Automotive UX (IVI, Clusters & Driver-Facing Design)

Driving is the primary task; everything in the cabin is secondary, competing for a safety-critical resource — the driver's eyes and attention. Design for glances and interruption, never immersion. (Voice depth: Voice & Conversational Interfaces; LLM-assistant UX: Designing AI-Powered Interfaces.)

The governing constraint: driver distraction

• NHTSA Visual-Manual Driver Distraction Guidelines (2013, voluntary; built on the earlier Alliance of Automobile Manufacturers "Driver Focus" principles): tasks performed while driving should be completable with single glances off-road of ≤2 seconds and cumulative eyes-off-road time of ≤12 seconds; the alternative occlusion test requires task completion through 1.5-second vision windows totaling ≤12 s shutter-open time.
• The 2-second budget rests on naturalistic-driving research linking long off-road glances to sharply elevated crash risk. Chunk every task into one-short-glance steps; the UI must hold state while the driver looks away; no timeouts that punish watching the road.
• Task-time lineage: SAE J2364 (the "15-second rule," Paul Green, UMTRI) capped navigation tasks at 15 s of static task time before NHTSA's eye-glance criteria became preferred; anything long belongs in park. NHTSA also recommends driving lockouts (video, manual text entry, long text) — the root of every platform's parked-vs-driving split.
• Measurement standards exist — don't reinvent: ISO 15007 (glance metrics, driver visual-behaviour measurement); ISO 16673 (the occlusion method for assessing visual demand of in-vehicle systems).

Physical vs. touch controls

• Euro NCAP 2026 protocols (in force January 2026) award safe-driving points for physical controls on core functions — Euro NCAP names turn indicators, hazard lights, horn, windscreen wipers, and SOS/eCall. Touchscreen-only cars can still reach five stars in 2026 (60% safe-driving threshold), but thresholds rise through 2027–28.
• Industry is re-adding buttons: VW's design chief announced (2025) physical controls for volume, fan, seat heating, temperature, and hazards on all future models from the ID.2 on, calling their removal a mistake; Hyundai and Porsche have publicly recommitted to hard keys.
• Why buttons win while driving: blind-operable — fixed location, tactile edges, and detents let muscle memory replace vision (zero glances vs. multi-glance menu dives). Glass demands visual confirmation for aim and outcome; haptics on glass recover confirmation, not aim. Split: frequent/urgent/safety → physical controls in reach; infrequent configuration → screen; never bury climate or defrost.

Instrument clusters & HUDs

• ISO 15008 covers legibility of dynamic visual information for drivers (character legibility, contrast, colour recognition; test procedures; technology-independent) — consult the standard for values.
• Tell-tales: ISO 2575 defines the standard symbols; US FMVSS 101 regulates location, identification, colour, and illumination (largely adopting ISO 2575 symbols). Codified colours: red = immediate danger (brake failure), amber/yellow = caution/malfunction — never repurpose red/amber for non-warnings in a cluster.
• Cluster hierarchy: speed and active tell-tales always visible regardless of user configuration; configurable zones (media, nav, trip) surround, never displace, them.
• Functional-safety context (engineering): ISO 26262 assigns safety-integrity levels (ASIL) — tell-tales/speed rate higher than infotainment, constraining what software may render where.
• HUDs: virtual image focused at distance to cut refocusing cost; minimal high-priority content (speed, next manoeuvre, critical warnings); never occlude hazards or sightlines; auto-adapt brightness. AR overlays must be conformal (stick to what they annotate).

Platform ecosystems (state as of mid-2026)

• Apple CarPlay (phone projection) is near-universal. CarPlay Ultra — next generation, also rendering cluster and vehicle functions — launched May 2025 with Aston Martin. Hyundai/Kia/Genesis, Porsche, Honda/Acura, Ford/Lincoln, Nissan/Infiniti, and JLR are committed; once-touted partners (Mercedes, Audi, Volvo, Polestar, Renault) stepped back, wary of ceding the UI/data/services layer. Don't assume Ultra.
• Android Auto vs. Android Automotive OS (AAOS) — keep distinct: Android Auto is phone projection (the app runs on the phone); AAOS is an embedded OS in the vehicle, with or without Google Automotive Services (GAS = Play Store, Maps, Assistant/Gemini built in, licensed separately; some OEMs run AAOS without Google).
• Google's guidance: "Design for Driving" (developers.google.com/cars/design) covers Android Auto + AAOS apps and OEM systems; migrating to docs.partner.android.com/drivingux (sole home for updates from March 2026).
• App model: driving-optimized categories (media, voice messaging, nav, EV charging, POI) on templated, distraction-tested frameworks — all else parked-only; templates make apps inherit glance-safe behaviour.

