Asleep at the Virtual Wheel

Hello. My name is Craig Anderton, I’m a PhD candidate at Birmingham City University. I’m here today presenting on behalf of my co-authors Professor Chris Creed, Doctor Sayan Sarcar, and Doctor Arthur Theil. Our paper is titled Asleep at the Virtual Wheel: The Increasing Inaccessibility of Virtual Reality Applications.

I’d like to start today by briefly presenting the agenda for the presentation. I’ll begin by exploring the background for the project, I’ll then discuss the selection of applications we looked at, an overview of our results, and a discussion of feature support and overall trends. I’ll then end today by presenting the key takeaways and future directions suggested by our research.

An important motivation for our research was to provide a source of continued academic engagement with the mainstream VR industry. We wanted to build an understanding of the decisions made by practitioners, in turn acting as a further reinforcement of the call to action in the development of inclusive VR titles. There is a great emphasis placed on industry practitioners in the implementation of software-level accessible design choices. However, interviews indicate that a lack of guideline knowledge within the industry has led to practitioners being largely unaware of best practices.

Mitigation of sickness has to date received the largest amount of accessibility research focus, whilst the lack of standardisation for defining how to implement non-visual feedback has led to the feeling of exclusion for users with sensory impairments. Our research aims to enhance the practical relevance of future academic accessibility research, in particular in the underexplored general VR accessibility field.

To allow for trend analysis, the 330 titles most used by VR end users since the release of the first wave of sensor-based consumer-level HMDs in 2016 were selected from all major VR storefronts. Features were derived from the systematic analysis of emerging VR-specific guidelines and takeaways from accessibility focused VR research findings.

Whilst OS-level settings can allow for alteration of a selection of the accessibility features analysed in this review, for example mono audio balance control, OS-level, platform-specific, and utility software functionalities may not be available to all users accessing the same application on an alternative HMD device, OS, or software platform.

All applications were launched by the first author to systematically understand how accessibility features were presented in each title. The picture to the right highlights a selection of runtime accessibility features that were noted in the 'Arizona Sunshine' game. These include white subtitles, a visual focus indicator white circle that appears when an object is within range to be interacted with, and a visual representation of the bimanual right hand controller highlighting button locations.

The overview of our results further highlights a selection of the key accessibility features we explored in the 330 titles. Interaction findings include 148 applications allowing for dominant hand selection, whilst more fine-grained alterations over interactions, such as input customisation and adjustments to input sensitivity, with for example smoothing of aiming inputs, is generally less widely supported. Temporal interaction analysis further highlights how remapping appears largely unavailable in applications released each year, with no clear growth trends in allowing users to customise interaction demands and minimal support for extensive remapping.

For inputs, bimanual handheld controllers are supported in all but 3 freehand-only titles, whilst low-friction non-hardware-based input, such as freehand, eye, or speech-based input, is rarely available. Along with the minimal non-hardware supported input, input hardware trends highlight how the number of applications which contain potentially accessible switch-based gamepad and keyboard interaction support has dramatically decreased since 2016, with users increasingly given no other option than to engage solely with bimanual tracked controllers.

In terms of physical positioning, seated access is generally well supported overall in 251 titles, with many applications either adapting to seated postures or providing options to choose between standing and seated access. Software-level lying down posture support however is available in just 4 titles, supporting the belief of bed-bound users that their access needs are not considered. Although seated access is generally well supported, temporal trends further highlight a slight but concerning trend in the number of applications which provide primary seated posture support, with these physical positioning results taken together suggesting that physical flexibility may be decreasing.

User interface results meanwhile show how users are given little control over resizing of text or UI elements, with more detailed analysis revealing that size adjustments at a fixed depth in particular are very rare.

Next, for output modalities, results show how haptic feedback is found in a large number of titles, although further analysis reveals that haptics associated with locomotion may be underexplored in comparison to interactions. Binaural volume and sound controls are also well supported in over 200 titles, whilst adjustments to colour and brightness are very rarely available, both of which are potentially vital considerations for users who are colourblind or have low vision.

Finally, subtitles and closed captions are included in 101 titles, whilst only 10 applications contain control over personal space, such as allowing for boundary bubbles, highlighting how harassment concerns are often not considered. Sign language meanwhile is not included in the graph as no title contains full signing support. Temporal trends further highlight wider communication concerns in VR, with the low level of closed captions and subtitles each year supporting suggestions that users with additional social and communication requirements are largely unaccounted for.

Overall, our results show numerous failures to accommodate the needs of users with disabilities across all categories. Although VR accessibility in academia has been extensively researched, with findings appearing in guidelines such as those from the W3C, our results show that this research is largely not being translated into practice. XR guidelines and platform-specific recommendations overall appear largely unsupported in mainstream titles. Whilst a selection of our key headline findings highlight a number of worrying accessibility trends, particularly in terms of interaction and physical positioning flexibility.

Next, I want to go into a bit of depth highlighting the key takeaways from our review. Firstly, for practitioners, prioritise inclusive input support by ensuring that your titles support a variety of input types. Enhance physical positioning flexibility by incorporating seated and lying down postures. Provide UI and output customisation options to allow for customisation to sensory-specific requirements. Support communication needs such as providing sign language support. And finally, adhere to accessibility guidelines and implement platform-specific recommendations early in the design process.

For researchers, explore alternative inputs and how to incorporate assistive technologies within the spatial interface. Investigate the impact of limited output customisation, particularly with users who have sensory impairments. Assess broader accessibility trends in collaboration with industry practitioners. Focus on locomotion from diverse perspectives in order to further understand the potential trade-offs in locomotion techniques. And finally, evaluate the role of guidelines and further establish new or refined recommendations.

In terms of cross collaboration efforts between academia and industry, joint research projects should be further explored to bridge the gap between research and practice. Resource sharing is also vital, with open-source toolkits and open access guidelines allowing for more efficient feature implementation. Further establish workshops and working groups to foster a community-driven approach to the creation of VR experiences that brings together researchers, practitioners, and users with disabilities at all stages of research and development. And finally, promote accessibility education by collaborating across educational institutions and industry bodies to integrate accessibility teaching and training into curricula and professional development programs.

I want to end today by highlighting future directions our work suggests. For example, future accessibility audits are required to understand the effectiveness of individual implementations. Whilst further accessibility analyses must be conducted by engaging with a wide range of users with disabilities to understand the perspectives of diverse audiences. Additional categorisation features influenced by emerging VR guidelines may be beneficial to record in future analyses to further understand the evolving state of accessibility in VR. Whilst future research is required in underexplored areas such as multi-modal and customisable inputs, the presentation and control of spatial UI elements, and alternative output modalities beyond visuals.

Thank you very much for listening. Please send any questions or queries you have to the provided email addresses, and please read our paper in the proceedings for more in-depth findings and takeaways.