Authoring Object-Bound Haptics

Haptic perception involves actively exploring physical properties of objects through touch and movement. Lederman and Klatzky [33] distinguish substance-related properties such as texture, hardness, and weight, structure-related properties such as shape, and functional properties such as part motion. ISO 9241-910 [21] similarly distinguishes geometric and material properties within tactile and kinesthetic interaction. These accounts provide a perceptual basis for the haptic targets used in our taxonomy, but do not address how such properties are specified, physically realized, evaluated, and refined during authoring.

Existing haptic classifications organize systems according to perceptual or interaction roles [9, 31], technology and physical realization [4, 53], or the fidelity of the resulting feedback [44]. These frameworks characterize what haptic systems produce and how their output can be understood, but generally do not capture the authoring process through which a desired haptic behavior is created.

Research on haptic authoring more directly examines this process. Studies of haptic practice describe recurring design activities and challenges [57, 62], while Regimbal et al. [54] characterize authoring tools according to dimensions such as hardware, design approach, interface, and playback. This work primarily concerns haptic effects rendered through programmable output technologies. Such systems often support rapid iteration between specification and physical playback, while object-bound haptic authoring additionally requires the artifact-specific geometry, materials, mechanisms, or physical configuration through which the behavior is realized.

Adjacent fabrication research examines how designers move between digital representations and physical artifacts [52], but does not organize these workflows around their intended haptic outcomes. Our work therefore complements both perspectives by characterizing object-bound haptic authoring as a complete workflow spanning specification, digital and physical representations, physical realization, evaluation, and refinement. Compared with existing classifications, which describe what haptic systems produce or which authoring tools exist, our taxonomy classifies complete workflows by what is authored and how digital and physical stages are sequenced, which allows comparing systems at the workflow level and locating unsupported transitions. This allows systems from haptics, fabrication, and tangible interaction to be compared on common terms, and reveals where digital and physical stages are well connected and where transitions or forms of evaluation are poorly supported, patterns that tool- or output-centred classifications cannot expose.

We operationalized the scope of the review using the definition of object-bound haptic authoring introduced in Section 2. We included contributions that intentionally support the authoring or engineering of haptic behavior through properties of a physical or hybrid artifact.

We include a contribution when it meets the following criteria:

1. The intended haptic outcome is bound to a physical or hybrid artifact, such as an object, object part, mechanism, surface, material configuration, or proxy, rather than being rendered solely through a general-purpose haptic output device.
   
2. One or more artifact-specific properties are intentionally treated as design variables that meaningfully contribute to the haptic outcome. These properties may include geometry, material composition, surface structure, mass distribution, mechanical behavior, assembly, physical configuration, fabrication parameters, or embedded actuation.
   
3. The work contributes, describes, or exposes an authoring or engineering workflow through which the object-bound haptic behavior is specified, manipulated, predicted, physically realized, evaluated, or refined.
   

These criteria require the haptic behavior of the artifact to be an explicit design concern rather than an incidental consequence of designing a physical object. General-purpose CAD modelling, fabrication, or engineering work is therefore not included solely because the resulting object can be touched or because its physical properties may influence the user experience. This boundary is particularly important for haptic aspects involving shape, since the geometry of an object can be experienced both visually and through touch and manipulation. We include shape-oriented work only when the physically perceived shape is intentionally treated as a tactile or kinesthetic outcome, which we later refer to as felt shape / form. HapTwist [88] and StrutModeling [35] therefore fall within scope as their resulting geometry is explicitly considered in terms of how the object is physically experienced. In contrast, general-purpose CAD systems [74] are excluded.

A system can contribute an object-bound haptic artifact without exposing how its haptic behavior is authored. Snake Charmer [2] uses a robotic arm to position textured physical surfaces at the location of virtual objects, so that users can touch them in VR. It contributes an object-bound haptic system, the textured surfaces, but does not present an authoring system for them, and is therefore not aligned with our inclusion criteria.

