The Road to Ubiquitous Personal Fabrication: Modeling-Free Instead of Increasingly Simple
SSpecial Issue on Pervasive Manufacturing
Florian Michahelles, Nadya Peek, and Simon Mayer
The Road to UbiquitousPersonal Fabrication:Modeling-free Instead ofIncreasingly Simple
Evgeny Stemasov
Institute of Media Informatics, Ulm University, Ulm, Germany
Enrico Rukzio
Institute of Media Informatics, Ulm University, Ulm, Germany
Jan Gugenheimer
T ´el ´ecom Paris - LTCI, Institut Polytechnique de Paris, Paris, France
Abstract —The tools for personal digital fabrication (DF) are on the verge of reachingmass-adoption beyond technology enthusiasts, empowering consumers to fabricatepersonalized artifacts. We argue that to achieve similar outreach and impact as personalcomputing, personal fabrication research may have to venture beyond ever-simpler interfaces forcreation, towards lowest-effort workflows for remixing. We surveyed novice-friendly DFworkflows from the perspective of HCI. Through this survey, we found two distinct approachesfor this challenge: 1) simplifying expert modeling tools (AutoCAD → Tinkercad), 2) enriching tools not involving primitive-based modeling with powerful customization (e.g., Thingiverse).Drawing parallels to content creation domains like photography, we argue that the bulk ofcontent is created via remixing (2). In this work, we argue that to be able to include the majorityof the population in DF, research should embrace omission of workflow steps, shifting towardsautomation, remixing, and templates, instead of modeling from the ground up. https://doi.org/10.1109/MPRV.2020.3029650©2021 IEEE. Personal use of this material is permitted. Permis-sion from IEEE must be obtained for all other uses, in any currentor future media, including reprinting/republishing this materialfor advertising or promotional purposes, creating new collectiveworks, for resale or redistribution to servers or lists, or reuse ofany copyrighted component of this work in other works P ERSONAL FABRICATION (PF) describesthe notion that machinery, workflows, and toolsfor industrial manufacturing become available toconsumers. This – ideally – includes not onlytechnology enthusiasts, but also less ”tech-savvy”
Pervasive Computing
Published by the IEEE Computer Society © 2021 IEEE a r X i v : . [ c s . H C ] J a n sers. They may still desire to benefit from theopportunities of PF, such as tailored artifactsthey are unable to order online easily (e.g.,non-standardized attachments [1]–[3]). However,these potential users may not be convinced toinvest time in skill acquisition and PF processes.They may not be ”makers” and may likewise notbe convinced to learn digital fabrication (DF) fortheir benefit – especially when their alternativesare ever-improving online shopping ”workflows”.Just like computing itself progressed fromcentralized use for few, often expert, user groupsto personal and ubiquitous computing, (personal)fabrication is likely on a similar path. Ultimately,as described by Gershenfeld, we may have ma-chines able to fabricate anything [4]. Such a de-vice is still constrained by the input it may receive– i.e., ”what to fabricate?”, currently answeredby the use of (computer-aided) design software.We want to approach these developments fromthe perspective of HCI. Namely, the assumptionthat machines able to fabricate in any materialand size will be available to users in the sameway powerful word, video, and image processorsbecame available to and actively used by them.This empowers users of PF devices to benefitfrom digital precision to create and shape matter– both for productive and mundane purposes.However, mere ownership of or access tosuch devices (e.g., 3D-printers) along with thesoftware needed (e.g., CAD software), does notmake a person a user. While increasingly moremachine knowledge can be embedded in the hard-or software itself [1], users have to precisely express requirements for future artifacts (e.g.,dimensions). We consider the established notionof (3D-) modeling – defining shapes based onsimple primitives such as lines or voxels – to be ahindrance for widespread adoption of PF. Defin-ing artifacts from the ground up is appropriatefor domain experts or users enjoying it and pos-sessing intrinsic motivation for the process itself,not necessarily the result [5]. While paradoxicalat first, we argue that PF must provide ways forfuture users to benefit from intricately tailored,personal artifacts, without resorting to definingthem in great detail. Likewise, DF has to providesimple, low-effort tools for content creation thatenable users to explore the possibilities of the technology while generating quick, yet viable,results. Our main argument is that these low-efforttools should not be a simplified version of an ex-pert tool which follows a creation paradigm ”fromscratch”, but should rather be radically simple in-terfaces which omit most modeling and requiredexpert knowledge, reducing artifact creation to asfew interactions as possible. We propose a modelto differentiate modeling and remixing. We see ”remixing” as a gradient between two extremes: ”getting” (purchases in stores) and ”modeling” (designing and defining artifacts from the groundup). We further survey a set of recent literature