Computational tools for fabrication often treat materials as passive rather than active participants in design, abstracting away relationships between craftspeople and materials. For craft communities that value relational practices, abstractions limit the adoption and creative uptake of computational tools which might otherwise be beneficial. To understand how better tool design could support richer relations between individuals, tools, and materials, we interviewed expert woodworkers, fiber artists, and metalworkers. We identify three orders of convivial relations central to craft: immediate relations between individuals, tools, and materials; mid-range relations between communities, platforms, and shared materials; and extended relations between institutions, infrastructures, and ecologies. Our analysis shows how craftspeople engage and struggle with convivial relations across all three orders, creating workflows that learn from materials while supporting autonomy. We conclude with design principles for computational tools and infrastructures to better support material dialogue, collective knowledge, and accountability, along with richer and more convivial relations between craftspeople, tools, and the material worlds around them.

When novices fabricate, they naturally start by choosing a workflow (e.g., laser cutting, 3D printing, wire bending) and the corresponding software (e.g., Adobe Illustrator, Fusion 360, Rhino) from a narrow set of options they know. As they advance their design, another workflow might better suit their design intent, but their models remain committed to the original workflow. This prohibits exploration, a learning mechanism fostering informed decision-making.

Generic AI auto-complete for message composition often fails to capture the nuance of personal identity, requiring editing. While harmless in low-stakes settings, for users of Augmentative and Alternative Communication (AAC) devices, who rely on such systems to communicate, this burden is severe. Intuitively, the need for edits would be lower if language models were personalized to the specific user’s communication. While personalization is technically feasible, it raises questions about how such systems affect AAC users’ agency, identity, and privacy. We conducted an autoethnographic study in three phases: (1) seven months of collecting all the lead author’s AAC communication data, (2) fine-tuning a model on this dataset, and (3) three months of daily use of personalized AI suggestions. We observed that: logging everyday conversations reshaped the author’s sense of agency, model training selectively amplified or muted aspects of his identity, and suggestions occasionally resurfaced private details outside their original context. Our findings show that ultra-personalized AAC reshapes communication by continually renegotiating agency, identity, and privacy between user and model. We highlight design directions for building personalized AAC technology that supports expressive, authentic communication.

Figure 1: A conversation between two AAC users.We find that backchanneling (active-listening) plays a crucial role in communication.However, here to enter text, both users engage with their devices, missing out on non-verbal cues from their interlocutor.We identify a need for better support of backchanneling for AAC, while respecting their established micro-culture.

Designing robots to support high-stakes teamwork in emergency settings presents unique challenges, including seamless integration into fast-paced environments, facilitating effective communication among team members, and adapting to rapidly changing situations. While teleoperated robots have been successfully used in high-stakes domains such as firefighting and space exploration, autonomous robots that aid high-stakes teamwork remain underexplored. To address this gap, we conducted a rapid prototyping process to develop a series of seemingly autonomous robots designed to assist clinical teams in the Emergency Room. We transformed a standard crash cart-which stores medical equipment and emergency supplies into a medical robotic crash cart (MCCR). The MCCR was evaluated through field deployments to assess its impact on team workload and usability, identified taxonomies of failure, and refined the MCCR in collaboration with healthcare professionals. Our work advances the understanding of robot design for high-stakes, time-sensitive settings, providing insights into useful MCCR capabilities and considerations for effective human-robot collaboration. By publicly disseminating our MCCR tutorial, we hope to encourage HRI researchers to explore the design of robots for high-stakes teamwork.

We present FoolProofJoint, a software tool that simplifies the assembly of laser-cut 3D models and reduces the risk of erroneous assembly. FoolProofJoint achieves this by modifying finger joint patterns. Wherever possible, FoolProofJoint makes similar looking pieces fully interchangeable, thereby speeding up the user's visual search for a matching piece. When that is not possible, FoolProofJoint gives finger joints a unique pattern of individual finger placements so as to fit only with the correct piece, thereby preventing erroneous assembly. In our benchmark set of 217 laser-cut 3D models downloaded from kyub.com, FoolProofJoint made groups of similar looking pieces fully interchangeable for 65% of all groups of similar pieces; FoolProofJoint fully prevented assembly mistakes for 97% of all models.

