What's Driving Drivers?
Understanding driver behavior and motivation and their impacts on ADAS engagement.
My Role:
- UX Researcher & Designer
For:
- Interaction Design
- University of Michigan Ann Arbor
- January - May 2022
Client:
Our client is a global technology company that designs and manufactures automotive components and systems for a range of vehicle types, including hybrid and electric vehicles. Driven by its commitment to making transportation safer, more sustainable, and more connected, our client is also at the forefront of the development of autonomous driving technology and Advanced Driver Assistance Systems (ADAS), with aims to expand research on driving in non-highway settings, including understanding the transition from highway to non-highway driving, how users perceive the automation, and when they choose to engage/ disengage the automation.
Project Goal:
This project aims to research driver behaviors and the use of partial automation on highway and non-highway roads and the transitions between these driving environments. To distinguish assisted driving experiences from the experience of driving as a whole, this project will be split into two phases:
- What's Driving Drivers?
Understand driver needs, motivation, and experience and how these factors influence the use of autonomous driving features;
- What's the current user experience?
Understand the user experience of current implementations of autonomous driving features using an existing commercial model as a benchmark.
For brevity, autonomous driving will be referred to as autopilot and/or ADAS (Advanced Driver Assistance Systems).
1. What's Driving Drivers?
Research Goals
For this phase, to better focus our research, we centered our efforts around the following questions:
- How do user needs and motivations affect their transportation preferences and needs?
- What are the differences in driving experience between highway and non‑highway environments; and how do these differences affect driver behavior across environments?
- What are the factors that affect the desirability and usage of ADAS features?
Methodology
To orient ourselves in the ADAS problem space and to gain a detailed understanding of the Tesla Model 3
(our client-selected benchmark) and its features, our group started with a round of secondary research
(on ADAS) and interaction mapping all of the Model 3’s ADAS related features.
We then furthered our understanding and contextualized our insight within the larger space of automation
within the Mobility industry through a comparative competitor analysis that covered 10 competitors (2x
each of direct, partial, indirect, parallel, and analogous competitors).
Finally, to better understand driver needs and motivations, the differences between driving in highway
and non-highway environments, and the use of ADAS features, we conducted a round of five interviews with
drivers with at least 2 years of experience using ADAS.
Synthesis
We synthesized our insight by creating an affinity diagram where we compiled our insights and sorted it into groups based on similarity, which allowed us to formulate key findings. To humanize our findings and contextualize driver needs, motivations, and frustrations; we created a set of personas that capture the represented demographic and attitudinal range.
To humanize our findings and contextualize driver needs, motivations, and frustrations; we created a set of personas that capture the represented demographic and attitudinal range.
Findings And Recommendations
Overarching Insight
Drivers currently only use ADAS features if they are in low-complexity environments and when the quality-of-life improvements that ADAS features bring to that driver outweighs the the added attention and energy it requires from drivers and the percieved risk in using autonomous driving features.
Key Findings
- Drivers do not currently trust autopilot systems, and, as a result, using autopilot features becomes more taxing because drivers feel the need to pay more attention when using autopilot features because drivers need to be already aware of the situation on the road if failures happen in order to react in time.
- Drivers that don’t fully trust autopilot systems will only use autopilot features under conditions that they believe to be perfect (long strips of low traffic highways in good weather)
- Due to current levels of trust, ADAS usage is mostly seen on longer trips rather than daily commutes, minimizing its benefits, namely making space for drivers to relax and providing convenience for tedious tasks.
Recommendations
- Provide drivers with feedback when the complexity of their driving environment increases so users
know when they need to pay more attention, create an internal monitor of the system's certainty of
its evaluation of the driving environment, and let users know when that certainty drops below a
certain threshold.
This would, ideally, eliminate the added attention draw that drivers face when using autopilot features
in complex environments or when transitioning into complex environments. -
Focus efforts on fixing ghost-braking and other failures that are not directly caused by
environmental complexity.
This aims at eliminating the added attention draw that drivers face when using autopilot features
in less complex environments due to not knowing when such events might occur. -
Create a log of all events and decisions that the system makes and encounters, categorize and
distill this into a log of events that the driver can review and understand.
This would allow drivers to more easily understand when, where, and why errors or failures might occur in the future, thereby, improving driver confidence and trust in the system under conditions where the system would not fail.
- Embrace regulation—especially when it comes to automation and safety. Well-regulated fields have less tolerance for risk, more standardized risk factors, and better and wider consensus on the perception of risk. Embracing regulation, decreases perceived risk and thereby increases willingness to purchase, test, and use automated products with high associated risk. Early proponents of regulation in lesser regulated fields are also often publicly perceived as safer than competitors.
2. What's the current user experience?
Research Goal and Methodology
To understand how the interface designs of autonomous vehicles affect the use of ADAS features and to gain insight into how to better design interfaces of autonomous vehicles, our team analyzed the interface of the Tesla Model 3 by conducting a heuristic analysis and a user test targeting the features and functions that were found to be highly relevant to building user’s trust towards the system during our user interviews. We selected the Model 3 as our ADAS benchmark to match our phase 1 benchmark and because Model 3 is a relative affordability model with an extremely high market share within the EV segment.
Overarching Insights
-
Ambiguity or lack of clarity in the design of the interface , such as the lack of visual differentiation between Tesla’s “Autosteer” and “Navigate on Autopilot” modes [Heuristic Issue 4],can induce user errors and confusion, such as believing that the system is failing when the system is actually functioning as intended . This can lead to decreased trust in the system as well as decreased willingness to test and use autonomous features. -
Lack of interface clarity and unintuitive design for features and actions outside of the autonomous system (such as adjusting mirrors, turning the car on/off, checking range, and opening windows) can lead to extraneous heuristic strain in learning the system.
