Home - Brand

PROBLEM ARTICULATION

How Might We Improve User's Trust on Autonomous Driving Vehicles?

To assist the autonomous driving design, a research is conducted in the psychology perspective (because of the strength of the BNUX program). The team started with learning different levels of autonomous driving.

The research direction is settled after literature review:

Most of the research was learning the "right of human override" in L3 autonomous driving scenarios
L4 Scenarios still need certain degree of human override, but there is a lack of research on the right of human override in L4 scenarios.
There is a similar psychological term to the "right of human override", the "sense of control", and the definition changes with the specific context.

Thus, the research question is : How might we facilitate the L4 autonomous driving design with "sense of control in driving"?

To answer the research question, the research question is split into the following questions:

What are the influencing factors of "sense of control in driving" ?
What are the scenarios with low sense of control ?
What are the user needs in the most needed scenarios?
How might the research results help out design process? Now and future.

Study Design:

Interview 1

What are the Influencing Factors of Sense of Control?

Interview Questions

The definition of "sense of control" was split into 5 dimensions referring to the psychology definition of "sense of control of learning". Interview questions were designed based on the definition.

The sense of being interrupted
The sense of the environment
The sense of suspicion
The sense of contributing effort
The sense of helplessness

Interview Process

Recruitment (N=19): the study focuses on the majority of people, so as long as the driver has more than one year's driving experience. All participants filled in a pre-study questionnaire to collect ethnography information and daily driving information.

To recall their driving experiences, all participants went through a stimulated driving experiment, in which people manually drove the car in the HMI lab as they would normally drive, experiencing the embedded typical driving scenarios.

Inspiration cards were used in the interview when the participant was less engaged in talking.

Data Analysis and Result of Interview

The interview data was transcript to proceed the analytical coding. The whole team engaged in the coding process with an emphasis on inferring the reasoning between the lines. The reasonings are the motivations that influence driving, so they're called influencing factors. The influencing factors were then categorized according to 1) scenarios and 2) power of influencing human driving according to a KANO model:

The results are collected to make a set of cards for design team's future brainstorming. This card set contains all the influencing factors of daily driving scenarios.

Focus group

What are the Scenarios with Low Sense of Control?

Based on the driving scenarios found in the last section, a focus group was conducted with driving experts to find out the scenarios with low sense of control. They were also welcomed to come up with more low sense of control scenarios.

Another set of card was made based on focus group data. The elements in influencing factors were extracted to prompt another perspective of categorization. This card set contains the influencing factors and elements of low sense of control driving scenarios. This card set is also the tool for design team's future brainstorming.

INTERVIEW 2

What are the User Needs of L4 Autonomous Driving in the Low Sense of Control Scenarios?

Interview Process

Recruitment (N=14, 11 as drivers + 3 as passengers users): The interview is going to reveal the user need in L4 autonomous driving scenarios, so there were 3 participants sit in the car as a passenger. Each participant experienced similar process as that in interview 1: pre-study survey -> stimulated HMI driving in low sense of control scenarios -> interview.

The interview collected data on participants’ gestures, experiences, feelings, and motivations in simulated scenarios.

Data Analysis and Result of Interview

Then the data were analyzed with open coding, axial coding, and selective coding.

A set of card was made based on the results: the low sense of control scenarios and corresponding user needs. This card was then used in the following design workshop.

Evaluation

Which Scenario Should We Focus On?

Since the time is limited for the future workshop, problem solving should be focused. The team invited experts and users to filter out scenarios that most worth working on. The team conducted an expert review and sent out user's survey to gain the data.

Four core scenarios were filtered out with the consideration of quantitatively, the frequency (user's survey), and qualitatively, the technology accessibility (expert review):

design workshop

How Might We Inspire the Design Team with Research Results?

Two brainstorming sessions were held to enlighten insights and solutions on the user needs listing above. Designers, researchers, and developers joined in the session. Due to the time limitation of the workshop, only 10 scenario-user need card were randomly chosen for each group.

Brainstorming 1 : Paper Passing

This activity aims to share ideas of solutions to user needs from every member's perspective, gathering design solutions from different participants. Each paper was passed on among all participants in the same group for writing down ideas that meet the user need on specific scenario-user need card.

Evaluation 1 : Discussion

Ideas were collected and shared. Then the group voted to select those insights worth designing, using "I want" and "I need" as criteria.

Brainstorming 2 : Idea Association

This activity aims to make connections between seemingly irrelevant ideas. Participants write down ideas inspired by the given inspirational card, then tried to apply these ideas into design solutions to solve the corresponding user needs.

Evaluation 2 : Role Play Discussion

Participants randomly selected one role play card, then shared in group from the perspective of corresponding roles to judge the design solutions.

idea proposal

How Might Translate the Design Ideas into Visualized Designs?

To communicate the design ideas in a better way, the team visualized the research results as examples that design team can refer to.

Infrastructure

Baidu company has developed models for autonomous driving that we can use in design.

Design 1. Parking

The gesture dragging: This can feel like more confident on identifying a position for parking, while the parking is totally autonomous. The dragging gesture gives driver more certainty.

Design 2. Passing Truck

Voice interaction instead of paying attention to the mirrors all the time: When there's a small car passing by the truck, the internet of vehicles identify the intention of passing and prompt what is going on. So here are two ways of noticing passing cars for truck drivers, knowing that can increase car drivers more confident about safety.

Design 3. Position of Interest

Take your time off to interact with the road: Since L4 autonomous driving allows people more time to paying attention off the road, it is an opportunity to create and view information based on location. More information, more confident about driving.

Design 4. Taking Over

Like watching a game: Although drivers won't have to control the vehicle, it is still important to know what is going on. The AR on windshield can tell detailed information to make drivers feel confident.

Design evaluation

How's the Design?

Expert Review

Experts (N=3) rated the designs according to feasibility, technology, and traffic regulations. They also left comments to explain the reasons. Mean score was calculated according to the criteria.

User Review

User participants (N=6) rated the designs according to the degree of meeting your needs and personal satisfaction. They also left comments to explain the reasons. Mean score was calculated according to the criteria.

self-reflection

How Might I Do Better?

Reduce redundancy: since it was still a practical project to a certain extent, and there were more members than needed, some of the evaluation phases were having redundancy. I can choose one evaluation method depends on time limitation instead of taking another user evaluation after already having an expert review.
The perspective of thinking: this study was proceeded with an assumption: the L4 autonomous vehicle driver's override is the same as that with driving daily regular cars. As I give a doubt on that, I would personally dig in to find out if this is true.
Refine the workshop: there should be a framework for discussions and evaluations.
Communication through out functional teams: I personally think the biggest contribution from the researchers are the "reasonings" of the influencing factors, but this is a black box for design teams, I prefer something like journey map that enables more discussions with design teams.

CHALLENGE

How might we improve user's trust in Autonomous Driving (L4)?

CLIENT & ACADEMIC

MY ROLE

Method & TOOL