inPF and categorize them within our gradient thatfocuses on effort as a core dimension.We partially ground our argument in paralleldevelopments that can be observed in the fields ofmusic, video, or image editing (e.g., GarageBand ⇔ Logic Pro, Instagram ⇔ Adobe Photoshop,and TikTok ⇔ Adobe Premiere). These facets of content creation have non-experts in photography,videography, or music, creating content for com-munities like TikTok or Soundcloud (Figure 1).By relying on automation and derivative work,users (unlikely to use expert tools) are enabledto explore and generate content without explicittraining, high entry barriers, and with low effort.One of the core arguments we propose in thiswork, is that one reason Instagram and TikTokwere able to bring content creation to the masses,is not only their associated communities, but alsotheir radical break with the ”creation paradigm”.Instead of giving users full control over the con-tent, as done by expert tools, creation follows ”getting” and ”remixing” paradigms, where theusers select from pre-defined tools and filters,which often leverage automation (e.g., face track-ing for videos). Standardizing these processes isan apparent reduction in expressivity. However,the combination of pre-defined operations stilloffers a sufficient variety to be able to pleasethe needs of the individual (who often evendepends on templates to realize and explore whatis possible). Flath et al. describe the growth ofThingiverse users as coinciding with the intro-duction of the customizer [6] – a way to enablenovices to create remixes without the skills tomodel the entire artifact. This customization as-pect enhanced a store-like interface (i.e., Thingi- Pervasive Computing eneral (online)stores (
Amazon )Simple contentcreation (
Instagram ) Complex contentcreation (
Photoshop ) Industrial CAD�
AutoCAD �Simplified CAD�
Tinkercad ) Effort P o t e n t i a l U s e r b a s e P r o x i m i t y t o D e s i r e d A r t i f ac t LowHighHighLow Low High
Efforts to simplifymodelling toolsEfforts toimproveremixing
Figure 1.
Simplified relation between effort (of a workflow) and the potential gap between desired andachievable artifact. Various prototypes simplify complex modeling tools (green arrow). We argue that theopposite direction (purple), based on lowest-effort interfaces for shopping and content creation, may involvethe population in PF. verse, ”getting” ) with options to tailor artifactsto one’s liking and is an inherently different stepthan the simplification of established industry-grade CAD tools (e.g., AutoCAD) to simpler ones(e.g., Tinkercad), as depicted in Figure 1. Basedon such parallels and our survey, we synthesizeconcepts crucial for the further dissemination ofPF as a relevant aspect of everyday scenarios:enabling derivative works, leveraging automation,crowds, and communities, along with a focus on ”remixing” and ”getting” , instead of ”modeling” by defining artifacts (or content) from the groundup. We see this work as a call to action thatPF may need a – possibly counterintuitive –paradigm shift to reach a wider audience beyondenthusiasts and domain experts, and to become a genuinely pervasive technology.
Ubiquitous Personal Fabrication
With this work, we focus on what Hudson etal. identify as ”casual makers”: people that careabout results, and less about processes [5], mak-ing them more akin to consumers than makers [1].With them being the majority of the popula-tion, their adoption of PF will enable ubiquitouspersonal fabrication – used for entertainment/ aesthetic purposes and functional, productivepurposes, as is the case with computing.Non-experts in video editing becameproficient generators of novel video content– not through ever simpler manifestations ofindustry-grade systems like Adobe Premiere, butrather through simple tools relying on derivative
January/March 2021 ineMechanismCrankPoint Point LineExteriorMusic Box Systemprovidesways toaggregateprimitivesSystemprovides2D or 3Dprimitives
Music Box
SystemprovidesalterationpossibilitiesSystemprovideslookup "Getting"
Systemprovideslookup (e.g., customizing a Thingiverse model)(e.g., via Amazon or Thingiverse) (i.e., designing from the ground up) "Remixing" "Modelling"
Figure 2. works coupled with communities (e.g., TikTok).The same can be presumed for images (i.e.,Instagram and Photoshop) or audio content (e.g.,through GarageBand). While the importanceof community and network effects can not bedisregarded, the tools these platforms deliver andembed in their workflow are compelling: theyrely heavily on derivative work (e.g., additionof imagery like stickers), automate previouslytedious processes (e.g., through filters) anddeliberately omit the majority of functions theirprofessional counterparts provide. The majorityof users will not generate sophisticated movieproductions or photographs. However, they aredeeply entwined in content creation for therespective domains. For PF, this may also bethe case: perfectly precise, industry-grade partsmay not be the aspect that drives the widespreadadoption of DF. Mundane, easy-to-create, low-effort artifacts that still generate value (e.g.,entertainment, tailored artifacts) for creators andinvolved communities will likely make up for abulk of artifacts made.
