CoChat
Sign Up
Cited by 75Digital Accessibility for DisabilitiesUsability and User Interface DesignAssistive Technology in Communication and Mobility

Reading Between the Guidelines: How Commercial Voice Assistant Guidelines Hinder Accessibility for Blind Users

Open access
Branham, Stacy M., Mukkath Roy, Antony Rishin·2019·Proceedings of the 21st International ACM SIGACCESS Conference on Computers and Accessibility
DOI 10.1145/3308561.3353797
Open this paper in CoChat — free accountOpen PDF

Abstract

Voice-Activated Personal Assistants (VAPAs)-like Apple Siri and Amazon Alexa - have rapidly become common features on mobile devices and in homes of millions of people around the world. They have proven to be particularly valuable to people with disabilities, chiefly among people with visual impairments. Yet, we still know relatively little about the fundamental metaphors and guidelines for designing voice assistants, and how they might empower and constrain visually impaired users. To address this need, we conducted a qualitative document review of VAPA design guidelines published by top commercial vendors Amazon, Google, Microsoft, Apple and Alibaba. We found that guidelines have many commonalities that surface an underlying assumption that VAPA interfaces should be modeled after human-human conversation. We draw on prior work about needs of people with visual impairments to critique this taken-for-granted human-human conversation metaphor and offer amendments to prevailing design guidelines that can make this now-pervasive platform more fully achieve its potential to become universally usable.

Integrity signals

Registry

Crossref · 10.1145/3308561.3353797

Cites

38 of 38 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

3 of 39 cited sources checked · 2 supported · 0 flagged · accuracy not yet measured

  • Premisecitation

    Many blind people can easily comprehend speech at higher listening rates than sighted people.

    “Studies conducted by various researchers have shown that many blind people can easily comprehend speech at higher listening rates than sighted people [6,12].”
    Quote in the paperSubstantial claim1 of 2 sources checked
    • cites [6] 6. Danielle Bragg, Cynthia Bennett, Katharina Reinecke, and Richard Ladner. 2018. A Large Inclusive Study of Human Listening Rates. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems (CHI '18), 444:1-444:12. https://doi.org/10.1145/3173574.3174018 source
    • cites [12] 12. Susanne Dietrich, Ingo Hertrich, and Hermann Ackermann. 2013. Ultra-fast speech comprehension in blind subjects engages primary visual cortex, fusiform gyrus, and pulvinar -a functional magnetic resonance imaging (fMRI) study. BMC Neuroscience 14: 74. https://doi.org/10.1186/1471-2202-14-74 source
  • Premisecitation

    Speech recognition often breaks down for children and older adults as well as people who are deaf or hard-of-hearing, because there is wide variation in pitch, pronunciation, patterns of stress, and intersyllabic pauses, leading to higher recognition error rates.

    “Prior work finds that speech recognition often breaks down for children and older adults [3,21] as well as people who are deaf or hard-of-hearing [16], because there is wide variation in pitch, pronunciation, patterns of stress, and intersyllabic pauses, leading to higher recognition error rates [16,21,35].”
    Quote in the paperNot a substantial claim1 of 4 sources checked
    • cites [3] 3. Saminda Sundeepa Balasuriya, Laurianne Sitbon, Andrew A. Bayor, Maria Hoogstrate, and Margot Brereton. 2018. Use of Voice Activated Interfaces by People with Intellectual Disability. In Proceedings of the 30th Australian Conference on Computer-Human Paper Session 8: Listen to this! ASSETS '19, October 28-30, 2019, Pittsburgh, PA, USA Interaction (OzCHI '18), 102-112. https://doi.org/10.1145/3292147.3292161 source
    • cites [21] 21. Silvia Lovato and Anne Marie Piper. 2015. "Siri, is this you?": Understanding young children's interactions with voice input systems. In Proceedings of the 14th International Conference on Interaction Design and Children (IDC '15). ACM, New York, NY, USA, 335- 338. http://dx.doi.org/10.1145/2771839.2771910 source
    • cites [16] 16. Abraham T. Glasser, Kesavan R. Kushalnagar, and Raja S. Kushalnagar. 2017. Feasibility of Using Automatic Speech Recognition with Voices of Deaf and Hard-of- Hearing Individuals. In Proceedings of the 19th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS '17). ACM, New York, NY, USA, 373-374. https://doi.org/10.1145/3132525.3134819 source

      Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

      Experimental — accuracy not yet measured
    • cites [35] 35. S. Schlögl, G. Chollet, M. Garschall, M. Tscheligi, and G. Legouverneur. 2013. Exploring Voice User Interfaces for Seniors. In Proceedings of the 6th International Conference on Pervasive Technologies Related to Assistive Environments (PETRA '13), 52:1-52:2. https://doi.org/10.1145/2504335.2504391 source
  • Findingnone given

    Public-facing VAPA design guidelines exist from Amazon, Google, Microsoft, Apple, and Alibaba.

