707 OS
Timeline
Mar 2026 - Jul 2026
Role
Solo HMI Designer
Project
Master Thesis
Collaborators
-
Overview
Problem
Drivers shouldn’t have to hunt for basic controls.
Moving a familiar control onto a screen changes more than its appearance. A direct action can become a sequence of navigation, visual checks and hesitation, demanding more attention from the driver.
I focused on where functions are placed and how many actions it takes to reach them. The challenge wasn’t to replace screens with buttons, but to understand which type of control best suits each function.
Research
I compared 10 vehicles and recreated their most frustrating feature flows
Across 10 infotainment systems, the same usability problems kept appearing. Try three everyday tasks in this simplified demo. The systems were chosen from 2 tests (shown below).
Tap the controls in the demo below.
A simple adjustment sits behind several menus. Try reaching it on the screen below.
Illustrative composite, not a replica or a measured vehicle test. Menu paths and the 1.2-second feedback delay are simulated. Your interactions here are not research results.
Research Basis: Vi Bilägare (2022), Carbuyer (2025)
What the comparison revealed
Across the 10 systems, recurring patterns emerged around control placement, interaction effort, familiarity and feedback.
Academic Research
The comparative analysis showed what worked. The existing research explained why
Research Basis: Wickens (2008), Anderson et al. (2024).
Solution
A framework for placing every control
I organised controls around two questions: how often is the function used while driving, and how much should the driver be able to customise it?
What each tier represents
Reusable beyond one interface
The framework asks teams to consider frequency, configurability and the cost of menu depth when deciding where a feature should live. These are design heuristics, not calibrated predictions of driver distraction.
Fixed physical controls
Used often, with no personalisation needed. Keep the control fixed and predictable.
Too many steps for frequent use. Consider direct access through a physical control.
How is this recommendation made?
Planning targets one access action for tasks used while driving. In the penalty grid, occasional use with 2–3 actions is Moderate; occasional use with 4+ or frequent use with 2+ is High. Feature names label the result; the selected inputs determine it.
Usage and configurability suggest a tier. Tap count checks the effort of reaching the function. Low effort does not require fewer taps; high effort calls for better access. “Never” means no use while driving, so that penalty does not apply. These are design suggestions, not measured safety ratings.
For occasional use with high configurability, Tier 3b describes detailed setup. A high access penalty still means the task needs easier access if performed while driving.
Implementation
Turning the framework into 707 OS
I translated the framework into an interface with fixed control locations, personal shortcuts and clear visual states, supported by a mix of digital, physical and hybrid controls.
Personal without becoming unpredicatable
The shortcut containers stay fixed, while their contents can be personalised. The upper stack links to pages, while the lower stack can also display live feature states. Shortcuts can only be remapped while the vehicle is stationary.
Keep the essentials in reach
Persistent controls stay anchored across screens, while hybrid controls combine fixed physical input with changing digital states. The climate/shortcut toggle updates the lower card without moving its position, keeping both access and feedback consistent.
A consistent design system
Geist typography, violet active states and an 8 px grid create consistency across the 1600 × 1200 prototype canvas. Final display dimensions and in-vehicle legibility would still need to be validated in a production environment.
Usability Test
Testing while the attention was divided
I tested the prototype using a driving simulator, asking participants to complete interface tasks while navigating a medium-traffic highway scenario on Asseto Corsa.
5 Participants
3 minutes familiarisation
5 tasks
Results
707 OS averaged 10.04 seconds, compared with 10.0 seconds for the Volvo V70 (2005) and 44.6 seconds for the MG Marvel R
The benchmark figures come from separate testing conditions and are included as reference points rather than a direct experimental comparison.
Takeaways
What I would do next + limitations
Next, I would build functional physical controls, test the system in a controlled vehicle environment and apply the model to another interface. A larger study with consistent comparison tasks and direct attention measures would test the framework more rigorously.
The current prototype cannot reproduce the full behaviour of production hardware, and Figma interaction constraints required compromises in some controls.
More Projects