We followed a literature review process with interpretive coding [77] and began by constructing a broad corpus of systems, tools, and workflows concerned with authoring object-bound haptics. Literature discovery took place between 2024 and August 2026, without restrictions on publication year. An initial scoping phase across haptic authoring was used to identify established work, relevant terminology, and the boundaries between object-bound authoring and haptic effect authoring. We expanded this initial collection through Google Scholar and ACM Digital Library searches, iteratively broadening the search vocabulary with terminology encountered in relevant publications. This included terms describing haptic qualities, artifact types, authoring activities, and realization methods, which subsequently informed the structured searches described below. Because relevant contributions are distributed across haptics, fabrication, tangible interaction, virtual reality, and mechanical design, and frequently do not describe themselves as haptic-authoring research, we subsequently used similarity-based exploration and backward and forward citation searches. Using ResearchRabbit, we explored publications related to the seed papers and their citation networks, assembling approximately 942 candidate records. Titles and abstracts were examined to identify potentially relevant contributions. We also manually browsed proceedings from selected haptics, HCI, fabrication, and virtual-reality venues, including ACM CHI, ACM UIST, ACM TEI, ACM DIS, ACM MUM, ACM VRST and ACM SCF.

Finally, we conducted structured searches in the ACM Digital Library and IEEE Xplore by combining terms describing the intended haptic quality, the type of physical artifact, the authoring activity, and the method used to realize the artifact. These terms included haptic, tactile, kinesthetic, texture, stiffness, and compliance; object, proxy, material, surface, and mechanism; authoring, design, prototyping, tuning, optimization, and inverse design; and digital fabrication, 3D printing, multi-material fabrication, and compliant mechanisms. Retrieved results were screened against the inclusion criteria described above. Given the interdisciplinary nature of object-bound haptic authoring and the varying terminology used across research communities, we combined structured database searches with purposive literature discovery [1, 67] through citation-network exploration and targeted venue browsing. Records were screened by title and abstract against the inclusion criteria, followed by full-text assessment of potentially eligible publications. Publications retained after this initial screening were subsequently assessed in full against the inclusion criteria.

Our literature search identified 2,150 records through ResearchRabbit (n = 942), structured database searches (n = 1,162), and supplementary citation and venue searches (n = 46). After removing 263 duplicates, 1,887 unique records underwent title and abstract screening, of which 1,705 were excluded. The remaining 182 publications were assessed in full against our inclusion criteria. Of these, 131 were excluded because they did not meet the criteria for object-bound haptic authoring. The final corpus comprised 51 papers, from which we identified 56 distinct authoring workflows.

The resulting corpus comprised 51 papers on object-bound haptic authoring. Most contributions appeared during or after 2010, consistent with the broader growth of haptic-authoring research reported by Regimbal et al. [54]. Our search focused on haptics, HCI, fabrication, and virtual-reality literature. The corpus therefore does not fully capture practice in domains where haptic quality has long been engineered, such as automotive interiors [32, 72], which we did not systematically consider.

Object-bound haptics authoring workflows formed our unit of analysis. Each paper in our corpus presents at least one workflow. We coded multiple workflows for a single paper when the work exposed substantially different authoring processes, either for the same haptic outcome or for different outcomes. Conversely, when several haptic qualities were addressed through the same underlying authoring process, we treated these as a single workflow.

The corpus was analyzed through several stages of iterative qualitative coding. One author first conducted open coding on a subset of 17 contributions. This stage identified provisional codes for recurring authoring activities and workflow elements, including parameter editing, demonstration of target haptic intent, simulation, optimization, fabrication, physical configuration, embodied testing, and post-fabrication fine-tuning of haptic results. The authors then collectively reviewed the initial codes, consolidated overlapping concepts, clarified distinctions, and developed a shared codebook.

After establishing the codebook, one author performed focused coding across the complete corpus. During coding, cases that did not fit the existing scheme were discussed, the codebook was iteratively refined, and previously coded contributions were revisited to maintain consistency. The corpus was coded by a single author, which may bias code assignment. To mitigate this risk, a second coder independently reviewed a subset of the coded corpus, 21 workflows, including contributions with varying workflows. Discrepancies were discussed and resolved through consensus, and the codebook was refined accordingly.

