UX Innovation > In-Vehicle UX Blog

Early AV Rider Research: Garbage In, Garbage Out

by Derek Viita | Jun 18, 2020

Over the past decade, Strategy Analytics has published several reports on the AV industry's effort to understand rider UX in a driverless car.  Or, more accurately, the industry's lack of effort on driverless UX.  Most industry research shared to the public has been limited to caveat-laden PR-driven demos of "first rides."  And most of that research has merely touted the obvious and well-worn relationship between exposure and trust.

As part of an ongoing series of webinars, this week the PAVE campaign (a special interest group for AV stakeholders, meant to educate the public on AV technology) showcased select industry research on "early rider experiences."  PAVE hosted speakers from Dataspeed, Aptiv, and Lyft to discuss their research.
- Dataspeed touted its partnership with SAE providing demo AV rides (on an indoor closed course) at last year's Detroit auto show.
- Aptiv touted its partnership with Lyft providing "driverless" (with safety driver) ride-sharing in Las Vegas USA.
- Lyft touted an additional partnership with Waymo providing "driverless" (with safety driver) ride-sharing in Phoenix USA.


Self-driving Lyft displays "Ready to Start" before ride begins


Rider or traveler UX absolutely deserves more attention than it is currently receiving in the AV space.  And all PAVE webinar participants deserve kudos for attempting to shine some light on their efforts thus far.  But this is a complex set of UX and human factors issues, requiring extensive research into human behavior and attitudes, technology development, and public policy.  This effort requires giant floodlights.  And unfortunately as with most industry-driven research thus far, the efforts described on the webinar were merely small candles.

The common thread through most existing research (including all of the research discussed during the PAVE webinar) has been a focus on "first rides."  And indeed, the first-ride experience is important for trust establishment.  But this is not new territory.  Strategy Analytics has tracked "trust" as a concern for many years.  Waymo and Intel, among others, have led the charge for AV stakeholders exploring trust factors in AVs.  Waymo has even established best UI design practices for trust.  More bluntly:  "Encouraging rider trust in AVs" has been the most over-researched topic in transport for the past 5 years.

Another common thread for most early efforts:  Sampling bias.  Many on-road services, including the Waymo One service and both Lyft trials in Las Vegas and Phoenix, are almost entirely reliant on early tech adopters.  By Aptiv and Lyft's own admission during the PAVE webinar, regarding their Las Vegas trial, riders are "more likely to be accepting of an AV before even riding" in one.  So the sample of Lyft users who are asked about such topics as "trust" and "satisfaction" are more likely to be giving skewed feedback.  Just like Tesla owners who forgive poor performance in Autopilot or Smart Summon because they are excited to "help the system learn," early adopters are more likely to feel invested in assuring that new products advance for wider development, regardless of readiness for market.  Research relying heavily on this population should take great care to avoid or account for this bias.

The early adopter bias shows up quite readily in certain UX metrics.  For example: During the PAVE webinar, Lyft's representative touted the outstanding performance of their "driverless" (with safety driver) rides on user-based metrics.  These included a high average score on Lyft's star-based rating system (4.95 out of 5), and positive open feedback in the Lyft app's comment box.  Though each of these metrics are important data points, the reliance on them exclusively to make claims about rider comfort (especially with early adopters) is ill-advised.  In ride-hailing star ratings, anything below a 5 tends to have a stigma attached to it, and open-ended qualitative feedback is often reserved for rides that leave an extreme impression.  So early adopters (who are excited about driverless technology) are almost guaranteed to create a ceiling effect in the data.

One other common thread in AV UX research:  The inclusion of a safety driver.  The vast majority of current on-road efforts (with the notable exception of one portion of Waymo's Phoenix fleet) are not truly driverless.  A safety driver, and sometimes an additional "chaperone" to monitor, are always present in the car.  This, in and of itself, presents a confound in any research attempting to make claims about rider UX.  Aptiv's representative even cited an example of such confounds during the webinar, when describing research on driverless experiences for visually-impaired populations:  "They (visually impaired people) are excited because they could just 'be' in the vehicle without interacting with passengers" who are trying to be helpful but chatty.  "They would say 'sometimes I'm not in the mood to chat with other people.'"

Ultimately, the biggest problem with all efforts described in this PAVE webinar (and with almost all UX research in the AV space) is a limited focus on the first ride experience.  Though the first ride is important, it is merely one aspect of an extensive range of UX issues which must be addressed (from hailing processes for limited mobility users, to assuring cleanliness and safety in the time of COVID-19, to objective ride comfort metrics, to unique CRM issues, and on and on).

Worst of all, far too many stakeholders extrapolate broad feelings about AV UX from an early-adopting rider's first formative experience.  And as Strategy Analytics has found via independent research with consumers, mere exposure is not the only roadblock for shared AVs.

Here is one example of how bad assumptions regarding "mere exposure" can lead to erroneous conclusions:  From 2018-2019, Drive.ai (now an Apple company) conducted an on-road pilot test of low-speed AV (with safety driver) shuttles in Frisco, Texas USA.  Researchers from Texas A&M Transportation Institute found that after first exposure, Drive.ai shuttle riders enjoyed the "driverless" experience, and felt positively about future usage.  But then the contract with Frisco ended, and the shuttle ceased operation.  Why?  Because for some reason, shuttle ridership was extremely light.  First-time rider enjoyment (measured by one single timepoint of satisfaction for a limited set of users) did not translate to evangelism or repeat widespread usage.  And no follow-on research on repeat ridership has been publicly shared.

Lyft's representative framed the broader issue quite clearly during PAVE's webinar:  "As we look at what people are looking at in an AV versus a regular ride-share, it's mostly the same thing."  People want to get where they are going quickly and safely.  In other words:  If we strip away the self-driving "magic," focusing on a "first ride" is no different than focusing on a first ride in any taxi or city bus.  It's an important "back-of-book summary" for understanding the varied factors of a rider's UX.  But great care should be taken when reading that summary.  Because it is not the whole story.

Previous Post: COVID-19: The Fate of the Fearless | Next Post: UXIP Launches Podcast: UX Soup

Let's talk

Now you know a little about us, get in touch and tell us what your business problem is.
Name:
Email:
Telephone:
Country:
Inquiry / Message: