PaperlayerView exact revision on arXiv

Source-linked AI summary

User Preferences for UI Anchoring in MR: Effects of Task Mobility and Interface Properties

João Belo, Sina Elahimanesh, Anna Maria Feit

arXiv:2608.28064v1cs.HC

TL;DR

MR anchoring is important, but user preferences across mobility contexts and interface types remain insufficiently understood. The paper addresses this gap with a mixed-methods study allowing participants to configure anchoring across conditions and tasks. It finds a shift from strong world-anchoring preferences when stationary toward body-relative anchoring during movement, without a single dominant body anchor.

  • Problem

    User preferences for MR anchoring across mobility contexts and interface types remain poorly understood because prior work has focused mainly on isolated tasks or fixed configurations.

  • Method

    The authors conducted a mixed-methods within-participants study in which users freely configured anchoring modes across three mobility conditions and multiple interface types.

  • Results

    Locomotion primarily determined preference for world versus body anchoring: stationary users strongly preferred world anchoring, whereas movement shifted preferences toward body-relative anchors, with no single body anchor dominating.

  • Takeaways & Limitations

    MR systems should prioritize flexible user customization and adaptive anchoring rather than relying on one universal default strategy.

  • Takeaways & Limitations

    The exploratory quantitative findings come from a sample of N=19, and the intertwined interface properties prevent isolating the effect of any single property.

Abstract

from arXiv · show

Anchoring - the choice of frame of reference for mixed reality (MR) interface elements - is a critical design decision involving trade-offs between accessibility, interaction comfort, and visual interference. Despite its importance, user preferences for anchoring across different mobility contexts and interface properties remain poorly understood, as prior work has largely focused on specific tasks or fixed interface configurations. We address this through a mixed-methods user study in which participants configure anchoring strategies across different mobility conditions and interface types. Combining behavioral analysis with structured qualitative inquiry, we analyze how participants select and reason about anchoring modes. Our results show a clear transition from world-anchored interfaces in stationary contexts to body-anchored interfaces during locomotion. However, no single body anchor consistently dominates, highlighting the personal nature of anchoring strategies. Our qualitative analysis reveals the factors users consider in their anchoring decision, including interface accessibility, stability during interaction, visual clutter, and individual mental models. These findings inform the design of adaptive and controllable MR interfaces and highlight the importance of supporting user customization.

1 INTRODUCTION

This study examines how mobility and interaction conditions shape user preferences for mixed-reality UI anchoring. It introduces customizable anchoring across representative scenarios and finds a shift from world anchoring when stationary toward body anchoring during movement.

  • Motivation: MR UI anchoring determines how virtual content is positioned relative to the user, body, and physical environment.The choice affects usability because different frames of reference offer different accessibility, stability, and interference trade-offs.
  • Research gap: Prior work often measured task performance in isolated conditions, leaving preferences across mobility, interaction modalities, and task characteristics poorly understood.The authors frame these experimental factors as potential reasons why prior findings vary or appear contradictory.
  • Approach: The study provides runtime controls for users to select and refine world-, head-, torso-, and arm-anchoring across mobility conditions and interaction scenarios.Participants could change anchoring during use rather than relying on fixed designer-defined choices.
  • Scope: Hand-input interfaces make anchoring especially important because interaction is coupled to hand and body motion rather than a stable controller.The study therefore focuses exclusively on hand input in everyday MR scenarios.
  • Findings: Stationary users unequivocally preferred world anchoring, while increasing mobility shifted preferences toward body anchoring.Semi-stationary use produced more varied choices between world- and body-anchored options.
  • Findings: Anchoring decisions reflected stability, visual clutter, interaction method, interface size, task demands, and individual preferences, with no universal optimal strategy.The findings motivate adaptive mechanisms that remain user-customizable.

2 RELATED WORK

Related work characterizes anchoring as a choice among world-, body-, head-, and object-relative reference frames with task-dependent trade-offs. The paper addresses limited evidence about user preferences during continuous movement and across UI types by emphasizing user-controlled anchoring.