The resulting codebook captured three complementary aspects of object-bound haptic authoring: the research perspective from which a contribution approached the problem, the haptic qualities intentionally being authored, and the sequence of authoring, realization, and evaluation activities that constituted each workflow. These aspects form the three dimensions of the taxonomy presented next.

Object-bound haptic authoring spans work from different disciplinary perspectives. We distinguish between haptics-oriented contributions, which foreground the design, authoring, understanding, or evaluation of haptic behavior, and fabrication-oriented contributions, which foreground the materials, structures, mechanisms, or computational methods through which such behavior is realized. The distinction concerns the research focus rather than the technology used. For example, VoxelHap [16] explicitly targets tactile and kinesthetic feedback in virtual reality, whereas LamiFold [34] primarily contributes a fabrication workflow for objects with integrated mechanisms. We determine primary research focus from the motivation, stated contributions, and evaluation of each work rather than its venue or disciplinary origin.

This dimension captures the haptic quality that a workflow intentionally seeks to design, reproduce, or modify. Building on classifications of object properties in haptic perception [21, 33], we distinguish five targets:

Texture / surface feel concerns local surface properties such as roughness, friction, patterns, or microgeometry, as in Tabby [68] and ConTextural [27].

Hardness / softness / compliance concerns deformation under applied force, including stiffness, elasticity, flexibility, and compressibility [30, 60].

Felt shape / form concerns geometry intentionally experienced through touch or physical manipulation, such as volume, contour, or graspable form [7, 88]. General geometric modelling is not sufficient. We code Felt shape / form only when how the geometry is experienced through touch or physical manipulation is itself an explicit design or evaluation concern.

Weight / mass / inertia concerns perceived mass properties, including weight, balance, mass distribution, and inertia [51, 71].

Part motion concerns the tactile and kinesthetic behavior of moving object parts, including motion, resistance, friction, spring behavior, detents, clicks, snaps, and end stops [49, 87].

A workflow may address multiple haptic targets. We assign a target only when it is an explicit part of the intended haptic outcome rather than an incidental consequence of another design decision. For example, Microstructures to Control Elasticity in 3D Printing [60] is coded for compliance but not texture, since only compliance is intentionally authored.

Digital and physical input capture how a designer specifies or revises the intended haptic behavior by modifying a digital or physical representation. We identify six authoring approaches that can occur in both physical and digital spaces.

Manual design and engineering develops the haptic realization largely from scratch, for example by designing geometry, materials, or mechanisms in CAD or by physically constructing them from existing components.

Parameter tuning modifies predefined variables of an existing design, such as dimensions, texture density, material distribution, tension, preload, or mass position. Parameters may expose technical values or more perceptual controls such as softer, stiffer, rougher, or smoother [71].

Composition of primitives combines reusable elements that already encode relevant geometric, material, mechanical, or haptic behavior, either digitally or physically [16, 34]. Primitives may themselves be parameterized, as in Kinergy [19], combining parameter tuning and composition of primitives.

Spatial annotation assigns a haptic property, structure, or behavior to a location or region of an existing representation rather than constructing its realization directly [18].

Demonstration of target shape or behavior specifies the intended outcome by demonstrating its motion, deformation, shape, or other behavior through digital or physical manipulation [24, 29].

Selection among presented alternatives lets the designer specify or refine an outcome by choosing among predefined, example-based, or computationally generated alternatives [7, 30, 71]. This differs from composition of primitives, where the designer constructs a result by combining or arranging reusable elements.

Physicalize captures the transition from a digital representation to a physically experienceable form. We distinguish three approaches:

Digital fabrication realizes a digital specification through computer-controlled fabrication, such as 3D printing, laser cutting, or CNC milling.