Modeling can be defined as ”to design or imi- tate forms” . We consider traditional 3D-modelingworkflows (CAD software) to be in line withthis definition: a user combining precise, funda-mental geometric primitives (e.g., lines) until anobject takes shape. When one takes existing arti-facts (e.g., a model from Thingiverse) to changethem, the primitive a user is working with isa finished, usable artifact. Alterations of it arethen a remix [6], [7]. While the extremes ofthe gradient between ”getting” and ”modeling” are distinct (i.e., getting a complete artifact withessentially ”one click”, compared to modelingit from primitives), the transition between thesetwo concepts is gradual, with the building blocks(i.e., primitives the users work with) progressingfrom fundamental shapes over their aggregationsto essentially finished artifacts – we treat thisspace as ”remixing” , where a degree of modelingwork is omitted through the system.To further clarify our definitions, we want tointroduce a tree-analogy to emphasize the fluidtransition between ”getting” over ”remixing” to ”modeling” artifacts (Figure 2). We considertrees as they are defined in computer science forthe process of defining / attaining an artifact: with Pervasive Computing root (desired artifact), nodes (subcomponents ofthe artifact) and leaves (fundamental geometricprimitives). We do not provide a formal way ofcreating such trees but argue that each artifact canbe decomposed into them. The more of the nodesusers have to define themselves, the closer theyoperate to ”modeling” (high effort, no pruning).The more nodes (and the closer they are to theroot) are provided to the users, the closer theyoperate to ”getting” (intermediate to low effort).In its simplest form, remixing is mainly ”getting” the object, the central paradigm for store inter-faces (low-effort, tree pruned right below root).An example can be seen in Figure 2 for a musicbox. It is comprised of mechanical (functional)components (a crank, gears, the cylinder definingthe melody) and an enclosure.If users are interested in ”getting” a(personalized) music box, they may refer tostores or repositories like Thingiverse. Theyprovide a lookup for the particular object theusers are searching for, and are able to deliverit to them. Thereby, users are enabled to pruneall subtrees below the finished artifact anddo not have to invest further work or occupythemselves with subtrees. This is possible, aslong as the lookup succeeded or the user acceptsan alternative artifact. We deliberately consideronline shopping to be a feasible ”branch” of PFwhere design and manufacturing have happenedalready without user intervention and maysurpass a user’s personal acceptance threshold(i.e., the product fits ”good enough”). To achievea higher degree of personalization, users mayconsider ”remixing” a personalized music box.For instance, by downloading a parametric designfrom Thingiverse and customizing it. A set ofsubtrees can be pruned, while the remainingsubtrees may require modeling work from theusers. Alternatively, they may require input ofparameters for a generative design (e.g., forembossed text). To achieve the highest degreeof control and personalization, users may resortto ”modeling” the music box. This requiresthe use of more sophisticated software thatprovides primitives to aggregate (e.g., via CSG,constructive solid geometry) to more complexparts. Users then combine 2D features like linesand rectangles to 3D features, which in turn are combined to higher-level components, likegears. Some systems provide both 2D and 3Dprimitives to reduce workload (allowing usersto prune leaves). The more work is omittedin the tree (i.e., through a lookup), the moreone can consider a process to be ”getting” .If an interface is able to deliver an artifactdirectly, one may consider it an interface for ”getting” objects. In contrast, if a user isrequired to aggregate a majority of componentsby combining primitives, one may consider theinterface to follow a ”modeling” approach.
