The Societal Case for Robotaxis

What if the AV industry sold on better cities instead of wishful thinking?

Last week I wrote about the Greenwashing of Robotaxis. In that post I pointed out that claims of robotaxi safety improvement, improved equity & accessibility, reduced congestion, and reduced energy/emissions were questionable. I still believe that if gears in the areas of company business plans, regulatory capture, and so on keep spinning as they are, we’ll end up net worse off than we are now in all those areas.

Not that it has to be that way, but rather that is where we’re headed.

Perhaps there will be some areas that improve a bit. But on the whole most of the promises will turn out to be more of the same mixed bag we got from ride hail, or different problems, or even worse problems. #1 on my list is exacerbating the downward spiral of some public transit services due to fare poaching by robot-ride-hails that are cheaper than human ride hail trips. This not only hurts those too poor to afford even a cheaper ride hail, but is also likely to make net transportation safety worse (because car trips tend to be more dangerous than mass transit trips).

And if we have the One Bad Crash headline amidst an atmosphere of distrust and a history of deflating hype with a side of deception (because, honestly, that is the hole the industry is trying to dig out of right now, regardless of which companies claim to be better actors than the others (*) ), we could lose any benefits we might have gotten. (*) Note: don’t forget Tesla proclaims itself to be a robotaxi company, regardless of what you might believe about the credibility of that claim.

What if — and I know this is crazy talk — the robotaxi industry decided to make its case to public stakeholders on actual benefits delivered to communities? Benefits that they are demonstrably standing up a business model to deliver in an accountable way?

First, let’s talk about the need for the industry to pivot hard away from the “saving lives” message. Then get to positive messages (and actions to back the words) that might be used instead.

Stop promising saving lives

Saying the #1 reason for forgiving all robotaxi operational sins is Saving Lives is, to put it bluntly, the worst possible path right now. It will be years before we know how fatality rates turn out for even one company, and no basis for crediting any other company with that outcome. There are good reasons to believe matching human driver safety for severe injuries and fatalities will be a struggle, especially when robotaxis start driving above 20 mph, which is when pedestrian fatalities increase dramatically. And let’s face it: pedestrians and other vulnerable road users are the ones being put at the highest risk here, not the robotaxi riders themselves.

Don’t get me wrong! I hope robotaxis will be safer. And they might eventually be safer. But even if they are safer we won’t have credible data to show that is true for years (perhaps a decade, depending on scaling rate). Meanwhile, every scary headline degrades trust in the industry.

It’s simple: people over-trust too soon, and backlash too hard just as quickly after adverse news. We’ve been through a full cycle (the Uber ATG fatality in 2018) and are mid-way through another cycle of distrust of vehicle automation safety (the Cruise pedestrian dragging in 2023). The drumbeat of Tesla safety issues, injuries, and fatalities involving AP/FSD is not helping. Saying that the industry has turned a corner and there will be no further such cycles is beyond wishful thinking, and into full-on delusion. So is doubling down on the tired playbook of shouting “we’re saving lives” when your own researchers publish papers that say way mo’ data is needed to know how fatality rates will eventually turn out.

Yes, documenting safety outcomes is important. But it is a weak selling point. The strongest credible message supported by data is a lackluster “not unsafe as far as we can tell so far,” which is a far cry from the “saving lives” claim their hypemasters and regulatory capture folks are pushing. Remember, Cruise kept saying they were multiples safer than human drivers — right up until they caused a severe injury due to a mistake (not remembering a pedestrian the car had just hit was still under their vehicle). A mistake their own investigation admits a human driver would not have made. Turns out that potentially negligent driving and failure to mitigate some specific risks can trump net statistical safety.

The safety message instead should not be “saving lives.” It should be a promise to do no harm (on a relative basis). Specifically, (1) be no worse than human drivers on average, (2) be extra-attentive to cleaning up any problems that do manage to happen regardless of blame, (3) provide accountability that they are actually doing the first two things. That arguably takes safety off the table in the short term so they can instead emphasize tangible benefits delivered in the here-and-now.

Promise transparent continuous improvement

The industry should get rid of the traditional automotive playbook that prefers to hide everything possible, blame drivers, blame victims, blame regulators, and then argue any safety improvement is too expensive. Seriously, this is automotive industry government relations 101, lecture 1, going back decades. I know that is a tried and true playbook for conventional vehicles, but it didn’t work out so well for Cruise, and it’s not going well for the industry overall because the don’t have a driver to blame but themselves.