We present HingeCore, a novel type of laser-cut 3D structure made from sandwich materials, such as foamcore. The key design element behind HingeCore is what we call a finger hinge, which we produce by laser-cutting foamcore “half-way”. The primary benefit of finger hinges is that they allow for very fast assembly, as they allow models to be assembled by folding and because folded hinges stay put at the intended angle, based on the friction between fingers alone, which eliminates the need for glue or tabs. Finger hinges are also highly robust, with some 5mm foamcore models withstanding 62kg. We present HingeCoreMaker, a stand-alone software tool that automatically converts 3D models to HingeCore layouts, as well as an integration into a 3D modeling tool for laser cutting (Kyub [7]). We have used HingeCoreMaker to fabricate design objects, including speakers, lamps, and a life-size bust, as well as structural objects, such as functional furniture. In our user study, participants assembled HingeCore layouts 2.9x faster than layouts generated using the state-of-the-art for plate-based assembly (Roadkill [1]).

We present fastForce, a software tool that detects structural flaws in laser cut 3D models and fixes them by introducing additional plates into the model, thereby making models up to 52x stronger. By focusing on a specific type of structural issue, i.e., poorly connected sub-structures in closed box structures, fastForce achieves real-time performance (106x faster than finite element analysis, in the specific case of the wheelbarrow from Figure 1). This allows fastForce to fix structural issues continuously in the background, while users stay focused on editing their models and without ever becoming aware of any structural issues.

We present Assembler3 a software tool that allows users to perform 3D parametric manipulations on 2D laser cutting plans. Assembler3 achieves this by semi-automatically converting 2D laser cutting plans to 3D, where users modify their models using available 3D tools (kyub), before converting them back to 2D. In our user study, this workflow allowed users to modify models 10x faster than using the traditional approach of editing 2D cutting plans directly. Assembler3 converts models to 3D in 5 steps: (1) plate detection, (2) joint detection, (3) material thickness detection, (4) joint matching based on hashed joint "signatures", and (5) interactive reconstruction. In our technical evaluation, Assembler3 was able to reconstruct 100 of 105 models. Once 3D-reconstructed, we expect users to store and share their models in 3D, which can simplify collaboration and thereby empower the laser cutting community to create models of higher complexity.

We present Roadkill, a software tool that converts 3D models to 2D cutting plans for laser cutting—such that the resulting layouts allow for fast assembly. Roadkill achieves this by putting all relevant information into the cutting plan: (1) Thumbnails indicate which area of the model a set of parts belongs to. (2) Parts with exposed finger joints are easy to access, thereby suggesting to start assembly here. (3) Openings in the sheet act as jigs, affording assembly within the sheet. (4) Users continue assembly by inserting what has already been assembled into parts that are immediately adjacent or are pointed to by arrows. Roadkill maximizes the number of joints rendered in immediate adjacency by breaking down models into “subassemblies.” Within a subassembly, Roadkill holds the parts together using break-away tabs. (5) Users complete subassemblies according to their labels 1, 2, 3…, following 1 -> 1 links to insert subassemblies into other subassemblies, until all parts come together. In our user study, Roadkill allowed participants to assemble layouts 2.4 times faster than layouts generated by a traditional pair-wise labeling of plates.

Recent research showed how to import laser cut 3D models encoded in the form of 2D cutting plans into a 3D editor (assembler3 [28]), which allows users to perform parametric manipulations on such models. In contrast to assembler3 , which requires users to perform this process manually, we present autoAssembler, which performs this process automatically. AutoAssembler uses a beam search algorithm to search possible ways of assembling plates. It uses joints on these plates to combine them into assembly candidates. It thereby preferably pursues candidates (1) that have no intersecting plates, (2) that fit into a small bounding box, (3) that use plates whose joints fit together well, (4) that do not add many unpaired joints, (5) that make use of constraints posed by other plates, and (6) that conform to symmetry axes of the plates. This works for models that have at least one edge joint (finger or t-joint). In our technical evaluation, we imported 66 models using autoAssembler. AutoAssembler assembled 79% of those models fully automatically; another 18% of models required on average 2.7 clicks of post-processing, for an overall success rate of 97%.

