Rideshare Ethnography 🚗
Ethnography & Co-Design Study
Project
Academic Research Paper
Conducted
May - September 2021
Published
Accepted to the ACM CHI Conference on Human Factors in Computing Systems 2022
Methods
Ethnography, Co-Design, Remote Research, Qualitative Analysis
Here I am presenting the project at the 2022 Good Systems Research Symposium
Project Overview
Inspired by prior work and to guide this research project, I developed two primary research questions:
RQ1: How do gig work’s algorithmic management and platform design affect worker well-being?
RQ2: What do gig workers desire to see in technology designs that support their well-being and work preferences?
We also framed this study along three aspects of well-being: psychological, financial, and physical.
In this project, we first conducted an ethnographic study with 19 rideshare drivers, learning about their work preferences, issues with algorithmic management, and documenting their lived experiences.
Then, we conducted participatory co-design sessions with 15 drivers to co-explore solutions to improve worker well-being regarding algorithmic management.
My work in this project was performed as part of my work in the University of Texas’ Human-AI Interaction Lab.
Why?
Algorithmic Control
Rideshare drivers are a marginalized and oft-forgotten group. Their work is characterized as temporary, short-term, and on-demand.
Platforms advertise the work as low commitment, hyper-flexible, and highly customizable.
In theory, because drivers are independent contractors, they should be able to accept tasks as they please.
However, drivers often find themselves controlled by heavy-handed algorithmic management.
This can take the form of harsh, punitive measures for rejecting gigs, directing and manipulation to accepted jobs they would not have otherwise, and exclusion or expulsion from driving rewards programs.
A recruitment ad from Uber’s website.
Research Questions
“It’s a stupid video game where it’s trying to keep you from winning”
Participant 6 on Uber’s Management Style
Methods
Focus Groups
Participants recruited through partnership with the Independent Driver’s Guild of Chicago and online on Reddit.
Conducted remotely using Zoom.
Example gigs were presented to drivers, with them evaluating each gig as it pertains to each aspect of well-being.
We also had participants evaluate Uber’s different management features - including Quests, reward structures, Surges, and rider information - along their well-being needs.
Co-Design Sessions
Quests promotions encourage drivers to work dangerously long hours.
Uber Pro has strict requirements that control driver behavior.
Storyboards
We drew 3 storyboards to show to participants. These contained potential interventions for collectively overcoming algorithmic management.
We also performed a pseudo contextual inquiry. Educated by our previous focus groups, we created five prompts that described situations that participants previously mentioned that negatively affect their well-being.
This storyboard educates drivers about collective data auditing.
Contexual Inquiry
Using Google Jamboards (a virtual whiteboard similar to Miro or FigJam), wrote participant responses live as a way of organizing and ensuring that our interpretations were accurate.
Using the contextual inquiry scenarios, we asked how each intervention could improve [Quests, for example] with respect to their well-being needs (psychological, physical, financial)?
This board asked participants to evaluate how collective auditing could help them better understand Quests.
Diverse Recruitment
50% - Ethnic Minorities
33% - Underrepresented Genders
71% - Drove Multiple Apps
54% - Held Another Job
83% - Used Ridesharing to Meet Essential Needs
P19 on being wrongly deactivated
"They’re [Uber] very quick to deactivate. They don’t seem to really fully appreciate the importance of our income."
P15 on trying to be re-activated
“Especially if Uber is your main source of income, to be deactivated can be incredibly stressful"
Quotes like these helped us get a grasp of the situations drivers found themselves in.
They also helped us develop the contextual inquiry scenarios we designed for the co-design.
Findings
Problems
Solutions
Lack of Well-Being Support
Data-Driven Insights
Through our focus groups and co-design session, we identified four primary sets of problems. In the following section, I describe the problems, as characterized by drivers, each paired with the solutions that drivers offered.
Reflecting on how platform features supported or hindered their financial well-being, drivers felt that platforms intentionally obscured statistics surrounding their work, forcing drivers to focus on short-term goals rather than taking a long-term view of their role and prospects.
"Uber discourages you from taking that long view of the past month or the past few previous months. It does remain very weekly focused."
- P7
When considering their physical and psychological well-being, drivers often felt they had to forgo well-being measures due to algorithmic management features, such as Uber’s tier system pressuring them to accept a high volume of rides.
"Mentally, if I see that dollars per hour going way down? That’s pretty disheartening. I mean, I’m doing this as a second job for extra income. If it’s not proftable, why am I wasting my time?"
- P12
During design sessions, participants found inspiration in the possibilities of their driver data being used to help themselves better understand their work patterns and make data-driven recommendations to them.