    “commercial VAPA vendors--led by Amazon, Google, Apple, Microsoft, and Alibaba--have been publishing their own public guidelines for years.”
    Quote in the paperSubstantial claim
  • Findingcitation

    Design guidelines encourage behaviors that are often exhibited in human-human conversations to develop lasting relationships (13 instances).

    “Design guidelines encourage behaviors that are often exhibited in human-human conversations to develop lasting relationships (13 instances, Go/Am).”
    Quote in the paperSubstantial claim
  • Findingown data

    Of significant concern across all guidelines was that developers strive to make conversations efficient and easy for users to perform (68 instances).

    “Of significant concern across all guidelines was that developers strive to make conversations efficient and easy for users to perform (68 instances, Go/Am/Mi/Ap/Al).”
    Quote in the paperSubstantial claim
  • May combine two claims. 24
  • Classified as probably not a substantial claim. 66
  • The quote only partly states this claim. 3

  • May combine two claims.

    Premisecitation

    Blind people can comprehend speech at higher listening rates, and blind VAPA users find speech rates frustratingly slow.

    “However, prior work shows that blind people can comprehend speech at higher listening rates [6,12], and that blind VAPA users find speech rates frustratingly slow [1].”
    Quote in the paperSubstantial claim1 of 3 sources checked
    • cites [6] 6. Danielle Bragg, Cynthia Bennett, Katharina Reinecke, and Richard Ladner. 2018. A Large Inclusive Study of Human Listening Rates. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems (CHI '18), 444:1-444:12. https://doi.org/10.1145/3173574.3174018 source
    • cites [12] 12. Susanne Dietrich, Ingo Hertrich, and Hermann Ackermann. 2013. Ultra-fast speech comprehension in blind subjects engages primary visual cortex, fusiform gyrus, and pulvinar -a functional magnetic resonance imaging (fMRI) study. BMC Neuroscience 14: 74. https://doi.org/10.1186/1471-2202-14-74 source
    • cites [1] 1. Ali Abdolrahmani, Ravi Kuber, and Stacy M. Branham. 2018. "Siri Talks at You": An Empirical Investigation of Voice-Activated Personal Assistant (VAPA) Usage by Individuals Who Are Blind. In Proceedings of the 20th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS '18), 249-258. https://doi.org/10.1145/3234695.3236344 source

  • Classified as probably not a substantial claim.

    Findingcitation

    Guidelines advise that the VAPA should never blame the user or a third party.

    “Perhaps the clearest guidance, reiterated in multiple instances, is that the VAPA should never blame the user or a third party:”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Lists are often used to constrain and therefore expedite the user's selection of a response.

    “In combination with questions, lists are often used to constrain and therefore expedite the user's selection of a response:”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Guidelines encourage colloquial language and favor natural-sounding conversation over grammatical accuracy.

    “Use of colloquial language is encouraged, and natural-sounding conversation is favored over grammatical accuracy.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Developers should create a description of the character, including a list of adjectives and personal traits, to help the persona mirror real human agents.

    “As part of this process, developers should create a description of the character, including a list of adjectives and personal traits, to help the persona mirror real human agents.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Findingnone given

    Google and Microsoft guidelines suggest that a good voice interaction should use a variety of confirmation techniques (10 instances).

    “Both Google and Microsoft suggest that a good voice interaction should use a variety of techniques for confirmation (10 instances).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    Google's guidelines note that too much content is as concerning as too little.

    “Google's guidelines note that too much content is as concerning as too little.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Findingreasoning

    Guidelines advocate against deceiving users through misrepresenting or hiding information using defaults.

    “they advocate against deceiving users through misrepresenting or hiding information using defaults.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based).

    Findingreasoning

    In a partial intent case, the VAPA should be able to detect the missing element and provide an immediate follow-up prompt towards fulfilling the service request.