I think more backlash is already winding up. We’ll see what happens in response to the next adverse event. Because driving has risks no matter how good the technology is. The next big crash is coming. Perhaps at an appropriate statistical time, but perhaps sooner. The question is whether the public will trust the crash happened in good faith, or will distrust an industry that stubbornly keeps operating according to its trust-degrading playbook.

Enough of playing safety messaging defense after crashes! Instead, play safety offense with a conspicuous display of transparency and commitment to continuous improvement. Don’t sell on saving lives. Say something more like in the far future it might save lives, but here are the benefits we can provide for you right now.

The industry needs to be seen as a trustworthy actor to make sure safety is really a top priority rather than being lip service. Perhaps safety really is #1 at some companies. But pretty clearly it has not been really #1 at all of them. It is the industry’s job to convince us, not our fault for being skeptical after what we’ve seen play out on the public stage.

After the next minor incident make a big deal of how the Safety Management System responded, the new test cases added, and so on. And do that for every incident (for example everything reported to SGO, but eventually even things that are not reactive to an embarrassing news story). Build trust via showing that the goal is safety improvement, not blaming others for mishaps. When there is embarrassing social media, take it in good humor and explain how (substantively) you’re fixing the problem. Don’t blame an “improperly towed” vehicle for getting hit by not one, but two of your robotaxis on the same tow truck trip (sheesh!). Own the imperfections and talk about the journey to improve; don’t try to deflect or deny them. And don’t lard an incident discussion with disinformation and marketing hype. A display of transparency does not work when it is sandwiched by weasel words, blame, and platitudes so obviously written by lawyers and publicists rather than technical safety experts. Because when you do that, you convey the message that every mishap is a negative news cycle to be escaped rather than a safety incident that needs to be addressed.

That way when something more severe happens and it really is not the company’s fault, people have enough trust to believe the explanation. The industry’s behavior patterns have degraded public trust tremendously in the last half-decade. And it is largely a self-inflicted wound. Time to change the approach.

I think we’ve seen enough that we should assume a lack of published metrics (or gamed metrics) means the industry is hiding a lack of safety. No news is not good news on safety metrics. Claims that safety outcomes are proprietary are really about trying to remain opaque about safety. (Tesla demands that NHTSA redact all the substantive fields in its crash reports released to the public. 2000+ records in the SGO database, and counting. Other companies are not as overt, but there is plenty of hiding, shading, and other withholding of data going on.)

Be held accountable for benefits

So you want to provide transportation to the disabled? Great! Where are the wheelchair ramps? I mean the real ones on vehicles serving actual people, not the prototype on the robotaxi model that has since been discontinued. To their credit Waymo has manually driven wheelchair-accessible rides, which shows great intentions, but also shows how far this technology has to go to meet the mobility promises the industry keeps making. Robotaxi funding subsidizing human-driven WAV rides does help those riders who need the service, but not at all the same as fulfilling the promise that AVs will offer independent movement to that community. For now AV accessibility is far more of a promise than a reality. Promises are great. Where are the progress metrics?

Publish goals for weekly rides for different communities with special access needs and chart how you’re growing to meet them. Otherwise it’s just marketing hype. Loss leader services won’t survive a market shake-out. And without a PUC mandate (which is not in place), those WAV rides could all-too-easily evaporate once market capture is completed and human-driven ride hail services mostly evaporate.

You want to reduce congestion? How exactly is that going to work? Why should we believe it will be any different than ride hail (which seems to have made things worse)?

Do you have an amazing idea to promote ride sharing? (Perhaps segmented, isolated luxury pods in a commuter AV minibus? Or just a front/seat back seat scheme with a hard divider — which won’t work on the bread loaf vehicles?) Do you have a routing algorithm that reduces the congestion induced by ride hail vehicles hunting for fares? Do you have a plan for not clogging streets with robotaxis that get confused and lose cell data connectivity? Show us the benefit, show us metrics for those benefits, and show us your progress.

You want to provide better safety for passengers who feel unsafe traveling with strangers? OK, that’s great, but make it a holistic safety experience, not just a talking point. Don’t dump women out in a scary neighborhood by ending a ride early or getting the drop-off point wrong. Don’t sell mere rides — sell the safety as the product. Again, where are the metrics? Where is the progress?