Design Tools for Personal Fabrication
To emphasize these two lines of approachescurrently explored in research ( ”modeling” – ”remixing” ) and the established paradigm of ”getting” artifacts, we selected papers relatedto simplifying the definition of artifacts fromthe last 10 years from venues such as CHI,UIST or TOG. The initial set consisted of papers, which were distilled to worksused to illustrate the gradient. Our selectionis not a holistic overview of the field, butconsists of papers that highlight unique aspectsand approaches of the community to PF.The core focus is on the design of artifacts,and less their fabrication . While fabricationitself poses intricate challenges, it is likelythe least ”personal” aspect of PF – ideallyto be automated and optimized without userintervention to reduce effort. Systems allowingoverlap between fabrication and design [8]–[10]or augmenting handcraft [11] remain part ofthe classification, as they can be considered anabstraction of CAD tools and their complexities.As mentioned in our motivation, effort neededto achieve a satisfactory result is the core dimen-sion we consider relevant for ubiquitous adoptionof PF. Ideally, the more effort users invest in amodeling process, the closer they should get totheir envisioned artifact. However, this relationis not linear or increasing monotonously: low-effort interfaces exist both for artifact acquisi-tion (e.g., online shopping) and content creation(e.g., Instagram). Therefore, Figure 3 abstracts the”proximity to desired artifact (c.f., Figure 1) toa general notion of ”expressivity” – a subjective January/March 2021 ffort LowHigh Low High P o t e n t i a l E x p r e ss i v i t y Legend: "Traditional"�3D� Modelling ModalityTransfer Handcraft& Tangible Outsourcing& Repositories Automation SituatedTools
Industry-gradeCAD (AutoCAD)
Amazon Dash Mix&Match
Stemasovetal.�2020�
DesignandFabri-cationbyExample
Schulzetal.�2014�
Robiot
Lietal.�2019�
CopyCAD
Follmeretal.�2010�
General PurposeOnline Stores Thingiverse
Thingiverse(with Customizer)
AutoConnect
Koyamaetal.�2015�
Grafter
Roumenetal.�2018�
KidCAD
Follmer&Ishii(2012�
MixFab
Weicheletal.�2014�
D-Coil
Pengetal.�2015�
Paper3D
Paczkowskietal.�2014�
StrutModeling
Leenetal.�2017�
Printy3D
Yungetal.�2018�
SketchChair
Sauletal.�2011�
Instagram ExoSkin
Gannonetal.�2016�Leeetal.�2016�
PosingandActingasInput...
Fab forms
Shugrinaetal.�2015�
Kyub
Baudischetal.�2019�
PARTs
Hoffmanetal.�2018�
MiragePrinter
Yamaoka&Kakehi�2016�
FlatFitFab
McCraeetal.�2014�
CraftML
Yeh&Kim(2018�Ballagasetal.�2019�
PervasiveMakingUsingGenerativeModeling...
Reprise
Chenetal.�2016�
RoMA
Pengetal.�2018�
Makers' Marks
Savageetal.�2015�
Hobbyist CAD(Tinkercad)
Turn-by-Wire
Tianetal.�2019�
FusePrint
Zhuetal.�2016�
Figure 3.
A selection of tools arranged by effort and potential expressivity. We decompose effort into threeranges from ’getting’ over ’remixing’ to ’modeling’. Expressivity is similarly segmented, ranging from tools meantfor a single artifact, tools for specific domains to tools limited to entire object ranges. We classify systems bytheir core approach to artifact definition (background) and secondary traits (corner). measure of the possibilities a tool enables.A tool with low expressivity is able to deliverone single artifact – an example is the AmazonDash button, which is pre-configured to ”get” a specific product, chosen once before. A stepabove such systems in terms of expressivity,are systems limited to a specific domain. Forexample, systems for customizing glasses [12]or chairs [13], reduce the effort needed, byconstraining the result domain of the process.Tools exclusively intended to remix toys [14] aredomain-specific, while tools intended to remixany static shape are more expressive. Tools thatadditionally account for dynamics/kinematics canbe considered as even more expressive.We consider the potential expressivityof industry-grade CAD tools to be close topotentially unlimited, as they allow the designof highly complex artifacts (e.g., entire engines),along with their assembly and simulation. Whilenot omnipotent yet, we consider them to beon the verge of having no set limits to theirexpressivity – as long as appropriate effort isinvested .The method with the least effort neededhas the highest potential degree of adoption, when considering the entire population. Low-effort means low friction, low entry barriers, withusers quickly being able to succeed with theirtasks. Examples for such an interface are theaforementioned Amazon Dash buttons, as moststeps to acquire an (non-personalized) artifactare omitted. A step above such a lowest-effortinterface are general interfaces. General, ”all-purpose” stores cover a high expressivity, whilespecialized ones cover a lower degree thereof.Thingiverse, can be regarded as a specializedstore, bound by the current limits of DF hardware.If augmented with customization options, storeinterfaces inherit aspects of remix- or tailoring-oriented interfaces. This mainly refers to para-metric designs, as found for instance on Thingi-verse, but also services that provide customerswith the ability to tailor product lines as theydesire (e.g., furniture). The more the customiza-tion options, the more effort the users may haveto invest. We consider interfaces that