"Uber is a data company, I mean they’re tracking everything that you do...so they’re monetizing that data in some form or fashion."
- P8
Drivers also imagined how data-driven insights could be applied to improve their experience with driving incentives like Quests. P15 suggested an intervention that would combine personal driving data and real-time driving information to give drivers an estimate on the attainability of each Quest option (i.e., predicted time to complete each Quest).
"[Uber] forces people to miss out on certain things with people who are important in their lives because they want to hit a Quest"
- P15
Gamification
Flexible Incentives
The gamification of their profession makes drivers feel like they work in an unequal system, a system with unclear rewards and objectives.
"They [Uber] gamify this [Quests] in a really unethical way. And they don’t have to do that.
- P17
Sentiments of all drivers invariably returned to their concerns over unfair driver treatment by algorithms that prioritized luring in new drivers or recently dormant ones instead of rewarding high-performing, loyal drivers.
"I’m not saying don’t offer incentives for new drivers. But I think you’re better off keeping drivers that have been driving for years, that are experienced, that have good ratings, that are consistent.
- P18
In discussions around how incentive differentials could be improved to center worker well-being, one theme of solutions that was discussed revolved around allowing drivers to play a role in configuring their incentive offers.
It’s like building a personal pizza, the toppings are rewards that drivers can combine and change depending on their circumstances that day.
- P6
Another driver-imagined configuration was the ability to switch between Quests offers instead of being locked into one Quest with incremental pay levels.
I may get into a situation where I can’t complete the Quest. And you know, I may do 98% of the Quest, but don’t get any of the bonus!
- P20
Uneven Information Access
Translucency in Task Assignment
The gamification of their profession makes drivers feel like they work in an unequal system, a system with unclear rewards and objectives.
"They [Uber] gamify this [Quests] in a really unethical way. And they don’t have to do that.
- P17
Sentiments of all drivers invariably returned to their concerns over unfair driver treatment by algorithms that prioritized luring in new drivers or recently dormant ones instead of rewarding high-performing, loyal drivers.
"I’m not saying don’t offer incentives for new drivers. But I think you’re better off keeping drivers that have been driving for years, that are experienced, that have good ratings, that are consistent.
- P18
In discussions around how incentive differentials could be improved to center worker well-being, one theme of solutions that was discussed revolved around allowing drivers to play a role in configuring their incentive offers.
It’s like building a personal pizza, the toppings are rewards that drivers can combine and change depending on their circumstances that day.
- P6
Another driver-imagined configuration was the ability to switch between Quests offers instead of being locked into one Quest with incremental pay levels.
I may get into a situation where I can’t complete the Quest. And you know, I may do 98% of the Quest, but don’t get any of the bonus!
- P20
Thematic Analysis
As part of the analysis of our results, we organized our ideas into this thematic analysis. We did this to better summarize our findings and ensure consistency.
Driver Identified Sub-Problems
Main Problems
Co-Designed Interventions
Specific Design Recommendations
Implications
In this study, we provided a case study for worker-oriented co-design. Utilizing digital methods and tools including digital whiteboards, artifact analysis, and online survey recruiting, paired with more traditional methods of ethnography, we conducted an entirely remote ethnographic study.
Case Study
We also provided several concrete recommendations for rideshare platforms to strengthen their relationship with their drivers.
Because of the unique psychological contract rideshare drivers have with platforms, it is crucial to ensure a healthy relationship with appropriate and interactive algorithmic management.
Recommendations
As a journal paper, this work was presented at the 2022 ACM CHI Conference on Human Factors in Computing Sysstems. A photo of me attending the conference is on the right.
I learned a lot at CHI, how to transition from academic research to industry-oriented UX research, heard interesting talks on a variety of fields, and learned a lot of new interesting research methods I hope to use in the future!
CHI Conference
Reflection
What went well:
We tried a very novel experimental research method, and it worked!
We targeted a non-expert domain user base for co-design, which can be very difficult because of the steep learning curve in design. However, we grew to develop better strategies for building a common vocabulary with participants, significantly reducing the burden on them.
We created specific and actionable design recommendations for rideshare platforms, some of which we’ve actually seen implemented in the time since the study was completed!
What could have gone better:
As my first publication-grade qualitative study, this project initially lacked focus. However, with the help of my co-authors, we narrowed and specified the scope, finding a gap in the existing literature that we hoped to fill.
We didn’t actually work to create or see through any of the solutions we proposed, we conducted a follow-up study that completed this!
We didn’t involve Uber or other rideshare platforms, meaning our suggestions may go unheard. We hope to work together with platforms with these findings in hand.