The challenge of robotaxis for the poor

While I’m very excited about the coming robocar world, there are still many unsolved problems. One I’ve been thinking about, particularly with my recent continued thinking on transit, is how to provide robotaxi service to the poor, which is to say people without much money and without credit and reputations.

In particular, we want to avoid situations where taxi fleet operators create major barriers to riding by the poor in the form of higher fees, special burdens, or simply not accepting the poor as customers. If you look at services like Uber today, they don’t let you ride unless you have a credit card, though in some cases prepaid debit cards will work.

Today a taxi (or a bus or Uber style vehicle) has a person in it, primarily to drive, but they perform another role — they constrain the behaviour of the rider or riders. They reduce the probability that somebody might trash the vehicle or harass or be violent to another passenger.

Of course, such things happen quite rarely, but that won’t stop operators from asking, “What do we do when it does happen? How can we stop it or get the person who does it to pay for any damage?” And further they will say, “I need a way to know that in the rare event something goes wrong, you can and will pay for it.” They do this in many similar situations. The problem is not that the poor will be judged dangerous or risky. The problem is that they will be judged less accountable for things that might go wrong. Rich people will throw up in the back of cars or damage them as much as the poor, perhaps more; the difference is there is a way to make them pay for it. So while I use the word poor here, I really mean “those it is hard to hold accountable” because there is a strong connection.

As I have outlined in one of my examinations of privacy a taxi can contain a camera with a physical shutter that is open only between riders. It can do a “before and after” photograph, mostly to spot if you left items behind, but also to spot if you’ve damaged or soiled the vehicle. Then the owner can have the vehicle go for cleaning, and send you the bill.

But they can only send you the bill if they know who you are and have a way to bill you. For the middle class and above, that’s no problem. This is the way things like Uber work — everybody is registered and has a credit card on file. This is not so easy for the poor. Many don’t have credit cards, and more to the point, they can’t show the resources to fix the damage they might do to a car, nor may they have whatever type of reputation is needed so fleet operators will trust them. The actions of a few damn the many.

The middle class don’t even need credit cards. Those of us wishing to retain our privacy could post a bond through a privacy protecting intermediary. The robotaxi company would know me only as “PrivacyProxy 12323423” and I would have an independent relationship with PrivacyProxy Inc. which would accept responsibility for any damage I do to the car, and bill me for it or take money from my bond if I’m truly anonymous.

Options for the poor

Without the proxy, robotaxi operators will want some sort of direct accountability from passengers for any problems they might cause. Even for the middle class, it mostly means being identified, so if damage is found, you can be tracked down and made to pay. The middle class have ability to pay, and credit. The poor don’t, at least many of them don’t.

People with some level of identity (an address, a job) have ways to be accountable. If the damage rises to the level where refusing to fix it is a crime at some level, fear of the justice system might work, but it’s unlikely the police are going to knock on somebody’s door for throwing up in a car.

In the future, I expect just about everybody of all income levels will have smartphones, and plans (though prepaid plans are more common at lower income levels.) One could volunteer to be accountable via the phone plan, losing your phone number if you aren’t. Indeed, it’s going to be hard to summon a car without a phone, though it will also be possible using internet terminals, kiosks and borrowing the phones of others.

More expensive rides

A likely solution, seen already in the car rental industry, is to charge extra for insurance for those who can’t prove accountability another way. Car rental company insurance is grossly overpriced, and I never buy it because I have personal insurance and credit cards to cover such issues. Those who don’t often have to pay this higher price.

It’s still a sad reality to imagine the poor having to pay more for rides than for the rich.

An option to mitigate this might be cars aimed at carrying those who are higher risk. These cars might be a bit more able to withstand wear and tear. Their interiors might be more like bus interiors, easily cleaned and harder to damage, rather than luxury leather which will probably be only for the wealthier. To get one, you might have to wait longer. While a middle-class customer ordering a cheap car might be sent a luxury car because that’s what’s spare at the time, it is less likely an untrusted and poor customer would get that.

Before we go do far, I predict the cost of robotaxi rides will get well below $1/mile, heading down to 30 cents/mile. Even with a 30% surcharge, that’s still cheaper than what we have today, in fact it’s cheaper than a bus ticket in many towns, certainly cheaper than an unsubsidized bus ticket which tends to run $5-$6. Still my hope for robotaxi service is that it makes good transportation more available to everybody, and having it cost more for the poor is a defect.