This extraneous learning strain diverts attention away from familiarization with the autonomous system/features and can also lead to a general sense of confusion, which can, in turn, lead to more misunderstandings and errors.
These insights highlight the importance of designing clear and intuitive interfaces in the effort towards widespread use and demand for autonomous driving systems and features.
Heuristic Analysis
In the heuristics evaluation, five team members each took turns taking the roles of one notetaker, one videographer, one driver, and three examiners. A Tesla Model 3 with a basic Tesla Autopilot software platform was rented for four hours for the evaluation and driven on a predetermined route from Ann Arbor to Kent, Michigan to include both highway and urban driving scenarios to meet our research needs. The data was collected during the driving with video clips and notes. The individual evaluation was the first step of this heuristic evaluation, in which each team member examined the system individually from the data collected based on the 10 heuristics principles. The findings from the individual evaluation were aggregated to a shared Google Sheet file for further analysis including debriefing the usability issues based on the 10 principles, in-team discussion for the debriefing, and assigning priorities and severities for the issues based on the 0-4 rating of the heuristic evaluation. In this study, the rating of the usability heuristics issues was decided on the level of affecting users’ ability to drive, driving performance, and safety.
Severity Level 4 Heuristic Violations:
Solutions:
-
Lane Assist and Lane Departure Avoidance work to prevent the vehicle from drifting out of the bounds of the lane, but gets in the way when drivers are changing lanes or intentionally drifting outside of the bounds of the lane (such as to avoid obstacles).
Heuristics violated: Visibility of system status | Match between system and the real world | User control and freedom | Consistency and standards | Error prevention | Flexibility and efficiency of use
Create a very quick and easy to temporarily disable Lane Departure Avoidance.
-
Testers were not able to discover how to turn off the car and could not determine if locking the vehicle and walking away would turn the car off.
Heuristics Violated: Visibility of System Status | Recognition Rather Than Recall | Match between system and the real world | User control and freedom | Flexibility of Use | Consistency and standards
Create a very quick and easy to temporarily disable Lane Departure Avoidance.
-
System fails at roundabouts without warning and does not accurately display the environment (environment disappears entirely from display).
Heuristics Violated: Visibility of System Status | Match between system & real world | Error prevention | Recognize, diagnose, and recover from errors
Better map wider arrays of traffic junctions.
-
Despite an extreme difference between "Autosteer" mode and "Navigate on Autopilot" mode, the two modes share the same icon, and are only differentiated by a slight change in active states of that shared icon.
Heuristics Violated: Visibility of system status | Consistency and standards | Error prevention | Recognition rather than recall | (Possibly: Match between system and the real world)
Create a greater difference in the visual indication for the two states.
-
The button/switch to adjust the car window and to open the door is identical, which confuses the users. One team member accidentally opened the door by accident while the car was moving.
Heuristics Violated: Aesthetic and Minimalist Design | Error prevention | Recognition rather than recall | Match between system and the real world
Visually and physically differentiate window switches from door switches.
Usability Test
This usability test aims to understand how new drivers interact and understand Tesla’s unique user interface through a straightforward onboarding experience. This study incorporates five participants with no or little experience with Tesla to perform six tasks simulating everyday driving scenarios using a stationary Model 3. The tasks tested the common features of daily driving scenarios on the car display and were followed by usability interviews to gather feedback on the ease of completing tasks, any issues encountered, and overall impressions of the car display.
Usability Issues:
Recommendations:
-
The car icon does not communicate as the button to bring up the control hub, where most adjustments are made.
Redesign the Controls icon to more directly communicate its function:
-
The left scroll wheel is used to adjust both mirrors, but users intuitively want to adjust the left mirror with the left scroll wheel and the right mirror with the right-side scroll wheel.
Allow users to use either scroll wheels to adjust either mirror.
-
Because doors and trunks are controlled through the 3D model, drivers intuitively believe that this is also true for adjusting mirrors and checking the tire pressure; however, this functionality is not built in, breaking the system’s own standards, and, when paired with Key Issue 1, leads to confusion about how to adjust mirrors and check tire pressure.
Enable the adjustment-of mirrors and checking-of tire pressure through the 3D model in the center stack of the Tesla. Because doors and trunks are controlled by the driver through the 3D model, drivers intuitively believe that this is also true for mirrors and tire pressure.
-
The only way to see how many miles the vehicle can drive for any given level of charge is to tap the tiny battery percent number (not the icon, the number) on the top right of the 3D model of the car. It cannot be done through the Controls hub, where all other similar tasks are done. Of all of the users tested, only one figured out how to check the battery status in terms of miles (they did this after having tried everything else remotely related to the battery or the system status).
Build in the functionality of allowing users to change battery-percent view to miles-left view in the Charging Tab of the Controls hub.
-
The battery percentage and battery icon are two different buttons but communicate as a single element that is not interactive.
Visually separate the battery icon from the battery-percentage/miles-remaining number such that they present as separate entities while maintaining their relationship to one another.
-
The documentation and labeling of information regarding speed are unclear, confusing, and incorrect, with "Set Speed" versus "Current Speed" and "Offset".
Rewrite the documentation for “Set Speed” and “Offset” using more comprehensible terminologies and sentence structures that fit the average person.
byes~ thank you!