectively expose particular functionality and how we choose various representational forms to support embodied, playful, and collaborative sensemaking of computational systems and artifacts.

Within dance and computing spaces, the exploration and sensemaking extends beyond the physicality of movement in physical space, into the ways that space and movements can be captured, extended, and transformed through digital technologies. The instructors and technology were constantly translating bidirectionally within and across the physical and digital spaces. Instructors coordinated movements and interactions in the physical environment to connect to how the computational system (including the camera, the pose model, and the user's code) saw and presented the space digitally. This iterative translation involved a reinterpretation of what physical movements signify within danceON's digital space, effectively blurring the boundaries between the digital and physical. Even ordinary activities, like taking a sip of water, became opportunities for creativity as they were translated by danceON.

Participants had to use their representational competence to understand how their information was getting mapped between their 3D world and the 2D computing system. Additionally, as Halverson would argue, they were engaging in a representational process as part of the art-making process as they made decisions about what they wanted to represent and how [18]. This work was also collaborative necessitating that participants had to negotiate the physical and digital representations and their meanings. The extensive engagement across these representations suggests it could be beneficial to explore how metarepresentational competence could support learners in these embodied computing environments as they learn to translate between, critique, and create mathematical, scientific, and technological representations [10].