embracederivative works to exhibit low to intermediateeffort, relative to established modeling paradigms.In PF, these are interfaces to repositories likeThingiverse [15], or interfaces for customizingparametric designs [16]. The definition of artifacts”from scratch” can be considered to be a high- Pervasive Computing ffort activity (with potentially high reward: aone-of-a-kind artifact). Primitives are aggregatedand combined with operations (e.g., CSG) and,with increasing detail, yield an exact rendition ofthe desired artifact. (3D) modeling can be consid-ered a dominant paradigm used in products andresearch prototypes to define arbitrary artifacts.Based on the retrieved set of literature (Figure3), we derived 6 approaches to reduce effort forPF: tools that are 1) situated, 2) automation-supported, 3) repository-based, 4) handcraft- ortangibility-oriented, 5) employing modality trans-fer and lastly 6) largely ”traditional” modelingtools. The horizontal arrangement of the works inFigure 3 is based on the degree of effort requiredand the degree of derivation a tool employs. Thevertical arrangement is based on the expressivityof the tools or their applicability to differentartifact classes. The positioning of the worksinside a sector was derived by mutual compar-isons with respect to their required effort andachieved expressivity. The following paragraphsintroduce systems that are representative for theseapproaches.
Situated tools (1) are an approach to bridgethe disconnect between the space a future artifactis meant to reside in, and the space it is beingspecified in. This allows easier embedding ofreal features [9], [15], [17], especially when theymight be hard to measure and digitize [18]. Theseaspects reduce effort in comparison to more dis-connected methods for modeling or remixing,while allowing users to preview their work beforeor during fabrication [8], [15]. Situated tools canbe either modeling tools incorporating the realworld as reference (e.g., [9], [19], pruning sub-trees of real-world features), or be remixing toolsembracing entire outsourced artifacts [15], [20](pruning at or close to the root). Situated toolsprovide the advantage of the correct physical con-text , thereby omitting steps like measurements,but may suffer from limited expressivity due toalternative input and output devices, compared toestablished 3D modeling workflows.
Automation (2) enables users to omit specificsteps after providing input to a design tool. Sys-tems relying on generative design can infer mod-els from sketches [13] or movement [21]. Theyare also able to transfer modalities like speech(i.e., descriptions of a product) to a personalized design [12]. Parametrized systems expose a lim-ited set of dimensions for users to explore andcustomize [16]. They may also rely on existingphysical objects as input to generate additionalgeometry [2], [22]. This class of systems eitherenriches modeling tools (pruning subtrees) orenables users to interact with simple interfaces toobject remixing [16] or generation [21] (pruningclose to the root). By automating steps, previ-ously complex procedures to define geometry areomitted. However, current automation approachesare constrained by the parametrized or previouslylearned geometries and can be limited to certainpredefined sets of objects.Works focusing on
Repositories and Out-sourcing (3) aim to benefit from finished modelsor features [15], [20], [23] at the expense ofexpressivity of the workflow, by including themin a design process. Systems may either rely onfeatures existing in the immediate surroundings ofthe user (e.g., [14], [15], [20]) or objects foundin crowd-based model repositories (e.g., [15],[23], [24]). Outsourcing may also happen throughautomation (work offloaded to the system) or byrelying on curated, centralized repositories (e.g.,[25]) or pre-defined parametric templates. Mostsuch systems can be assigned to the range of ”remixing” but may either require low effort [20]or are meant for more complex domains requiringmore modeling-like procedures [24]. Tools thatembrace outsourced work are generally limitedto the databases of artifacts they rely on – bothin terms of object diversity and how much isencoded in the objects (e.g., static geometry orparametrized geometry). They furthermore needrobust retrieval methods to benefit the user.Systems relying on
Handcraft and tangibletools (4) enable users to interact with their de-sign in a more immediate and tangible fashion.By providing tangible, pre-defined components,users may not only manipulate them directly,but also allow a system to replace placeholderswith complex geometry [26], allowing users toomit their design. This enables a more uncon-strained interaction with the design materials [27].Interactive fabrication is enabled by mediatedinput to the fabrication device [10] or augment-ing previously manual processes with comput-erized support [11]. This likewise enables usersto rely on existing features in their designs, or