Guided physical realization manually realizes a digitally specified design using physical materials, components, or primitives, with the digital representation guiding construction or assembly. HapTwist [88], for example, computationally derives a structure that is subsequently assembled physically.

Active haptic rendering uses a programmable haptic device or actuator to physically render all or part of a digitally specified behavior. It can serve as a preview before the final object-bound artifact is realized. Interactive Material Design [80], for example, renders the mechanical response of a digital material before fabrication. It can also be part of the final artifact, where programmed actuation is combined with artifact-specific properties. In Shifty [85], for example, the mechanism and movable mass are engineered into the physical proxy, while actuation renders a changing mass distribution. Haptic-effect authoring can therefore form part of an object-bound haptic authoring workflow, either as an intermediate step or as part of the final result.

Multiple Physicalize labels may occur within the same workflow, for example when active haptic rendering precedes digital fabrication or fabricated parts subsequently require guided assembly.

Digitize captures the transfer of haptic-relevant information from a physical artifact or interaction into digital form for subsequent authoring. We distinguish three approaches:

Model-linked transfer applies when information from a physical artifact or interaction is used to update an existing digital representation. This may happen indirectly, when physically evaluating an artifact informs which parameter or feature of its digital model should be changed, or directly, when changes to a physical representation are automatically reflected in a linked digital model as in tangible modeling.

Capture and reconstruction creates new digital data or representations by capturing properties of a physical artifact, for example through photography, scanning, tracking, or reconstruction. BrickStARt [64], for example, reconstructs a physically assembled brick model as digital geometry for subsequent refinement.

Haptic probing deliberately stimulates or manipulates a physical artifact and measures its haptic or mechanical response, such as texture, deformation, stiffness, friction, motion, or force response [12, 42].

Digital evaluation captures how a designer assesses an object-bound haptic design through a digital representation. We distinguish three approaches:

Visual inspection assesses visible design properties such as geometry, surface structure, material distribution, or component configuration, including conventional CAD renderings.

Kinematic or motion preview visualizes how an artifact or its parts move or change configuration, without necessarily predicting forces or deformation.

Physics-based simulation predicts physical or mechanical behavior relevant to the intended haptic outcome, such as deformation, stiffness, force response, contact, or friction.

Physical evaluation captures when a designer assesses the haptic behavior of a physically realized artifact. We identify one recurring approach:

Physical try-out involves directly experiencing the artifact by touching, deforming, lifting, or manipulating it. We include both explicit evaluation steps and informal handling that was not reported as evaluation but informs subsequent refinement, such as flexing a printed part and noticing that it is too stiff.

Our coding did not reveal sufficiently distinct recurring forms of physical evaluation to warrant additional labels. Across the corpus, physical evaluation predominantly takes the form of direct physical try-out, with comparatively little author-facing support for more structured or quantitative characterization of the realized haptic behavior, such as force measurements, stiffness or strength tests, friction measurements, or instrumented motion analysis. We briefly describe how this physical evaluation is reported at the paper-level (Section 7).

Combining the six stages and their labels allows us to represent complete authoring workflows. For example, Stay Tuned [49] supports digital design and parameter tuning of a bistable mechanism, digital inspection, fabrication, physical evaluation, and post-fabrication tuning. We represent this workflow as:

Digital input (Manual design and engineering, Parameter tuning) → Digital evaluation (Physics-based Simulation) → Physicalize (Digital fabrication) → Physical evaluation → Physical input (Parameter tuning).

Applying this representation across the corpus resulted in 56 workflows. Every stage occurrence in these workflows could be described with the labels defined above, and no additional labels were required. The complete coding is provided in Appendix A.

Across the 56 workflows, Part motion was the most common haptic target, occurring in 24/56 workflows. Its prominence may partly reflect the breadth of this category, which encompasses different types of motion, constraints, and mechanism behavior, together with the substantial body of fabrication research concerned with engineering moving mechanisms. Felt shape / form occurred in 18/56 workflows and Texture / surface feel in 15/56. In contrast, Hardness / softness / compliance occurred in 9/56 workflows and Weight / mass / inertia in 7/56. These lower frequencies suggest that compliance and mass-related properties remain comparatively underexplored as haptic qualities in object-bound haptic authoring.