    “In this case, the VAPA should be able to detect the missing element and provide an immediate follow-up prompt towards fulfilling the service request.”
    Quote in the paperNot a substantial claim

  • The quote only partly states this claim.

    Findingown data

    The authors identified design guideline documents for 5 of 7 vendors: Google, Amazon, Microsoft, Apple, and Alibaba.

    “Using this method, we were able to successfully identify design guideline documents for 5 of the 7 vendors: Google 2 (Go), Amazon 3 (Am), Microsoft (Mi) 4”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    Making conversation personal is about designing VAPAs so that they exercise agency and tailor interactions to the user (50 instances).

    “Making conversation personal is about designing VAPAs so that they exercise agency and tailor interactions to the user (50 instances, Go/Am/Mi/Ap/Al).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    Guidelines recommend maintaining variation in the VAPA's responses (7 instances).

    “In addition to supporting varied request phrasing, guidelines recommend maintaining variation in the VAPA's responses (7 instances).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    One guideline recommended creating a human-human dialog script and testing it by conducting roleplay with real humans.

    “For example, one guideline recommended creating a human-human dialog script and testing it by conducting roleplay with real humans.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Guidelines advise that the system should request confirmation if exiting would result in significant loss of progress.

    “One caveat to this advice is that the system should request confirmation if exiting would result in significant loss of progress.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Guidelines rationalize list length by claiming the limited capacity of the human brain to remember instructional information.

    “The guidelines further rationalize this choice by claiming the limited capacity of the human brain to remember instructional information:”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Findingreasoning

    Context can extend beyond the bounds of the current invocation.

    “Context can extend beyond the bounds of the current invocation.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    Guidelines offer various techniques to improve efficiency while reading lists (13 instances).

    “Guidelines offer various techniques to improve efficiency while reading lists (13 instances).”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Findingnone given

    Guidelines state that the system should be able to identify when a user's tone of voice reveals frustration or impatience, and then change to a more concise interaction mode.

    “These guidelines also state that the system should be able to identify when a user's tone of voice reveals frustration or impatience, and then change to a more concise interaction mode.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    A question placed at the end of the utterance prompts the user to begin their turn while also discouraging interruption.

    “In the guideline above, we see that a question placed at the end of the utterance both prompts the user to begin their turn, while also discouraging interruption.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The guidelines articulate two types of confirmation: implicit and explicit.

    “They articulate two types of confirmation--implicit and explicit.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    Maintaining efficient conversation entails clean information hand-offs between the VAPA and user, in which each takes a turn without either interrupting or pausing between utterances (12 instances).

    “Maintaining efficient conversation entails clean information hand-offs between the VAPA and user, in which each takes a turn without either interrupting or pausing between utterances (12 instances).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Most platforms provide a variety of synthesized text-to-speech voice libraries to choose from.

    “To achieve this, most of the platforms provide a variety of synthesized text to speech (TTS) voice libraries to choose from.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    Microsoft and Amazon guidelines both encourage keeping conversations friendly (3 instances).

    “Microsoft and Amazon guidelines both encourage keeping conversations friendly (3 instances).”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Findingcitation

    Guidelines state that VAPAs should be able to handle even subtle variations and mispronunciations in the user's response.

    “The guidelines also state that VAPAs should be able to handle even subtle variations and mispronunciations in the user's response.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Findingnone given

    VAPAs often format responses as lists.

    “VAPAs often format responses as lists--lists of top items for purchase, lists of nearby businesses, lists of search results, lists of music stations, and more.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingnone given

    Microsoft, Amazon, and Alibaba all remark on the number of list items to present, generally around three items, although Amazon suggests a range of 2-5.

    “Microsoft, Amazon, and Alibaba all remark on the number of list items to present, generally around three items, although Amazon suggests a range of 2-5.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Findingcitation

    Guidelines caution against overusing earcons because they can impose cognitive load and become overwhelming, limiting use to a few easily distinguishable sounds.

    “They therefore caution against overusing earcons because they can”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingnone given

    The duration of pauses between lists of items should be based on item length and complexity; there should be a 350ms pause for a normal scenario and a 400ms pause for a complex scenario.

    “Furthermore, the duration of pauses between lists of items should be based on item length and complexity.For example, there should be a 350ms pause for a normal scenario and a 400ms pause for a complex scenario.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    There is an option to record and use a real professional human voice.