January/March 2021 ase primitive-based modeling processes. Mostsuch systems rely on modeling as a paradigm– however, they are closer to its original def-inition, grounded in handcraft. With generativeaspects [26] and reliance on simpler buildingblocks [27], some venture close to remixing-like procedures. While tangible tools re-introducefeedback lost with most digital fabrication tech-niques, they likewise re-introduce a skill andlearning curve, thereby increasing effort in somecases. Modality Transfer (5) is likewise a commonoccurrence in literature. By avoiding traditionalCAD metaphors (e.g., aggregation of primitives),systems enable novices to express their require-ments. Sketching-based systems enable users toomit precision in their design process, whichis either not necessary to achieve a functionalartifact [13], or is inferred by a system relyingon artificial intelligence [21]. Examples includethe use of metaphors known from natural ma-terials [28], gestures [19], [29], speech [12] orprogramming [30]. Some systems that employ amodality transfer, still employ modeling as theircore paradigm [19], [28], [30], while others rad-ically omit explicit steps to define them from theground up [12], [29]. Tools employing modalitytransfer without omitting modeling processes gen-erally reduce effort with respect to learning, butless so with respect to effort during the processitself. Approaches that combine novel modalitieswith low-effort processes, reduce effort through-out the entire design process [12], [29]. ”Traditional” Modeling tools (6) areapproaches that, at their core, aim to simplifymodeling as a paradigm, thereby making it moreaccessible to novices. This includes the reductionof available primitives [31], operations [32],or fidelity [33]. While they generally achievea simpler (and often less expressive) processthan industry-grade toolchains, they rely onthe paradigm of modeling, which, while themost expressive, requires effort for the completedefinition of artifacts nonetheless.All previously presented approaches have incommon that they reduce the required effort forPF. However, a majority of them still employsmodeling-like approaches. Prototypes employinghigh degrees of automation [12], [22]) or ones that outsource modeling work to crowds [15],[24], venture close to ”modeling-free personalfabrication” and ”getting” . They enable users toquickly and often effortlessly receive uniquelypersonalized results without having to definethem from the ground up. These low-effort / high-expressivity procedures are crucial for personaldigital fabrication, both for productive, functionalartifacts, but also for content creation as such.DF can be a synthesis of mere artifact acquisition(e.g., online shopping) and content creation (e.g.,for platforms like Instagram). If appropriate, low-effort procedures are provided to users, DF itselfmay become a ubiquitous and inclusive endeavorfor users currently not involved in it.
Conclusion – A Call to Action
With this work, we wanted to emphasize thatpersonal fabrication research may and shouldfocus on ways to circumvent work and effortneeded to achieve one’s goals. We argue that ifPF is meant to be adopted by a wide range ofusers, the benefits (e.g., a personalized product)do not dictate the process (e.g., 3D modeling).The bulk of objects designed with the meansof personal fabrication will likely not exhibit acomplexity that would dictate precise and lengthyprocesses. We argue that both researchers andpractitioners should consider finding novel waysto omit processes akin to modeling, instead ofaiming to simplify them further and further.This is likely a crucial component to achievewidespread adoption in households, instead of fa-blabs and other (technology-)enthusiasts’ spaces,making PF as ubiquitous and relevant as personalcomputing itself became over time. A second,crucial component to widespread adoption of PFis the embedding thereof into less serious contentcreation contexts, linked to vibrant communities.Our main argument is that these low-effort toolsshould not be a simplified version of an experttool, but rather radically simple interfaces whichomit most of the process and required expertknowledge, reducing artifact creation to as fewinteractions as possible – ”getting” .Based on our survey, we derived specificaspects that may support system-driven HCI re-search to achieve, or venture towards modeling-free personal fabrication: a focus on deriva-tive works leveraging automation, crowds, and Pervasive Computing ommunities, with a general direction towards ”remixing” and ”getting” , instead of ”model-ing” .Pervasive PF can not ignore the environmentalimplications. One important future challenge forpervasive PF will be the potential environmentalimpact. Physical artifacts can not be as easilyremoved as mass-produced digital content suchas images. Therefore, our call for action is alsodirected to sustainability research that addressesthis issue. We argue that the path towards massusage is a realistic one (with many researchersfinding ways to enable and accelerate PF fornovices) and, therefore, the question of sustain-ability has to be addressed now, before it becomesubiquitous .For an actual novice in DF, the decision isnot between Tinkercad and AutoCAD, but ratherbetween ”Product A” and ”Product B” online, dueto the low-effort experience of ”getting” . Theinclusion of a majority of people in personal DFrequires different approaches than are currentlydominant in research. Low-effort approaches toDF will likely be the ones to enable ubiquitouspersonal fabrication.