Because a workflow may receive more than one haptic-target label, these counts are not mutually exclusive. This occurred in 11/56 workflows, the most common combination of haptic target that was authored within a single workflow was Part motion and Felt shape / form (5/56). Make It Stand [51], for example, exposes both shape and mass distribution as editable properties within the same authoring workflow. Overall, however, most workflows focused on a single haptic target. This is notable because object-bound haptic artifacts simultaneously exhibit properties such as shape, surface texture, compliance, weight, and mechanical behavior, which may also interact perceptually [39]. Authoring several of these haptic properties together therefore remains comparatively underexplored in state-of-the-art literature. When different haptic targets involve even slightly different authoring processes, we represent them as separate workflows; workflows with multiple targets authoring may therefore be slightly understated.

Digital input appears in 45/56 workflows and is dominated by Parameter tuning (32/45), followed by Composition of primitives (19/45) and Spatial annotation (17/45). Manual design and engineering (9/45) and Demonstration of target shape or behavior (9/45) occur less frequently, while Selection among presented alternatives is rare (2/45). Individual digital input approaches, however, still vary considerably in how they are instantiated. Digital parameter tuning, for example, ranges from parameters controlling characteristics of geometric mechanism, such as spring dimensions and travel in Shape-Haptics [87], to more perceptually meaningful controls such as the perceived stiffness in Ondulé [18]. Kinergy [19] illustrates how multiple digital input approaches can also be combined by composing predefined mechanical primitives and subsequently tuning their behavior.

Physical input occurs in 28/56 workflows and shows a different distribution. Composition of primitives is most common (18/28), followed by Parameter tuning (7/28) and Manual design and engineering (6/28), while the remaining approaches occur only sporadically. Physical composition typically relies on tangible building blocks, as in VoxelHap [16] and Mod2Hap [20]. This contrast reflects two common ways of authoring object-bound haptics: digital tools frequently expose editable parameters or regions of a design, while physical authoring more often proceeds by assembling or reconfiguring tangible elements.

The preferred input approaches also differ across haptic targets (Full overview in Appendix B). Texture / surface feel and Hardness / softness / compliance rely strongly on digital Parameter tuning (9/13 and 8/9 workflows respectively) and Spatial annotation (5/13 and 5/9). Touch-n-Go [65], for example, lets designers assign and tune texture properties over selected regions of a model. Part motion authoring similarly makes extensive use of digital Parameter tuning (16/19), followed by Composition of primitives (12/19), reflecting the frequent use of reusable mechanisms to construct motion behavior. Physical authoring of part motion is also primarily based on Composition of primitives (8/14).

Felt shape / form shows a different pattern. Digital input is spread relatively evenly across several approaches, including Parameter tuning (4/10), Composition of primitives (4/10), Manual design and engineering (3/10) and Spatial annotation (3/10). Physical input, in contrast, is strongly dominated by Composition of primitives (12/14), as in TanGi [15], VoxelHap [16], and BrickStARt [64]. This suggests that felt shape can be specified digitally through several representations, while physical authoring more often relies on directly constructing or reconfiguring tangible geometry. Weight / mass / inertia, by contrast, is predominantly authored digitally through Parameter tuning (4/7) and Spatial annotation (3/7), with relatively little physical input. HapticPrint [71], for example, allows designers to assign different weight properties to regions of a digital model, which are subsequently realized through fill material.

Digital and physical input can also complement one another within the same workflow. ReCompFig [81] first lets designers specify kinematic modes and degrees of freedom digitally. After fabrication, the mechanism remains physically adjustable by tightening cables to selectively constrain these degrees of freedom. The workflow therefore combines digital specification with subsequent physical tuning of the realized mechanism.