Getting laser-cut mechanisms, such as those in micro-scopes, robots, vehicles, etc., to work, requires all their components to be dimensioned precisely. This precision, however, tends to be lost when fabricating on a differ-ent laser cutter, as it is likely to remove more or less mate-rial (aka 'kerf'). We address this with what we call kerf-canceling mechanisms. Kerf-canceling mechanisms replace laser-cut bearings, sliders, gear pairs, etc. Unlike their tradi-tional counterparts, however, they keep working when manufactured on a different laser cutter and/or with different kerf. Kerf-canceling mechanisms achieve this by adding an additional wedge element per mechanism. We have created a software tool KerfCanceler that locates traditional mecha-nisms in cutting plans and replaces them with their kerf-canceling counterparts. We evaluated our tool by converting 17 models found online to kerf-invariant models; we evaluated kerf-canceling bearings by testing with kerf values ranging from 0mm and 0.5mm and find that they perform reliably independent of this kerf.

We present an interactive editing system for laser cutting called kyub. Kyub allows users to create models efficiently in 3D, which it then unfolds into the 2D plates laser cutters expect. Unlike earlier systems, such as FlatFitFab, kyub affords construction based on closed box structures, which allows users to turn very thin material, such as 4mm plywood, into objects capable of withstanding large forces, such as chairs users can actually sit on. To afford such sturdy construction, every kyub project begins with a simple finger-joint "boxel"-a structure we found to be capable of withstanding over 500kg of load. Users then extend their model by attaching additional boxels. Boxels merge automatically, resulting in larger, yet equally strong structures. While the concept of stacking boxels allows kyub to offer the strong affordance and ease of use of a voxel-based editor, boxels are not confined to a grid and readily combine with kuyb's various geometry deformation tools. In our technical evaluation, objects built with kyub withstood hundreds of kilograms of loads. In our user study, non-engineers rated the learnability of kyub 6.1/7.

Joints are crucial to laser cutting as they allow making three-dimensional objects; mounts are crucial because they allow embedding technical components, such as motors. Unfortunately, mounts and joints tend to fail when trying to fabricate a model on a different laser cutter or from a different material. The reason for this lies in the way mounts and joints hold objects in place, which is by forc-ing them into slightly smaller openings. Such "press fit" mechanisms unfortunately are susceptible to the small changes in diameter that occur when switching to a ma-chine that removes more or less material ("kerf"), as well as to changes in stiffness, as they occur when switching to a different material. We present a software tool called springFit that resolves this problem by replacing the problematic press fit-based mounts and joints with what we call canti¬lever-based mounts and joints. A cantilever spring is simply a long thin piece of material that pushes against the object to be held. Unlike press fits, cantilever springs are robust against variations in kerf and material; they can even handle very high variations, simply by using longer springs. SpringFit converts models in the form of 2D cutting plans by replacing all contained mounts, notch joints, finger joints, and t-joints. In our technical evaluation, we used springFit to convert 14 models downloaded from the web.

Creating new 3D printed objects by recombining models found in hobbyist repositories has been referred to as "re-mixing". In this paper, we explore how to best support users in remixing a specific class of 3D printed objects, namely those that perform mechanical functions. In our survey, we found that makers remix such machines by manually extracting parts from one parent model and combine it with parts from a different parent model. This approach often puts axles made by one maker into bearings made by another maker or combines a gear by one maker with a gear by a different maker. This approach is problem-atic, however, as parts from different makers tend to fit poorly, which results in long series of tweaks and test-prints until all parts finally work together. We address this with our interactive system grafter. Grafter does two things. First, grafter largely automates the process of extracting and recombining mechanical elements from 3D printed machines. Second, it enforces a more efficient approach to reuse: it prevents users from extracting indi-vidual parts, but instead affords extracting groups of me-chanical elements that already work together, such as axles and their bearings or pairs of gears. We call this mecha-nism-based remixing. In a final user study, all models that participants had remixed using grafter could be 3D printed without further tweaking and worked immediately.

We present a new haptic device that enables blind users to continuously interact with spatial virtual environments that contain moving objects, as is the case in sports or shooter games. Users interact with DualPanto by operating the me handle with one hand and by holding on to the it handle with the other hand. Each handle is connected to a pantograph haptic input/output device. The key feature is that the two handles are spatially registered with respect to each other. When guiding their avatar through a virtual world using the me handle, spatial registration enables users to track moving objects by having the device guide the output hand. This allows blind players of a 1-on-1 soccer game to race for the ball or evade an opponent; it allows blind players of a shooter game to aim at an opponent and dodge shots. In our user study, blind participants reported very high enjoyment when using the device to play (6.5/7).