    “There is also an option to record and use a real professional human voice.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingreasoning

    Guidelines suggest third-party app developers should create their own personas rather than impersonating or mimicking the system persona.

    “third-party app developers should create their own personas rather than impersonating or mimicking the system persona.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Guidelines suggest prompts given to users should be carefully constructed so that possible answers are predictable and unambiguous.

    “Another is to carefully construct the prompts given to users, so that possible answers are predictable and unambiguous:”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Findingcitation

    Guidelines suggest the illusion of user control can be achieved by avoiding teaching commands or giving a tutorial-like introduction to a VAPA's capabilities.

    “One is to avoid teaching commands or giving a tutorial-like introduction to a VAPA's capabilities:”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Findingnone given

    The VAPA should be able to identify all the key attributes from the utterance and respond meaningfully to the user request.

    “The VAPA should be able to identify all the key attributes from the utterance and respond meaningfully to the user request.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Developers are cautioned by one guideline to refrain from asking users a question the system is not prepared to handle the answer to.

    “Developers are cautioned by one guideline to refrain from asking users a question the system is not prepared to handle the answer to.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingreasoning

    Conversational interfaces are ideal for enabling users to multitask, as they support hands-free, eyes-free interaction with low cognitive load (7 instances).

    “Conversational interfaces are ideal for enabling users to multitask, as they support hand-free, eyes-free interaction with low cognitive load (7 instances).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    Guidelines identify voice as a safe and efficient modality for navigating--walking or driving--without distraction.

    “Similarly, guidelines identify voice as a safe and efficient modality for navigating--walking or driving--without distraction.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims. The quote only partly states this claim.

    Findingcitation

    Guidelines suggest defining a range of utterances people may say to the VAPA, with a standard benchmark around 30 alternate utterances per intent.

    “They suggest a standard benchmark to be around 30 alternate utterances per intent.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims. The quote only partly states this claim.

    Findingown data

    The axial codes and instance counts were: Make conversations human 39; Make conversations personal 50; Make conversations efficient 68; Make conversations relational 13; Give user a sense of control 20.

    “Make conversations human 39”
    • Also stated in: § Make conversations human 39 · plain
      “Make conversations personal 50 Make conversations efficient 68”
    • Also stated in: § Make conversations relational 13 · plain
      “Make conversations relational 13”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Guidelines identify strategies for personas, such as developing personas based on familiar roles enacted by real-life service providers.

    “To achieve this, guidelines identify various strategies.For example, personas can be developed based on familiar roles enacted by real-life service providers:”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Findingnone given

    Microsoft recommends that the VAPA should not take more than 20 seconds to read the first few items in the list.

    “The VAPA should not take more than 20 seconds to read the first few items in the list.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    Guidelines advocate these personalization qualities as beneficial traits of human conversational partners.

    “Guidelines advocate for these qualities as beneficial traits of human conversational partners.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingown data

    The thematic analysis resulted in 190 instances, 18 open codes, and five axial codes.

    “This process resulted in 190 instances, 18 open codes, and five axial codes.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Findingnone given

    Guidelines advise always allowing the user to make course corrections or discontinue a task immediately when the user gives commands like exit, cancel, stop, nevermind, and goodbye.

    “Finally, always allow the user to make course corrections (”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Guidelines state that VAPAs should be able to handle requests not only of various phrasing, but also of various completeness (12 instances).

    “Guidelines state that VAPAs should be able to handle requests not only of various phrasing, but also of various completeness.Completeness can range from no intent to multiple intents at once (12 instances).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingreasoning

    What constitutes a relevant default value varies according to the context of use.

    “What constitutes a relevant default value varies according to the context of use.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingreasoning

    Guidelines recommend using explicit confirmation when the cost of misunderstanding is high, as when booking a flight or making a financial transaction.

    “They also recommend using explicit confirmation when the cost of misunderstanding is high, as when booking a flight or making a financial transaction.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    Guidelines from Amazon and Google suggest that VAPAs should maintain transparency in conversation with users when handling errors and take responsibility when necessary (10 instances).

    “Guidelines from Amazon and Google suggest that VAPAs should maintain transparency in conversation with users when handling errors, and take responsibility when necessary (10 instances).”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingcitation

    Guidelines warn that context changes even when it appears otherwise, and it is particularly important to recognize this in error scenarios to reduce frustration.

    “Guidelines warn that context changes even when it appears otherwise; it is particularly important to recognize this in error scenarios to reduce frustration:”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    A smart default may be derived from previous interactions.