The Physicalize and Digitize stages show a pronounced asymmetry. Physicalize occurs in 53/56 workflows, whereas Digitize (15/56) is much less common. Object-bound haptic authoring therefore predominantly moves from digital representations toward physical realization, while transferring information from physical artifacts back into digital form occurs far less frequently.

Separating workflows by their paper's primary research focus shows that this asymmetry also persists within haptics-oriented research. Among these workflows, Physicalize occurs in 22/25, whereas Digitize occurs in 10/25. The difference is more pronounced among fabrication-oriented workflows, where Physicalize occurs in 31/31 and Digitize in only 5/31. The asymmetry is therefore present in both subsets, rather than being limited to fabrication-oriented research.

Our review cannot establish why the asymmetry exists, but two factors plausibly contribute. First, our inclusion criteria require an artifact-bound outcome and most of the corpus is fabrication-oriented, so digital-to-physical transitions may be more common given the definitions and inclusion criteria of our review. Second, Haptic probing appears uncommon, with Haptic probing being the least common form of Digitize (3/15).

Physicalize is predominantly realized through Digital fabrication (44/53), while Guided physical realization (13/53) and Active haptic rendering (6/53) occur less frequently. This reflects the strong role of digitally specified geometry, material structures, and mechanisms in the corpus. Different forms of physicalization can also complement one another. FabriClick [17], for example, combines digitally fabricated textile components with guided manual alignment and assembly. Active haptic rendering is comparatively rare and instead uses programmable haptic devices to make aspects of a digitally specified behavior physically experienceable before or without fabricating the complete artifact.

Among 15 workflows containing Digitize, Capture and reconstruction is most common (9/15), followed by Model-linked transfer (6/15) and Haptic probing (3/15). Digitization serves two broad roles. It can provide an initial digital representation of a physical reference, as in TactStyle [14], or connect physical authoring back to subsequent digital stages, as in Makers’ Marks [56]. Notably, most digitization workflows capture geometry, configuration, or other information whose relation to the haptic outcome is already known. Directly measuring haptic or mechanical behavior through Haptic probing is uncommon. This suggests that closing the loop from physically experienced haptic behavior back to a digital representation remains comparatively underexplored in object-bound haptic authoring.

Some differences also emerge across haptic targets (full details in Appendix C). Texture / surface feel is almost exclusively physicalized through Digital fabrication (15/15), with only one workflow additionally using Guided physical realization and none using Active haptic rendering. In contrast, digitization of texture is comparatively diverse, with Capture and reconstruction occurring in 4/6 workflows, Haptic probing in 2/6, and Model-linked transfer in 1/6. Capturing Tactile Properties [12] is notable in that it directly probes real surfaces to capture microgeometry for subsequent digital reproduction. Haptic Target Felt shape / form shows a broader range of transitions. It is physicalized through Digital fabrication (10/15), Guided physical realization (5/15), and Active haptic rendering (2/15). Digitization is also more common for felt shape, particularly through Capture and reconstruction (5/7) and Model-linked transfer (3/7). VoxelHap [16], for example, continuously tracks the configuration of physical voxels in a corresponding digital model, while PneuSeries [8] illustrates the less common use of Active haptic rendering to physically instantiate a digitally specified shape.

By contrast, Digitize is uncommon for Hardness / softness / compliance, Weight / mass / inertia, and Part motion. For compliance, only one workflow uses Digitize, through Model-linked transfer in Reconfigurable Elastic Metamaterials [83]. For weight, only one workflow uses Digitize, through Capture and reconstruction in BrickStARt [64]. For part motion, Model-linked transfer, Capture and reconstruction and Haptic probing each occur in only 1/2 workflows. Makers’ Marks [56], for example, uses Capture and reconstruction to capture annotated functional elements such as hinges and reconstructs them digitally. VoxelHap [16] uses Haptic Probing to record interaction-induced vibrations from real-world objects using a contact vibration microphone. Model-linked transfer, also for VoxelHap [16] captures the assembled Voxel configuration, allowing the placement of RotationVoxels to directly define rotational degrees of freedom in the corresponding digital model. These low frequencies are notable because these properties are, however, routinely physically experienced after realization, yet the resulting physical behavior is rarely transferred back into digital form. Overall, these observations suggest that current object-bound haptic authoring provides extensive support for turning digital designs into physical artifacts, but substantially less support for capturing the resulting haptic behavior and feeding it back into subsequent digital authoring.