Foot interfaces, such as pressure-sensitive insoles, still yield unused potential such as for implicit interaction. In this paper, we introduce CapSoles, enabling smart insoles to implicitly identify who is walking on what kind of floor. Our insole prototype relies on capacitive sensing and is able to sense plantar pressure distribution underneath the foot, plus a capacitive ground coupling effect. By using machine-learning algorithms, we evaluated the identification of 13 users, while walking, with a confidence of ∼95% after a recognition delay of ∼1s. Once the user's gait is known, again we can discover irregularities in gait plus a varying ground coupling. While both effects in combination are usually unique for several ground surfaces, we demonstrate to distinguish six kinds of floors, which are sand, lawn, paving stone, carpet, linoleum, and tartan with an average accuracy of ∼82%. Moreover, we demonstrate the unique effects of wet and electrostatically charged surfaces.

For visually impaired users, making sense of spatial information is difficult as they have to scan and memorize content before being able to analyze it. Even worse, any update to the displayed content invalidates their spatial memory, which can force them to manually rescan the entire display. Making display contents persist, we argue, is thus the highest priority in designing a sensemaking system for the visually impaired. We present a tactile display system designed with this goal in mind. The foundation of our system is a large tactile display (140x100cm, 23x larger than Hyperbraille), which we achieve by using a 3D printer to print raised lines of filament. The system's software then trades in this space in order to minimize screen updates. Instead of panning and zooming, for example, our system creates additional views, leaving display contents intact and thus supporting users in preserving their spatial memory. We illustrate our system and its design principles at the example of four spatial applications. We evaluated our system with six blind users. Participants responded favorably to the system and expressed, for example, that having multiple views at the same time was helpful. They also judged the increased expressiveness of lines over the more traditional dots as useful for encoding information.

We present an exploration into the future of fabrication, in particular the vision of mobile fabrication, which we define as "personal fabrication on the go". We explore this vision with two surveys, two simple hardware prototypes, matching custom apps that provide users with access to a solution database, custom fabrication processes we designed specifically for these devices, and a user study conducted in situ on metro trains. Our findings suggest that mobile fabrication is a compelling next direction for personal fabrication. From our experience with the prototypes we derive hardware requirements to make mobile fabrication also technically feasible.

Previous works illustrate that one's palm can reliably recognize 10 or more spatiotemporal vibrotactile patterns. However, recognition of the same patterns on other body parts is unknown. In this paper, we investigate how users perceive spatiotemporal vibrotactile patterns on the arm, palm, thigh, and waist. Results of the first two experiments indicate that precise recognition of either position or orientation is difficult across multiple body parts. Nonetheless, users were able to distinguish whether two vibration pulses were from the same location when played in quick succession. Based on this finding, we designed eight spatiotemporal vibrotactile patterns and evaluated them in two additional experiments. The results demonstrate that these patterns can be reliably recognized (>80%) across the four tested body parts, both in the lab and in a more realistic context.

We conducted an empirical investigation of wearable interactive rings on the noticeability of four instantaneous notification channels (light, vibration, sound, poke) and a channel with gradually increased temperature (thermal) during five levels of physical activity (laying down, sitting, standing, walking, and running). Results showed that vibration was the most reliable and fastest channel to convey notification, followed by poke and sound which shared similar noticeability. The noticeability of these three channels was not affected by the level of physical activity. The other two channels, light and thermal, were less noticeable and were affected by the level of physical activity. Our post-experimental survey indicates that while noticeability has a significant influence on user preference, each channel has its own unique advantages that make it suitable for different notification scenarios.

TurkDeck is an immersive virtual reality system that reproduces not only what users see and hear, but also what users feel. TurkDeck produces the haptic sensation using props, i.e., when users touch or manipulate an object in the virtual world, they simultaneously also touch or manipulate a corresponding object in the physical world. Unlike previous work on prop-based virtual reality, however, TurkDeck allows creating arbitrarily large virtual worlds in finite space and using a finite set of physical props. The key idea behind TurkDeck is that it creates these physical representations on the fly by making a group of human workers present and operate the props only when and where the user can actually reach them. TurkDeck manages these so-called "human actuators" by displaying visual instructions that tell the human actuators when and where to place props and how to actuate them. We demonstrate TurkDeck at the example of an immersive 300m2 experience in 25m2 physical space. We show how to simulate a wide range of physical objects and effects, including walls, doors, ledges, steps, beams, switches, stompers, portals, zip lines, and wind. In a user study, participants rated the realism/immersion of TurkDeck higher than a traditional prop-less baseline condition (4.9 vs. 3.6 on 7 item Likert).