  • Anchoring approaches: MR anchoring research spans environment-, task-object-, body-, and wrist-relative placements, whose effectiveness depends on task, interaction, and reference-frame demands.These approaches support different forms of spatial understanding, accessibility, and contextual interaction.
  • Anchoring approaches: World anchoring can support faster completion, lower cognitive load, and improved target-acquisition precision through environmental stability.Its benefits are tied to maintaining a stable reference frame in the physical environment.
  • Anchoring approaches: Head anchoring offers quick access and persistent visibility but may increase visual interference and input disruption during demanding movement or dual-task use.Prior work associates head-fixed content with walking interference and cognitive load from anchored text.
  • Anchoring approaches: Object anchoring aligns digital information with relevant physical items, supporting contextual awareness, engagement, and precision in professional or safety-critical settings.Its value is particularly associated with task-relevant real-world objects.
  • Open gap: Adaptive anchoring methods have largely relied on system-driven rules without incorporating user preferences during continuous movement or across UI types.This gap motivates studying user choices rather than only performance under fixed conditions.
  • User control: Existing work supports repositioning and runtime adaptation, but surrounding environment and user-specific ergonomic differences can affect the effectiveness of transitions and layouts.The literature increasingly advocates customizable interfaces instead of fixed layouts.
  • Contribution: In contrast to most prior studies using fixed anchoring schemes, this work emphasizes user customization across anchoring choices.The distinction is central to the paper’s contribution.

3 STUDY

The study investigates four MR anchoring modes through an exploratory mixed-methods design centered on user choice. It combines runtime customization with qualitative inquiry into how mobility, interface properties, and personal preferences shape anchoring decisions.

  • Research questions: The study asks how mobility and interface type affect anchoring preferences and which factors users weigh when selecting a strategy.The research questions also examine how these factors interact across contexts.
  • Anchoring modes: The evaluated anchoring modes are world, head, torso, and arm anchoring, selected from prior MR use, hardware feasibility, and prototype experience.Figure 2 illustrates how these modes behave as the user’s pose changes over time.
  • Anchoring modes: World anchoring keeps virtual objects spatially stable relative to tracked physical-environment feature points.It is described as the standard mode for persistent UIs in major MR operating systems.
  • Anchoring modes: Head anchoring maintains a fixed field-of-view position, torso anchoring follows the torso coordinate system, and hand anchoring follows a selected wrist.These modes trade visibility, clutter, placement space, coordination demands, and interaction stability differently.
  • Design scope: Hybrid and movement-dependent anchoring variants were explored but excluded because their behavior was sensitive to implementation details and expanded the design space combinatorially.The pilot observations nevertheless informed the study’s focus on controllable anchoring choices.
  • Customization: The prototype lets users switch anchoring modes on demand and adjust UI position within the selected reference frame.This supports individual differences in preference, ergonomics, and situational context.

3.2 Study design

The study uses repeated within-participants comparisons across stationary, semi-stationary, and moving MR conditions, with three interface tasks experienced in each condition. The tasks vary interaction demands, urgency, size, frequency, and location dependence to represent distinct everyday use cases while preserving experimental control.

  • Study design: Participants completed repeated task trials in stationary, semi-stationary, and moving conditions, with three interface types presented in each trial.The design captures preferences across multiple mobility levels and task contexts within participants.
  • Mobility conditions: Mobility represents the degree of physical movement during interaction, a factor linked to perceptual stability, interaction performance, and anchoring preference.The conditions approximate everyday MR use in contexts such as navigation or messaging while walking.
  • Task design: The three tasks approximate realistic MR interactions while requiring users to coordinate virtual interaction with attention to the physical world.They were designed to retain experimental control across mobility settings.
  • Task design: The interfaces differ in input modality, interaction frequency, urgency, size, and location dependence, but the design cannot attribute observed differences to any single property.These properties are intentionally varied together to represent distinct interface demands.
  • Key task: The Key task combines visual attention and frequent hand input in a large, location-bound interface with low urgency.Participants identify a target key among alternatives and place it in a target box.
  • Visual task: The Visual task requires visual attention without hand input, using a low-frequency, non-location-bound interface for periodic information checks.Participants verbally report the number of red faces on rotating geometric objects when prompted.
  • Controls task: The Controls task uses manual input on a small, non-location-bound interface with low interaction frequency and high urgency.It emulates intermittent actions such as answering calls, controlling playback, or changing notification settings.