Evaluation is prevalent throughout the corpus. Digital evaluation occurs in 53/56 workflows, while Physical evaluation occurs in all 56. The latter partly follows from the object-bound scope of the corpus, since each workflow ultimately produces a physically experienceable haptic outcome and we also code handling of the resulting artifact as Physical evaluation. Digital evaluation most commonly relies on Visual inspection (37/53), while Kinematic or motion preview (12/53) and Physics-based simulation (4/53) are considerably less frequent (full overview in Appendix D).

Although Physical evaluation occurs in all 56 workflows, this does not imply that every paper explicitly describes it as an authoring step. To examine this, we separately assessed how physical evaluation is reported across the 51 papers. Physical assessment is explicitly reported in 43/51 papers, while 22/51 report objective physical measurements. The latter include laboratory characterization and research experiments, which may validate research contributions without necessarily being author-facing stages in the resulting workflow. However, even when physical assessment is not explicitly reported, users of the workflow can still physically evaluate their artifacts and use their observations to refine subsequent designs. Explicit transfers of information from physical artifacts back into digital models are coded separately as Model-linked transfer.

Visual inspection occurs across all haptic targets, but what can be assessed visually differs considerably. It is particularly common for Texture / surface feel (14/15) and Felt shape / form (15/15), where surface structure or geometry can often be inspected directly. Part motion is also visually inspected in 7/21 workflows, typically by examining the geometry or configuration that enables the motion rather than the resulting haptic behavior itself [18, 65]. It also occurs for Hardness / softness / compliance (8/9) and Weight / mass / inertia (7/7), but here the visualization generally exposes design features that determine the eventual haptic outcome, such as material structure or mass distribution, rather than the felt property itself. Visual inspection therefore provides broad support for evaluating object-bound haptic designs, but often only indirectly communicates the resulting haptic experience.

More behavior-specific forms of digital evaluation are concentrated around Part motion. Kinematic or motion preview occurs in 11/21 Part motion workflows and Physics-based simulation in 3/21. Kinergy [19], for example, lets designers preview the motion produced by embedded kinetic units before fabrication, while Metamaterial Mechanisms [22] simulates the mechanical response of compliant cell structures. Texture / surface feel uses Kinematic or motion preview once (1/15): Metamaterial Textures [23] previews how the surface texture changes as the structure compresses. Physics-based simulation occurs only once for Hardness / softness / compliance (1/9), in Interactive Material Design [80], where deformation behavior is simulated interactively during authoring. Neither motion preview nor physics-based simulation occurs for Felt shape / form or Weight / mass / inertia in our corpus.

Physical evaluation, by contrast, occurs for every haptic target and directly exposes the resulting haptic behavior through interaction with the artifact. Depending on the target, this may involve rubbing a surface, pressing a compliant structure, grasping or lifting an object, or manipulating a moving part. Our coding did not reveal recurring forms of physical evaluation beyond this direct physical experience. Taken together, these findings show that final haptic outcomes are typically assessed through direct physical try-out, with little support for more structured or quantitative physical characterization. Digital evaluation, meanwhile, predominantly supports inspection of the underlying design and only occasionally predicts the resulting haptic behavior before physical realization.

Examining individual stages reveals which authoring activities are supported, but not how these activities are combined into complete workflows. We therefore compare the ordered stage sequences across the corpus and identify recurring higher-level workflow structures. These structures summarize recognizable ways of moving from haptic intent to a physically experienceable result. They are not additional taxonomy labels and are not necessarily mutually exclusive, since a single system may support more than one such workflow.