REFERENCES
1. Patrick Baudisch and Stefanie Mueller, “Personal Fabri-cation”,
HCI , 2017.2. Xiang ’Anthony’ Chen, et al., “Reprise: A Design Tool forSpecifying, Generating, and Customizing 3D PrintableAdaptations on Everyday Objects”, in
Proceedings ofthe 29th Annual Symposium on User Interface Softwareand Technology , 2016.3. Rita Shewbridge, et al., “Everyday Making: IdentifyingFuture Uses for 3D Printing in the Home”, in
Proceed-ings of the 2014 Conference on Designing InteractiveSystems , 2014.4. Neil Gershenfeld, et al.,
Designing Reality: How toSurvive and Thrive in the Third Digital Revolution , BasicBooks, Inc., USA, 2017.5. Nathaniel Hudson, et al., “Understanding Newcomersto 3D Printing: Motivations, Workflows, and Barriersof Casual Makers”, in
Proceedings of the 2016 CHIConference on Human Factors in Computing Systems ,2016.6. Christoph M. Flath, et al., “Copy, Transform, Combine:Exploring the Remix as a Form of Innovation”,
J InfTechnol , 2017. 7. Celena Alcock, et al., “Barriers to Using, Customizing,and Printing 3D Designs on Thingiverse”, in
Proceed-ings of the 19th International Conference on SupportingGroup Work , 2016.8. Huaishu Peng, et al., “RoMA: Interactive Fabricationwith Augmented Reality and a Robotic 3D Printer”, in
Proceedings of the 2018 CHI Conference on HumanFactors in Computing Systems , 2018.9. Junichi Yamaoka and Yasuaki Kakehi, “MiragePrinter:Interactive Fabrication on a 3D Printer with a Mid-AirDisplay”, in
ACM SIGGRAPH 2016 Studio , 2016.10. Rundong Tian, et al., “Turn-by-Wire: ComputationallyMediated Physical Fabrication”, in
Proceedings of the32nd Annual ACM Symposium on User Interface Soft-ware and Technology , 2019.11. Huaishu Peng, et al., “D-Coil: A Hands-on Approachto Digital 3D Models Design”, in
Proceedings of the33rd Annual ACM Conference on Human Factors inComputing Systems , 2015.12. Rafael Ballagas, et al., “Exploring Pervasive MakingUsing Generative Modeling and Speech Input”,
IEEEPervasive Computing , 2019.13. Greg Saul, et al., “SketchChair: An All-in-One ChairDesign System for End Users”, in
Proceedings of theFifth International Conference on Tangible, Embedded,and Embodied Interaction , 2011.14. Sean Follmer and Hiroshi Ishii, “KidCAD: DigitallyRemixing Toys through Tangible Tools”, in
Proceedingsof the 2012 SIGCHI Conference on Human Factors inComputing Systems , 2012.15. Evgeny Stemasov, et al., “Mix&Match: Towards OmittingModelling Through In-Situ Remixing of Model Reposi-tory Artifacts in Mixed Reality”, in
Proceedings of the2020 CHI Conference on Human Factors in ComputingSystems , 2020.16. Maria Shugrina, et al., “Fab Forms: CustomizableObjects for Fabrication with Validity and GeometryCaching”,
ACM Trans Graph , 2015.17. Kening Zhu, et al., “FusePrint: A DIY 2.5D Printing Tech-nique Embracing Everyday Artifacts”, in
Proceedingsof the 2016 ACM Conference on Designing InteractiveSystems , 2016.18. Madeline Gannon, et al., “ExoSkin: On-Body Fabrica-tion”, in
Proceedings of the 2016 CHI Conference onHuman Factors in Computing Systems , 2016.19. Christian Weichel, et al., “MixFab: A Mixed-Reality Envi-ronment for Personal Fabrication”, in
Proceedings of theSIGCHI Conference on Human Factors in ComputingSystems , 2014.20. Sean Follmer, et al., “CopyCAD: Remixing Physical