3.3 Apparatus

The study used a hand-tracked passthrough MR prototype on a Meta Quest 3, with participants configuring anchoring modes across stationary, semi-stationary, and moving conditions. Sessions included training, counterbalanced experimental blocks, breaks, questionnaires, interviews, and debriefing.

  • Apparatus: The prototype used Unity, Meta Core, Meta Interaction SDK, Meta UI Set, hand tracking, and poke-based input.Passthrough MR and the Movement SDK supported real-world visualization, movement capture, and torso anchoring.
  • Participants: Nineteen participants aged 23–60, with normal or corrected vision and no walking impairments, took part.Participants reported relatively low prior MR experience, averaging 1.53 on a 1–7 scale.
  • Procedure: Each session lasted approximately 90 minutes and began with an overview, consent, demographic questionnaire, and MR-experience measure.The experiment occurred in a quiet, well-lit room with a chair marking the starting point.
  • Procedure: Participants trained on repositioning interfaces and switching anchoring modes before completing three counterbalanced mobility-condition blocks.The stationary, semi-stationary, and moving blocks each contained ten trials, including exploration and experimental trials.
  • Procedure: Only the seven experimental trials per block were analyzed, while participants could change anchoring configurations throughout those trials.The first three trials encouraged trying all four methods and reset interfaces to world anchoring after practice.
  • Procedure: After each mobility condition, participants completed a questionnaire and semi-structured interview, then received a debriefing and compensation.The study was conducted in Portuguese, with responses translated into English before analysis.

3.6 Data collection and Analysis

The study combined structured usability and preference measures with behavioral recording and reflexive thematic analysis of participants’ explanations. Statistical tests were selected for ordinal, repeated-measures data, while participants prioritized comfort and correctness over speed.

  • Questionnaire data: Participants rated each anchoring method for each interface type on a 1–7 UMUX-adapted ease-of-use scale and ranked their preferences.Questionnaires were administered after each mobility condition.
  • Quantitative analysis: Friedman tests compared the four related anchoring methods within each mobility-condition and interface-type combination.The non-parametric repeated-measures approach matched ordinal ratings and the within-participants design.
  • Qualitative analysis: Researchers analyzed 171 interview explanations using reflexive thematic analysis.The interviews asked participants to explain anchor choices for each interface type and their overall impressions.
  • Behavioral measures: Task completion times and accuracy were recorded, but participants were instructed to prioritize comfort and correctness over speed.

4 RESULTS

Anchoring preferences shifted from world anchoring while stationary toward body anchoring during movement, while specific body-anchor choices remained task- and participant-dependent. Participants’ reasoning centered on access, movement context, predictability, visual clutter, and field-of-view management.

  • 4.1 Anchoring Preferences: 94.7–100% of preferred anchor selections were world-anchored when participants were stationary across all interface types.Preferences progressively shifted toward body anchoring as locomotion increased.
  • 4.1 Anchoring Preferences: In semi-stationary conditions, 42.1% still preferred world anchoring for the Key task, reflecting the task’s spatial relationship with real-world elements.
  • 4.1 Anchoring Preferences: Moving conditions produced entirely body-anchored preferences, but no single body anchor dominated across all tasks.Controls favored Arm anchoring at 57.9%, compared with 15.8% for Key and 21.1% for Visual, because proximity reduced reach time and interaction latency.
  • 4.1 Anchoring Preferences: World-anchor usability dominated stationary ratings, mixed world- and body-anchor ratings appeared semi-stationary, and body anchors rated highest while moving.This gradient was measured with 1–7 UMUX-adapted ease-of-use ratings.
  • 4.1 Anchoring Preferences: World-anchor ease ratings differed significantly across mobility conditions for all three interfaces, with χ2 = 28.56-32.38, all p < 10^-6, and Kendall’s W = 0.75-0.85.
  • 4.1 Anchoring Preferences: Moving reduced World medians to 1.0–2.0, while body-anchored references reached medians of 4.0–6.0.In stationary conditions, Head and Torso medians were 4.0–5.0, whereas Arm medians were 2.0–3.0.
  • 4.1 Anchoring Preferences: Semi-stationary usability ratings bridged the stationary and moving extremes, with Torso reaching medians of 7.0 for Visual, 6.0 for Key, and 5.0 for Controls.Arm reached a 7.0 median for Controls but had a wider IQR of 3.5–7.0.
  • 4.2 Anchoring Usage Patterns: Only 6 of 57 experimental instances involved re-adjusting anchoring choices after practice, indicating that participants usually retained their initial configurations.