    “A smart default may also be derived from previous interactions”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The most common recommendation among guidelines to achieve variety is by adding variety in prompts that repeatedly occur during the conversation.

    “The most common recommendation among guidelines to achieve variety is by adding variety in prompts that repeatedly occur during the conversation.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Amazon introduces the concept of a one-breath test to evaluate the length of conversation.

    “Amazon introduces the concept of a 'one-breath test' (can the utterance be stated in one breath?) to evaluate the length of conversation.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Guidelines commonly recommend strategic placement of questions in the VAPA's script.

    “To scaffold efficient turn-taking, guidelines commonly recommend strategic placement of questions in the VAPA's script:”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    There are multiple segments of the guidelines regarding the importance of leveraging context to communicate in a more cooperative manner with individuals (12 instances).

    “There are multiple segments of the guidelines regarding the importance of leveraging context to communicate in a more cooperative manner with individuals (12 instances).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Voice can be further customized with Speech Synthesis Markup Language (SSML) to achieve more personalization.

    “This voice can be further customized with Speech Synthesis Markup Language (SSML) to achieve more personalization.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingreasoning

    Amazon suggests that items be clustered into groups of two or three.

    “Amazon suggests that items be clustered into groups of two or three”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Findingnone given

    When the VAPA asks for more information to complete a no intent or partial intent request, if the user does not provide any response, the VAPA should re-prompt with a slight rewording.

    “The VAPA should handle such situations by re-prompting with a slight rewording.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

    Findingnone given

    As the user engages a specific app or skill over time, a greeting should become less verbose and more familiar.

    “Additionally, as the user engages a specific app or skill over time, a greeting should become less verbose and more familiar.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Findingnone given

    In a no intent case, VAPAs should not simply dismiss the user's request or ask them to restate the request correctly; they should walk users through the interaction by providing a list of three options from which to choose.

    “In this case, VAPAs should not simply dismiss the user's request or ask them to restate the request correctly; they should walk users through the interaction by providing a list of three options from which to choose.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Findingreasoning

    The guidelines suggest that earcons are a mismatch for conversational interfaces because they are not intuitive.

    “The guidelines suggest that earcons are a mismatch for conversational interfaces, because they are not intuitive.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Guidelines suggest that personality traits should be kept consistent within each persona.

    “Guidelines suggest that personality traits should be kept consistent within each persona;”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Guidelines state that it is the developer's responsibility to build the right balance between implicit and explicit confirmation.

    “Guidelines state that it is the responsibility of the developer to build in the right balance between both implicit and explicit confirmation.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Findingnone given

    Guidelines recommend that VAPAs be made to speak using everyday language, complete with slang, contractions, naturalistic pauses, and turn-by-turn cooperative dialogue.

    “Guidelines recommend that VAPAs be made to speak using everyday language, complete with slang, contractions, naturalistic pauses, and turn-by-turn cooperative dialogue.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Findingnone given

    VAPA guidelines recommend being able to handle complex commands from the user, which include all necessary information for a service request in a single utterance.

    “Second, regarding complexity, VAPA guidelines recommend being able to handle complex commands from the user, which include all necessary information for a service request in a single utterance.”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Findingreasoning

    Guidelines justify their length limits by arguing that longer responses will be difficult to follow, overwhelm the human brain, and result in frustration and impatience for users.

    “Guidelines justify these limits by arguing that longer responses will be difficult to follow, as this will overwhelm the human brain, and it may result in”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S4 (based). May combine two claims.

    Findingnone given

    The first few list items should be spoken for a maximum of 20 seconds, with pause durations ranging from 350ms to 400ms.

    “Our findings revealed that the first few list items should be spoken for a maximum of 20 seconds, with pause durations ranging from 350ms to 400ms.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim. May combine two claims.

    Findingreasoning

    Guidelines emphasize using natural and intuitive speech, avoiding robotic structures, and minimizing earcons, because these may lead to cognitive overload.

    “Guidelines additionally emphasize using”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Findingnone given

    Guidelines recommend concrete limits to VAPA responses, including a maximum of two keywords per turn, three items per list, and the one-breath test.

    “For example, various guidelines recommend a maximum of two keywords per turn, three items per list, and the”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    The voice-only nature of VAPAs is crucial for blind users; while sighted users may switch to a visual interface for in-depth tasks or error correction, blind users do not always have accessible alternatives.