January/March 2021 bjects with Copy and Paste from the Real World”, in Adjunct Proceedings of the 23Nd Annual ACM Sympo-sium on User Interface Software and Technology , 2010.21. Jiahao Li, et al., “Robiot: A Design Tool for Actuat-ing Everyday Objects with Automatically Generated 3DPrintable Mechanisms”, in
Proceedings of the 32ndAnnual ACM Symposium on User Interface Softwareand Technology , 2019.22. Yuki Koyama, et al., “AutoConnect: Computational De-sign of 3D-Printable Connectors”,
ACM Transactions onGraphics , 2015.23. Megan Hofmann, et al., “Greater Than the Sum of ItsPARTs: Expressing and Reusing Design Intent in 3DModels”, in
Proceedings of the 2018 CHI Conferenceon Human Factors in Computing Systems , 2018.24. Thijs Jan Roumen, et al., “Grafter: Remixing 3D-PrintedMachines”, in
Proceedings of the 2018 CHI Conferenceon Human Factors in Computing Systems , 2018.25. Adriana Schulz, et al., “Design and Fabrication by Ex-ample”,
ACM Transactions on Graphics , 2014.26. Valkyrie Savage, et al., “Makers’ Marks: PhysicalMarkup for Designing and Fabricating Functional Ob-jects”, in
Proceedings of the 28th Annual ACM Sympo-sium on User Interface Software & Technology , 2015.27. Amanda K. Yung, et al., “Printy3D: In-Situ TangibleThree-Dimensional Design for Augmented Fabrication”,in
Proceedings of the 17th ACM Conference on Interac-tion Design and Children , 2018.28. Patrick Paczkowski, et al., “Paper3D: Bringing Casual3D Modeling to a Multi-Touch Interface”, in
Proceedingsof the 27th Annual ACM Symposium on User InterfaceSoftware and Technology , 2014.29. Bokyung Lee, et al., “Posing and Acting As Input for Per-sonalizing Furniture”, in
Proceedings of the 9th NordicConference on Human-Computer Interaction , 2016.30. Tom Yeh and Jeeeun Kim, “CraftML: 3D Modeling IsWeb Programming”, in
Proceedings of the 2018 CHIConference on Human Factors in Computing Systems ,2018.31. Patrick Baudisch, et al., “Kyub: A 3D Editor for ModelingSturdy Laser-Cut Objects”, in
Proceedings of the 2019CHI Conference on Human Factors in Computing Sys-tems , 2019.32. James McCrae, et al., “FlatFitFab: Interactive Modelingwith Planar Sections”, in
Proceedings of the 27th An-nual ACM Symposium on User Interface Software andTechnology , 2014.33. Danny Leen, et al., “StrutModeling: A Low-Fidelity Con-struction Kit to Iteratively Model, Test, and Adapt 3DObjects”, in
Proceedings of the 30th Annual ACM Sym- posium on User Interface Software and Technology ,2017.
Evgeny Stemasov is a second-year PhD student atUlm University in Germany. He is interested in per-sonal fabrication along with the design and engineer-ing processes enabling it for novices and experiencedusers alike. Evgeny holds a Master’s Degree in MediaComputer Science from Ulm University. Contact himat [email protected]
Enrico Rukzio is full professor in the Institute ofMedia Informatics at Ulm University, Germany. He isinterested in designing intelligent interactive systemsthat enable people to be more efficient, satisfied andexpressive in their daily lives. Enrico received his PhDin computer science from the University of Munich.Contact him at [email protected]
Jan Gugenheimer is Assistant Professor for Com-puter Science at T ´el´ecom Paris (Institut Polytech-nique de Paris) in the DIVA group. He is workingon several topics around Mixed Reality (AugmentedReality and Virtual Reality) with focus on Human-Computer Interaction. Jan received his PhD in com-puter science from Ulm University. Contact him [email protected]