5 DISCUSSION

Locomotion is the main determinant of anchoring preference: users favor stable world anchoring when stationary but increasingly prefer body-relative anchoring during movement. Interface properties, task demands, and personal preferences shape which body anchor users choose, while early configurations tend to persist.

  • Locomotion: Stationary users overwhelmingly preferred world anchoring, with near-unanimous usage rates of 94.7–100% across tasks.World anchoring was also associated with stability, precision, predictability, and lower cognitive load.
  • Locomotion: As mobility increased, world anchoring became less usable and body-anchored interfaces better preserved accessibility and spatial coherence.Semi-stationary settings showed a mixed design space: 42% still preferred world anchoring for the key interface.
  • Task: During movement, interface type primarily influenced the preferred body anchor, but no single anchor prevailed across all interfaces.Urgent tasks favored immediate availability, whereas frequently used interfaces favored accessible viewing and reachability despite possible clutter.
  • UI Properties: Smaller interfaces were more acceptable when head-anchored, whereas larger interfaces shifted preferences toward torso, arm, or world anchors to reduce visual clutter.For precise hand interaction, world and arm anchoring were perceived as more stable than head or torso anchoring because involuntary motion could introduce jitter.
  • User personalization and implications for design: Users selected different body anchors based on personal preferences, perceived task demands, and comfort rather than following a universal strategy.Although torso often received higher usability ratings during locomotion, participants’ choices remained individualized and context dependent.
  • User personalization and implications for design: Participants generally converged on an anchoring configuration early and rarely changed it when the context stayed the same.This makes first-use onboarding and guided exploration important opportunities for helping users discover effective configurations.
  • Limitations: The exploratory study used 19 participants and could not isolate the effects of intertwined UI properties, limiting strong causal and population-level claims.The selected interfaces also covered only part of the broader MR UI design space.
  • Future Work: Future work could explore adaptive transitions between world- and body-anchored modes, context-aware clutter reduction, personalized placement, and longitudinal changes in preference.These directions address stability, accessibility, attentional load, spatial positioning, and the persistence of early configurations.

6 CONCLUSION

The study found that locomotion primarily determines anchoring preference: users favor stable, predictable world anchoring when stationary and increasingly prefer body-relative anchors as movement increases. Because preferences vary by interface and individual, MR systems should support customization and flexible anchoring, especially during onboarding.

  • 6 CONCLUSION: Participants freely configured anchoring modes across multiple interface types to examine how mobility and UI properties shape preference.The study considered world- and body-relative anchoring and allowed participants to refine their choices.
  • 6 CONCLUSION: Stationary users strongly preferred world-anchored interfaces, while increasing movement shifted preferences toward body-relative anchors.World anchoring was associated with stability and predictability, whereas body-relative anchors better maintained accessibility and spatial coherence during movement.
  • 6 CONCLUSION: No single body anchor dominated across interfaces; users selected anchors according to interface characteristics and personal preference.Once users found a suitable configuration, they rarely changed it during later interactions.
  • 6 CONCLUSION: MR systems should prioritize user customization and flexible anchoring rather than relying on one default strategy.The conclusion also identifies onboarding as especially important because initial anchoring choices tend to persist.

A STUDY DETAILS

The study collected anchoring preferences, ease-of-use ratings, correction effort, and interview explanations for three interfaces across stationary, semi-stationary, and moving conditions. Participants evaluated world, head, torso, and arm anchoring modes in a room-based task setup.