    “The voice-only nature of VAPAs is crucial for blind users; while sighted users may switch to a visual interface for in-depth tasks or error correction, blind users do not always have accessible alternatives [29].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [29] 29. Alisha Pradhan, Kanika Mehta, and Leah Findlater. 2018. "Accessibility Came by Accident": Use of Voice- Controlled Intelligent Personal Assistants by People with Disabilities. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems (CHI '18), 459:1-459:13. https://doi.org/10.1145/3173574.3174033 source

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    The ability to define voice macros will be particularly useful to blind users with regards to increasing complexity or reducing length.

    “We believe the ability to define these”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    While sighted people can comprehend speech at a maximum of 10 syllables per second, some blind people can comprehend up to 25 syllables per second.

    “While sighted people can comprehend speech at a maximum of 10 syllables per second, some blind people can comprehend up to 25 syllables per second [14].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [14] 14. R. Douglas Fields. Why Can Some Blind People Process Speech Far Faster Than Sighted Persons? Scientific American. Retrieved April 30, 2019 from https://www.scientificamerican.com/article/why-can- some-blind-people-process/

      Couldn’t be matched to a paper in Crossref, so it can’t be checked.

  • May combine two claims.

    Interpretationreasoning

    These guidelines conflict significantly with the voice output design of screen readers for people who are blind, which require expert interaction techniques, use robotic-sounding synthesized voice, and speak output at a rate far faster than a human could.

    “Interestingly, these guidelines conflict significantly with the voice output design of screen readers for people who are blind, which require expert interaction techniques, use robotic-sounding synthesized voice, and speak output at a rate far faster than a human could.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Like people who are blind, other groups such as older adults, children, people with Alzheimer's, and people with intellectual disabilities would benefit from VAPAs that adopt more inclusive models of conversation length, complexity, speed, and other aspects of interaction.

    “Like people who are blind, these groups would benefit from VAPAs that adopt more inclusive models of conversation length, complexity, speed, and other aspects of interaction.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationreasoning

    Blind people may have sufficient motivation to learn and leverage longer, more complex commands and to deploy them at top speed.

    “it stands to reason that they may have sufficient motivation to learn and leverage longer, more complex commands and to deploy them at top speed.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    Many blind people are remarkably superior to sighted peers in performing serial memory tasks and are better able to recall longer word sequences.

    “many blind people are remarkably superior to sighted peers in performing serial memory tasks and are better able to recall”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [33] 33. Noa Raz, Ella Striem, Golan Pundak, Tanya Orlov, and Ehud Zohary. 2007. Superior serial memory in the blind: a case of cognitive compensatory adjustment. Current Biology: CB 17, 13: 1129-1133. https://doi.org/10.1016/j.cub.2007.05.060 source

  • May combine two claims.

    Interpretationcitation

    Longer turns may be possible for blind users, while previous studies suggest that more succinct turns may be desirable for blind users.

    “Above, we draw out the ways in which longer turns may be possible for blind users, while previous studies suggest that more succinct turns may be desirable for blind users [1].”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [1] 1. Ali Abdolrahmani, Ravi Kuber, and Stacy M. Branham. 2018. "Siri Talks at You": An Empirical Investigation of Voice-Activated Personal Assistant (VAPA) Usage by Individuals Who Are Blind. In Proceedings of the 20th International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS '18), 249-258. https://doi.org/10.1145/3234695.3236344 source

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Premisecitation

    Murad et al. published early work identifying design guidelines for hands-free speech interaction, targeting adoption within the HCI research community.

    “Murad et al. [25] published early work identifying design guidelines for hands-free speech interaction, targeting adoption within the HCI research community.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [25] 25. Christine Murad, Cosmin Munteanu, Leigh Clark, and Benjamin R. Cowan. 2018. Design Guidelines for Hands-free Speech Interaction. In Proceedings of the 20th International Conference on Human-Computer Interaction with Mobile Devices and Services Adjunct (MobileHCI '18), 269-276. https://doi.org/10.1145/3236112.3236149 source

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    The three proposed recommendations have not been user tested.

    “It is important to note that these have not been user tested.”
    Quote in the paperNot a substantial claim
77 of 77

  • The paper relies on this without support.

    Understanding a user's utterance and making a VAPA's response relevant require context from prior utterances and conversations.

