Grounded

Week 18

Building a voice and identity

Research question. How can Grounded sound clear and adult when airport information changes, while still feeling welcoming and offering spoken guidance for people who choose it?

This page develops the voice research after the brand synthesis in GDE751: Week 18: Brand research. The audience is autistic adults travelling independently. The Critical Report includes ten self-declared autistic adults; participants described independence and control, late or changing information, sensory environments and fragmented systems. Those findings explain why clarity and choice matter here. They do not show that every autistic adult wants the same tone, pace, voice or amount of help. The words and audio below are therefore design hypotheses, not participant-approved language.

Why voice changes meaning

“Voice” has two connected meanings in this project. Brand voice is the character carried by words and visual expression. Spoken voice is information heard through a system voice, a recorded speaker or a character. The same sentence can feel different when its order, punctuation, pace, pitch and context change. A smiley mascot or a warm voice can soften a message, but it can also make a serious alert seem less reliable. A formal voice can feel authoritative, but may sound distant or leave a person to work out the next step.

The key design question is not “How do I make everything sound friendly?” It is “What does the traveller need to understand at this moment, and what tone helps without adding assumptions?” During an urgent change, the current fact and action should lead. A less time-sensitive preparation screen can offer more explanation. A support prompt can feel warm while leaving the decision with the traveller.

Three voices to compare

| Reference voice | How it communicates | Useful learning | What does not transfer |
| --- | --- | --- | --- |
| Apple interface writing | Apple’s Human Interface Guidelines say to set a consistent voice and vary tone by context. Their examples contrast a direct fall notification with a light, congratulatory activity message. Apple also recommends clear words, important information first and action-led labels. | Keep core wording consistent, then let the situation set the tone. A gate change should state the new gate and action. A completed task can be more celebratory. | Apple’s examples are not specific to autistic travellers or airport changes. Following a style guide does not prove that the copy is understood. |
| Therapeutic and wellbeing apps | Headspace uses short, named activities and an illustrated, recognisable identity. Finch describes its bird as a gentle, encouraging companion and frames goals as small steps. These brands use reassurance and companionship as part of an ongoing self-care relationship. | Offer support in an approachable way; name what the support is for; make the product recognisable across different moments. | Airport directions are operational information, not a therapy session. “You’ve got this” or “take a deep breath” could distract from a gate change, presume a feeling or sound childish when repeated. |
| Public service and accessible information | NHS England guidance asks for communication that is personal and direct, understandable when spoken aloud, respectful and empowering. W3C cognitive accessibility guidance recommends familiar, literal wording, concise content and one instruction at a time. | Write for a real person, say what is known, and make the next action explicit. Plain language can remain adult and informative. | These are broad guidance sources. They do not decide the app’s exact personality or replace testing with the intended audience. |

This comparison suggests a direct core voice with optional warmth. It can use the clarity of an operating system or public service for time-sensitive facts, then take the therapeutic-app lesson of approachable support in optional moments. The brand should not imitate a clinician or speak as if it knows the traveller’s emotions. The companion can express the identity; it should not deliver the only version of essential information.

Accessible language without talking down

Plain language is not childish language. It means removing unnecessary work from the sentence: use familiar words where they preserve the meaning, explain unavoidable terms, keep sentence structure straightforward and place one instruction at a time. An adult reader should still receive the full fact and enough context to make a decision. Avoid babyish praise, excessive exclamation marks, diminutives and commands that sound like a teacher addressing a child.

The W3C guide for people with cognitive and learning disabilities is supplemental guidance, not a substitute for WCAG conformance or user research. It recommends including people in design and testing because cognitive accessibility needs vary. Its clear-content guidance includes using literal language, simple sentence structures and separate instructions. NHS tone guidance makes a related point for public communication: write personally and directly, avoid jargon and make complex subjects understandable. Neither source says that all readers need the same amount of detail. The design should give the essential instruction first and let a traveller open more detail if they want it.

For this app, “Your flight has been reassigned to gate A04” is grammatically formal and hides the useful word “changed.” “Gate changed. Go to A04.” is shorter and action-led, but still adult. The message should also identify the former gate if that context matters: “Previous gate: B12.” A traveller should not need a warm introduction before reaching the fact.

Avoid language that labels or diagnoses the user. “You seem stressed,” “we know this is hard,” and “calm down” are not grounded in what the app can know; they may feel intrusive or patronising. Prefer an offer that describes the feature: “Would a step-by-step route help?” with a clear “Not now” option. The wording respects the person’s choice without suggesting that they are unable to manage the journey.

Written and spoken information

The app’s auditory option needs to be considered separately from VoiceOver, Apple’s screen reader. VoiceOver speaks interface labels and content for people who use that assistive technology; it should receive a logical reading order and useful labels. An in-app “listen” control is a separate, optional way to hear the current travel instruction. Decorative mascot content should not be announced as if it were operational information.

The W3C’s 2026 cognitive accessibility research module for voice systems is a draft note, so its recommendations remain under development. It describes how speech interfaces can place demands on memory, processing speed, attention, listening and the ability to answer in real time. It also discusses user control, simpler prompts and fallback paths. This matters even if Grounded does not become a conversational assistant: spoken instructions can disappear as soon as they are heard, and a traveller may be in a loud terminal, using headphones, processing other announcements or unable to listen at that moment.

For a first version, the most grounded exploration is one-way, on-demand spoken guidance, not a voice chatbot. A traveller could tap a clearly labelled speaker button to hear the same current instruction shown on screen. The text remains visible and is not replaced by speech. Audio would not autoplay. The listener could pause or replay it, and the app would respect the device’s speech settings where possible. A short phrase and pause between steps may be easier to follow than a long paragraph. No route should depend on hearing the instruction, and no essential information should exist only in audio.