  • Questionnaires and interviews: After each mobility condition, participants ranked World, Head, Torso, and Arm anchoring for the Controls interface and explained their decisions.They also reported ease of use and time spent correcting the interface on 1–7 scales.
  • Questionnaires and interviews: The same ranking, explanation, ease-of-use, and correction-effort measures were collected for the Key instruction and Visual interfaces.Interview prompts also asked whether preferences changed after planning and what could improve anchoring or positioning.
  • Experimental setting: The experiment used stationary, semi-stationary, and moving conditions in a 6x4.5-meter room, with seated tasks, two working zones, or movement throughout the room.The interface panels ranged from approximately 6x7 cm to 24x15 cm.

B FULL STATISTICAL RESULTS

The appendix reports repeated-measures inferential analyses of ease-of-use ratings from the same 19 participants, using omnibus Friedman tests and corrected paired comparisons when appropriate.

  • Analysis procedure: All ease-of-use analyses used repeated measures from the same 19 participants.The analyses therefore compared anchoring conditions within participants.
  • Analysis procedure: Friedman tests assessed omnibus differences in ease-of-use ratings across anchoring conditions.Kendall’s W was reported for omnibus effects.
  • Analysis procedure: Significant omnibus results were followed by paired Wilcoxon signed-rank tests with Holm-corrected pairwise p-values.Absolute rank-biserial correlation was reported for pairwise effects.

B.1 Comparisons Among Anchoring Methods

Across mobility conditions and interface types, users rated different anchoring methods differently, with significant contrasts reported for several moving-condition tasks. The strongest pattern favored body anchors over world anchoring during moving visual and controls tasks.

  • Moving key trials showed higher ratings for torso than arm anchoring.The Holm-corrected comparison reported pHolm = .013 and |rrb| = 0.77.
  • Moving visual trials favored head, torso, and arm over world anchoring in significant contrasts.Torso exceeded world with pHolm = 7.42×10−4 and |rrb| = 1.00; head and arm also rated higher than world.
  • Moving visual trials also favored torso over head and torso over arm in significant pairwise comparisons.The torso-over-head contrast had pHolm = .031 and |rrb| = 0.64; torso-over-arm had pHolm = .022 and |rrb| = 0.74.
  • Moving controls trials favored each body anchor over world anchoring, with no significant differences among head, torso, and arm.World was rated lower than head, torso, and arm, while all three body-anchor pairings were non-significant.

B.2 Comparisons Among Mobility Conditions

Mobility condition affected ratings differently depending on interface type and anchoring method. World anchoring was rated higher when stationary for controls, while several body anchors received higher ratings in more mobile conditions.

  • Visual trials with head or arm anchoring showed no significant mobility-condition differences.The reported pairwise comparisons for head and arm anchoring were non-significant across the mobility conditions.
  • Visual trials with torso anchoring were rated higher in semi-stationary and moving conditions than when stationary.Both significant comparisons had pHolm = .002, with |rrb| = 1.00 for semi-stationary versus stationary and |rrb| = 0.93 for moving versus stationary.
  • Controls with world anchoring were rated higher when stationary than in semi-stationary and moving conditions.The stationary-versus-semi-stationary comparison had pHolm = 3.56×10−4 and |rrb| = 1.00; stationary-versus-moving had pHolm = 3.21×10−4 and |rrb| = 1.00.
  • Controls with head anchoring were rated higher when moving than when stationary, while the semi-stationary comparisons were non-significant.Moving exceeded stationary with pHolm = .017 and |rrb| = 0.81.
  • Controls with torso anchoring were rated higher in semi-stationary and moving conditions than when stationary.Both significant contrasts had pHolm = .033, with |rrb| = 0.69 for semi-stationary versus stationary and |rrb| = 0.70 for moving versus stationary.
  • Controls with arm anchoring were rated higher in semi-stationary and moving conditions than when stationary.The significant effects were pHolm = .015 and |rrb| = 0.71 for semi-stationary versus stationary, and pHolm = .002 and |rrb| = 0.93 for moving versus stationary.
Loading 2608.28064v1…