    “Due to the ambiguous nature of language, a user's utterance cannot be understood in isolation, i.e. without the context of prior utterances and conversations.Neither can the VAPA's response be made relevant.”

  • The paper relies on this without support.

    Commercial VAPA guidelines are a unique window into how developed systems operate and may allow pinpointing fundamental sources of documented accessibility and usability issues.

    “Commercial VAPA guidelines stand to offer a unique window into how the developed systems operate and may allow us to pinpoint fundamental sources of documented accessibility and usability issues [1,29].”

  • The paper relies on this without support.

    The five selected vendors' guidelines are representative of commercial VAPA design guidelines.

    “We chose to search for guidelines produced by recognized industry leaders on mobile voice, as well as smart speaker platforms.”

  • The paper relies on this without support.

    The guidelines collected in September 2018 reflect current commercial VAPA design guidelines.

    “Given the dynamic nature of online content, we collected static copies of the guidelines for offline viewing.”

  • The paper relies on this without support.

    The number of coding instances reflects the prevalence or importance of a theme in the guidelines.

    “When reporting open codes, we will identify the number of instances in parentheses.”

  • The paper relies on this without support.

    Length (the number of words) and complexity (the number of intents) directly impact the efficiency of communication with VAPAs.

    “Length--the number of words--and complexity--the number of intents--directly impact the efficiency of communication with VAPAs.”
6 of 6

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Voice assistants hold the potential to enable individuals with disabilities to govern their houses without constantly depending on caregivers.

    “Such systems hold the potential to significantly help users with motor and cognitive disabilities who currently depend on their caregiver even for basic needs (eg, opening a door).”

    From Investigating the Accessibility of Voice Assistants With Impaired Users: Mixed Methods Study (2020) · cites this paper as [12]

    The Hands-free Interaction and Preference for Voice Input sections describe voice assistants as promising for people with disabilities and report independent mobile-device use by people with limited hand dexterity (page unavailable). They do not address controlling a house, opening doors, or reducing reliance on caregivers for basic needs. (LLM, conf=0.85)

    Experimental — accuracy not yet measured

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

  • Microsoft is omitted from the relational-theme vendor list

    Minor

    The relational theme is attributed only to Google and Amazon, but its friendly-conversation subsection says Microsoft also recommends that behavior.

  • The cited early report is a listening-rates study

    Minor

    The discussion cites reference [6] as an earlier report on voice-design guidelines. The reference list identifies [6] as a study of human listening rates; the paper elsewhere identifies its earlier guideline review as [24].

2 of 2

Cites

Flagged first, then by citations

1 of the 39 references couldn’t be matched to a known paper.

Research works printed but not linked: 4

  1. 10.1186/1471--2202--14--74 · DOI
  2. 10.1007/978--3-030--15742--5_9 · DOI
  3. 10.1007/s00779-011-0470--5 · DOI
  4. 10.1016/0364-0213(88)90023--7 · DOI

Cited by

Sorted by citations

Related

Authors
Branham, S. M., & Mukkath Roy, A. R.
Affiliations
  • University of California, Irvine (US)
  • University of Maryland, Baltimore County, Baltimore, MD, USA
ORCID
Branham, Stacy M.
Venue
Proceedings of the 21st International ACM SIGACCESS Conference on Computers and Accessibility
Publisher
ACM
Published
2019-10-24
Pages
446-458
DOI
10.1145/3308561.3353797
OpenAlex
W2981912314
Semantic Scholar
8e75ce87feb968d67b7a7429cd7445d45339b4e8
Cited by
75
References
39
Citations by year
2026: 10 · 2025: 10 · 2024: 10 · 2023: 17 · 2022: 10 · 2021: 9 · 2020: 9
FWCI
13.45
Type
conference paper
Language
en
Open access
Licence
www.acm.org/publications/policies/copyright_policy#Background (vor)
Locations
International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS) (publishedVersion)
Funders
  • TRX Systems, Inc.
  • Maryland Industrial Partnerships Program
Topics
  • Social Sciences › Human Factors and Ergonomics › Digital Accessibility for Disabilities
  • Physical Sciences › Computer Science › Human-Computer Interaction › Usability and User Interface Design
  • Health Sciences › Health Professions › Occupational Therapy › Assistive Technology in Communication and Mobility
Keywords
design guidelines, visual impairment, accessibility, universal usability, assistive technology, voice user interface design

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-11.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-11.

Cited sources are checked by gpt-6-sol.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review