Object-bound haptic authoring has largely developed from fabrication, computational design, and mechanism engineering. Most papers in our corpus are fabrication-oriented, and Digitally Authored Fabrication is the dominant recurring workflow. Correspondingly, digital authoring commonly exposes geometry, material structures, mechanism parameters, or fabrication parameters that indirectly determine what will ultimately be felt.

This creates a gap between haptic intent and the representations through which it is authored. Parameter tuning dominates digital input, while approaches based on demonstrating a desired behavior or selecting among perceptually meaningful alternatives remain comparatively uncommon. Future tools could expose higher-level haptic properties more directly, for example allowing designers to specify that a mechanism should feel stiffer, more pronounced, or smoother while computational methods translate these intentions into appropriate artifact-specific parameters.

Our findings show a strong asymmetry between moving from digital representations to physical artifacts and bringing information from realized artifacts back into digital authoring. Physicalize occurs in almost every workflow, whereas Digitize is much less common. Even when digitization occurs, it predominantly captures geometry or physical configuration rather than directly characterizing the resulting haptic behavior.

This is particularly limiting because object-bound haptic behavior often becomes fully apparent only after physical realization. The resulting artifact should therefore be considered not only an output of the authoring process, but also an important source of information for subsequent iterations. Future systems could capture properties such as force response, stiffness, friction, deformation, or mass distribution and associate them with the corresponding digital model. Such information could help designers understand differences between intended and realized behavior and inform subsequent design iterations. Future work could also extend the review to industrial engineering practice using the terms mentioned.

The unified model can also serve as a lens for considering how existing workflows might be extended with additional forms of authoring support. StrutModeling [35], for example, transfers physical modifications to a digital representation. A future extension could additionally analyse that digital model and provide concrete guidance for modifying the physical configuration, thereby supporting a more tightly coupled digital–physical authoring loop.

A related limitation concerns how designers evaluate haptic behavior. Although Digital evaluation is widespread, it predominantly consists of visual inspection. Designers can therefore inspect the geometry, structure, or material distribution responsible for a haptic outcome, but only rarely receive a prediction of the resulting haptic behavior. Kinematic preview and physics-based simulation provide richer feedback, but occur comparatively infrequently.

At the physical side of the workflow, every artifact is evaluated through direct physical experience, yet this evaluation is predominantly informal. Our coding revealed little recurring authoring support for systematically characterizing the realized behavior. Future systems could therefore strengthen evaluation on both sides of physical realization. Simulation or active haptic preview could provide earlier feedback on expected behavior, while instrumented measurements could complement direct physical try-out after fabrication.

Active haptic rendering already demonstrates one possible bridge, but occurs in only a small subset of workflows. Programmable haptic devices, reconfigurable interfaces, or other intermediate physical representations could allow designers to experience selected aspects of a design before committing to its final fabrication. Such previews need not reproduce the complete artifact to be useful. Their purpose could instead be to expose the haptic properties most relevant to the current design decision.

Current workflows also tend to address individual haptic qualities in isolation. Only a minority intentionally author multiple haptic targets, even though real artifacts simultaneously exhibit shape, texture, compliance, weight, and mechanical behavior. Future authoring environments could make dependencies between these properties more explicit and help designers reason about trade-offs when modifying one property affects several aspects of the resulting experience.

Several of these opportunities also suggest closer integration with haptic-effect authoring. Haptic-effect authoring has developed methods for rapid replay, perceptually meaningful representations, reusable effects, and programmable rendering. Object-bound authoring contributes complementary strengths in artifact-specific geometry, materials, mechanisms, fabrication, and post-fabrication refinement. Future systems could combine these approaches, allowing designers to rapidly specify and preview intended behavior before translating it into a physically realized artifact and subsequently refining that artifact based on physical experience.

More broadly, the taxonomy and unified model provide a way to reason about where an authoring system provides support and where transitions remain unsupported. They may also offer a basis for comparing future systems at the workflow level, considering not only whether a desired haptic outcome can be produced, but also where iteration occurs, how readily alternatives can be explored, and how effectively information from one iteration informs the next.