Voice & multimodal in the car

• Voice is the distraction-mitigation channel — eyes on road, hands on wheel — but not cognitively free: the AAA Foundation / Strayer (Univ. of Utah) program showed voice interactions impose real mental workload, with residual impairment persisting up to ~27 s after a demanding task (phase III, 2015). A bad voice UI is a distraction.
• Push-to-talk on the wheel plus wake word is the standard dual entry; confirm consequential actions verbally (Voice & Conversational Interfaces, confirmation-by- stakes); pair voice input with glanceable visuals, not monologues.
• LLM assistants shipping (verified): Mercedes added ChatGPT to MBUX (US beta 2023, then production, via Azure OpenAI); VW's IDA integrates ChatGPT via Cerence (CES 2024); Mercedes' 2025 CLA uses Google Cloud's Automotive AI Agent (Gemini-based); Google announced Gemini for Android Auto and Google-built-in cars (2025). Designing AI-Powered Interfaces risks — verbosity, confident error — at highway speed.

ADAS & automation handoff

• SAE J3016 levels, shorthand: L0 warnings only; L1 steering or speed support; L2 both but driver supervises constantly; L3 system drives within its domain, human is fallback on request; L4 no human fallback within domain; L5 everywhere. The L2/L3 boundary — who is responsible right now — is the UX battleground.
• Takeover requests: multimodal escalating alerts (visual → auditory → haptic seat/wheel) with unmistakable urgency. Time-budget research yields no single number: Eriksson & Stanton (2017) found experimental budgets of ~0–30 s (mean ~6 s), takeover times averaging ~3 s, longer when drivers are deep in other tasks. Budget generously; design the alert to rebuild situation awareness, not just demand hands-on.
• Driver monitoring (camera gaze/attention tracking) is becoming standard — Euro NCAP rewards it — and is the honest precondition for L2+: attention warnings must escalate and eventually restrict the feature, not nag ignorably.
• Mode confusion is the central hazard: the driver must always know, at a glance, who is driving and what the system won't do. Ambiguous names and subtle state icons cause automation surprises — make mode transitions loud, explicit, confirmed. Trust calibration = the Designing AI-Powered Interfaces problem at higher stakes.

EV-specific UX

• Range honesty: displayed range is a prediction ("guess-o-meter" is the owner-community verdict on optimistic ones). Adjust for temperature, terrain, driving style; a hidden buffer below zero is fine, but never imply 0 = stranded-now. %-vs-miles is convention, not standard: % is stable, miles is what drivers plan with; show both where possible.
• Charging-stop planning belongs in native nav: automatic stop insertion, arrival state-of-charge per stop, charger power/ availability, predicted charge time (Tesla set the pattern).
• Plug & Charge (ISO 15118): cable-in = authenticated and billed, no app or card — the target public-charging UX; EU AFIR regulation now mandates ISO 15118 support on new public chargers.
• Preconditioning: batteries must be warm for fast DC charging — the car should prompt or auto-precondition when a fast charger is the nav destination, and say so (silent preconditioning reads as range loss).

Rear/passenger screens & motion sickness

• Passenger displays must not distract the driver: NHTSA guidance treats driver-visible video as a lockout item, and shipping solutions make it impossible — Porsche's passenger screen uses a polarized/privacy film the driver can't see; BMW's (2026) uses the driver-monitoring camera to dim passenger video when the driver glances over. Rules vary by market; "driver watches video in motion" is never acceptable.
• Reading in motion causes sickness via visual–vestibular conflict: eyes on a stable near screen while the inner ear feels the car move (same mechanism as VR — see Virtual Reality Design). Mitigate: mount screens high so peripheral vision keeps the outside world; stable UI, minimal in-screen motion; placement and breaks are the reliable tools today.

Sources

• NHTSA, "Visual-Manual NHTSA Driver Distraction Guidelines for In-Vehicle Electronic Devices" (Federal Register, April 2013); lineage: Alliance of Automobile Manufacturers Driver Focus principles.
• ISO 15007 (glance metrics); ISO 16673:2017 (occlusion); ISO 15008:2017 (in-vehicle visual presentation); ISO 2575 (tell-tale symbols); ISO 26262 (functional safety); ISO 15118 (Plug & Charge) — iso.org.
• FMVSS No. 101 "Controls and Displays," 49 CFR 571.101 (ecfr.gov); SAE J2364 + Green, P. (1999, UMTRI/HFES); SAE J3016 (sae.org).
• Euro NCAP 2026 protocols (euroncap.com); ETSC coverage (etsc.eu).
• Apple Newsroom, CarPlay Ultra launch (May 2025); 9to5Mac adoption tracking (Dec 2025); Google, "Design for Driving" (developers.google.com/cars/design → docs.partner.android.com/drivingux).
• Strayer et al. / AAA Foundation, "Measuring Cognitive Distraction in the Automobile" I–III (2013–2015), aaafoundation.org.
• Eriksson, A. & Stanton, N. (2017), Human Factors 59(4); Zhang et al. (2019) take-over meta-analysis, Transp. Research Part F.
• Mercedes-Benz press (ChatGPT in MBUX 2023–24; CLA/Google Cloud agent, 2025); VW–Cerence IDA (CES 2024); InsideEVs (VW buttons, 2025); TechRadar (Hyundai buttons, 2025); Porsche/BMW passenger-display docs.