The distinction between in-app speech and system accessibility also matters technically and ethically. The app could use the device’s built-in text-to-speech voice to read a structured text string; this is consistent with current platform settings and avoids recording a fixed “mascot voice.” A recorded human voice could sound more natural but is harder to localise, update when information changes, and personalise for different listening preferences. These are implementation options to investigate, not a decision that the app already supports them. W3C guidance on audio control recommends giving people control when audio plays; its cognitive-accessibility guidance also stresses that audio can compete with screen-reader speech. A tap-to-play model avoids many interruptions, but still needs testing with VoiceOver and common airport noise.

Same gate change, different meanings

Event used in these examples: BA256 was at B12 and is now at A04. The airport data is fictional.

| Treatment | Written version | Spoken version and likely effect | Question to test |
| --- | --- | --- | --- |
| Institutional | “Gate amendment: BA256 A04. Refer to airport signage.” | Read in a neutral system voice, the nouns and abbreviations may sound official but compressed. “Amendment” and “refer” make the traveller infer what changed and what to do. | Can the traveller identify the new gate without interpreting airport jargon? |
| Overly therapeutic | “Oh no, your gate has changed! Don’t worry, we’re here for you. Take a deep breath and we’ll get through this together.” | A warm or expressive recording may make this sound caring, but it adds several sentences before the useful destination and assumes distress. The same reassurance can feel comforting to one person and intrusive to another. | Does the supportive tone delay the fact or presume how the traveller feels? |
| Direct and companionable | “Gate changed. Go to A04. Previous gate: B12. Would a step-by-step route help?” | A calm, unhurried system voice can read the same words with a short pause between the change and the optional route offer. The traveller can stop after the essential fact or continue to hear the option. | Is the change immediately clear, and does the optional offer feel respectful? |

The preferred direction is the third treatment because it gives the destination first and leaves support optional. This is a reasoned candidate, not a tested winner. Pauses, pace and pronunciation may change how it feels. In particular, the spoken message should say “A zero four” or “A oh four” consistently only after checking what travellers naturally understand from the gate code. We should not assume that written abbreviations sound clear aloud.

Working aims and rules

| Aim | Do | Avoid |
| --- | --- | --- |
| Clear before comforting | State what changed, the current fact and the available action. | Opening an alert with a joke, mascot line or reassurance before the gate. |
| Adult and respectful | Use ordinary, direct language and give the full relevant information. | Baby talk, “good job!” for ordinary actions, or wording that implies the user cannot cope. |
| Warm without guessing | Offer support as a question: “Would a step-by-step route help?” | “You seem anxious,” “we know you’re overwhelmed,” or “calm down.” |
| Specific and honest | Mark information as current, previous, sample, sent or confirmed. | Saying “help is coming” before a service has confirmed it. |
| One step at a time | Break a route into short actions and let the traveller choose when to hear or reveal detail. | Long spoken paragraphs, several choices in one breath or a fast countdown. |
| Optional sound | Provide a visible play/pause/replay control and retain the same information in text. | Autoplay, sound-only meaning, or audio that competes with screen-reader speech. |
| Consistent terms | Use “gate,” “previous gate,” “route” and “next step” consistently. | Switching between “destination,” “terminal,” “location” and “gate” when they mean the same thing. |

Voice options to test

These are three possible expressions of the same underlying voice. They are not three separate brand personalities.

A · Neutral and efficient

“Gate changed. Go to A04. Previous gate: B12.”

This maximises speed and keeps the alert factual. It could feel cold if the interface offers no way to get more guidance.

B · Direct with a human touch

“Your gate changed to A04. The previous gate was B12. Would a step-by-step route help?”

This is warmer because it addresses the traveller directly and offers help, while keeping the fact first. This is the strongest starting hypothesis for the written interface.

C · Companion-led

“Gate changed to A04. I can show you the next step if you want.”

This gives the brand a more recognisable speaker. It could suit onboarding or optional support, but repeated first-person character speech may feel childish or distract during a time-sensitive alert. If explored, the same route must work with the character hidden or muted.

For audio, the same message can be spoken by the device’s selected system voice, a single recorded human voice, or a character voice. The first is more adjustable and updateable; the second could sound more naturally paced but is costly to revise and localise; the third has the strongest brand personality and the highest risk of competing with the seriousness of the task. No voice has been chosen. The first experiment should compare the same short recording or speech-synthesis output at a measured pace, with a visible transcript and a play/pause/replay control.

This existing type-and-copy specimen is an early hierarchy experiment. Its useful question is whether the gate number and action remain clear before illustration or character voice is considered. It is not a final screen or a test result.

How I would check the voice

A next research round should ask autistic adults to compare written and spoken versions of the same gate change. Ask what changed, which gate is current, what action is available, how the wording feels, and whether the audio pace and voice are comfortable. Include a no-audio path, replay and pause controls, and a way to ask for more detail. Test the pronunciation of gate codes rather than deciding it from the written form alone. Record disagreement as useful evidence: the aim is to offer enough control for different preferences, not to identify one universal “autistic voice.”

Provisional direction: Grounded speaks clearly and respectfully, like a capable travel service, with warmth available in optional support moments. Spoken guidance should repeat the current written instruction on request, with user control over listening and an equally complete visual route. The examples and rules on this page are hypotheses for testing, not findings from the Critical Report.

Sources