And what exactly is the story for ride hail and taxi drivers displaced by robotaxis? Because obviously the whole point of this exercise is to put them out of a job. (Or if I’m wrong, what is the story, and what are the metrics to show it’s working?)

A simple ask by state & local regulators is: “What KPIs are you using to measure benefit to the community?” Regulators should ignore any claim without a KPI as mere puffery — and demand reporting of KPI progress in exchange for any regulatory concessions. If company X is claiming jobs and economic benefits in exchange for using public roads as a free testing space, where are the public quarterly reports of delivering on those promises? And for proof that negative externalities have been limited?

If companies don’t have community benefit KPIs, their management won’t be incentivized to benefit the community. So you won’t get it. (It will be even worse if the KPIs are community acceptance instead of benefits — because that is a PR goal, not a benefit goal.) If they won’t reveal how they are doing toward improving KPIs, then we should not assume the news is good.

Some crazy ideas and metrics

Here is a mix of crazy ideas and metrics. Some will require business model changes — which is the point! Apply the proper ratio, scaling, or other details to make these the right metrics. These are ideas and topics for metrics, not mathematical proposals for exact measurement formulas. For some up is good; for others down is good. I’m sure you can figure out which if you live in a community with robotaxis.

• Rides that connect a passenger from/to mass transit

• Personal vehicle rides that switch to mass transit + robotaxi

• Personal vehicle rides that switch to robotaxi only

• Deadhead robotaxi miles & idle time in parking/marshalling areas

• Rides for people who do not have the ability to drive a car (can’t drive; no car)

• Shared rides (different pickup and/or dropoff points for passengers on one ride)

• Avoided rides (e.g., homebound leg avoided for a human driver after dropping someone off at work/school)

• Rides taken by vulnerable riders (e.g., women late at night; under-18; elderly)

• Food deliveries to locations more than walking distance to a full-service grocery store

• Riders with special accessibility needs served by compatible equipment in an actual AV (not a manually driven compliance ride) (wheelchair ramp, vision-impaired interface, large grocery load, child seat)

• Congestion increase/decrease

• Energy reduction compared to alternatives (ride hail EV; public transit), accounting for data center energy used for machine learning training

• Conformance to best practices for serving special accessibility needs

• Adverse rider experiences (dropped at wrong location; stranded; demanded emergency egress, other distress situations)

• Adverse interactions with emergency responders

• Substantive traffic rule violations (red lights run, encroachment on pedestrians in crosswalks, etc.)

• Fraction of industry-standard SMS metric goals met (SMS = Safety Management System) and improvement strategy

• Costumer complaint groupings and progress on identified issue resolution

• Other road stakeholder complaint groupings and progress on identified issue resolution

• Resolution of hazards identified via near-miss events, with general event description placed into a shared industry database for continuous industry improvement

• Conformance to industry safety standards for automobiles, active safety systems, and autonomous vehicle features

• Metrics showing ride hail drivers getting better jobs in the robotaxi industry vs. being displaced by robotaxis

Even better metric topics are welcome in the comments! Not all of these are ones we should track in real life, but some probably all, and most if not all give starting points for discussions that need to be undertaken. The point is these are all about providing claimed benefits/lack of negative externalities, and not about the details of the technology itself. This list is demonstrating we can think about benefits in a measurable way. If you don’t like metrics at all, what is your suggestion for creating accountability beyond the robotaxi company leadership’s fiduciary responsibility to their shareholders?

There should also be safety-oriented metrics, but that is a whole different discussion.

I realize robotaxi companies will claim proprietary/business secret etc. But at some point if they want to earn back trust they’re going to have to have a plan. “We’re saving lives” is a sinking ship of a plan that is waiting to catch fire after the next high profile mishap.

Perhaps being transparent about real benefits instead of safety hype is the way to go.

Something I have not addressed is whether robotaxis can be a profitable business model. There are tens of billions of dollars betting it can. Perhaps. Perhaps in the end it is just a technology development platform for other opportunities. Perhaps we’ll find out we need to wait another technology generation. I’m willing to wait and see on that one — but only if the Grand Robotaxi Experiment is not fueled by avoidable societal harm.