In addition, as long as damage levels remain low, as a comment points out, perhaps the added cost on every ride would be small enough that you don’t need worry about this for poor or rich. (Though having no cost to doing so does mean more spilled food, drink and sadly, vomit.)


Over time, fortunately, poor riders could develop reputations for treating vehicles well. Build enough reputation and you might have access to the same fleet and prices that the middle class do, or at least much cheaper insurance. Cause a problem and you might lose the reputation. It would be possible to build such a reputation anonymously, though I suspect most people and companies would prefer to tie it to identity, erasing privacy. Anonymous reputations in particular can be sold or stolen which presents an issue. One option is to tie the reputation to a photo, but not a name. When you get in the car, it would confirm you match the photo, but would not immediately know your name. (In the future, though, police and database companies will be able to turn the photo into a name easily enough.)

Poor riders would still have to pay more to start, probably, or suffer the other indignities of the lower class ride. However, a poor rider who develops a sterling reservation might be able to get some of that early surcharge back later. (Not if it’s insurance. You can’t get insurance back if you don’t use it, it doesn’t work that way!)

It could also be possible for the poor to get friends to vouch for them and give them some starter reputation.

Unfortunately, poor who squander their reputation (or worse, just ride with friends who trash a car) could find themselves unable to travel except at high cost they can’t afford. It could be like losing your car.

The government

The government will have an interest in making sure the poor are not left out of this mobility revolution. As such, there might be some subsidy program to help people get going, and a safety net for loss of reputation. This of course comes with a cost. Taxes would pay for the insurance to fix cars that are damaged by riders unable to be held accountable.

The alternative, after all, is needing to continue otherwise unprofitable transit services with human drivers just for the sake of these people who can’t get private robocar rides. Transit may continue (though without human drivers) at peak times, but it almost surely vanishes off-peak if not for this.

Governments could also simply force this on operators as part of the cost of doing business, effectively making regular riders subsidize the higher insurance rates of the poor.


A more serious concern of fleet owners might be non-accidental damage, ie. vandalism or even theft. In such rare events cars could move into a high-surveillance mode, beaming video of everything back home. This is something the police might investigate, and could cause the piercing of a privacy proxy service. (The most common question we get at Starship Technologies is how the small delivery robots will deal with theft or vandalism, though in reality it is far from the biggest issue.)


Some will propose solving this with surveillance. Making it clear to riders that they are under surveillance, mitigated by the mandatory destruction of the video after a day has passed. Sadly, most businesses and governments like to retain data so it’s hard to accept promises it really will be destroyed, and even so, people would feel much less free in general in the vehicles. Who hasn’t engaged in something private in a car? I suspect a large fraction of people have had some sort of sex in one. There are a lot of other activities that would be self-censored.

Shared vehicles and crime

The question gets more serious with shared-ride vehicles. Today, with things like UberPool, you have both a driver in the car and the knowledge that your seat-mate is paying with a credit card and is in theory identified. That could be an issue in a car with no driver and the risk of harassment or crime. We’ve often seen people get unruly on a bus (especially late at night on the weekend when people are drunk and being responsible by riding the bus.)

In spite of the fact that assaults in Uber are rare, there is a great deal of press and concern about them. (This is in part because the Taxis that compete with Uber want to convince people that riding in it is much more dangerous than monopoly licenced taxi.)

There are a few obvious options here. Riders can have a panic button which invokes surveillance or even a videoconference with an operations center. The remote party can’t physically intervene but can stop the vehicle, call police and speak to the offender, as well as record video and any identity information.

Vehicles designed for sharing with untrustable parties can also have dividers. While not offering perfect security — you also need to be able to open the dividers to get out the other side in case of an accident — they can improve the privacy and security quite a bit. If the other rider is a risk, you are still vulnerable getting in or out, even if you can lock down the divider.

Shared transportation is important for the poor as it offers the lowest cost option, and it’s also important at rush hour. We might see a world, however, in which the middle class refuse to share vehicles with people who are less accountable or have no reputation for reliability. Today, you often see the homeless on transit, but many on the transit wish they weren’t there. It would be a strange reversal if that wish were translated into a world where nobody will ride with the homeless or poor and they have to ride alone — at higher cost.

Get a criminal record and your chances of riding with other people will go way down. Indeed, get any negative reputation and you may ride only with other people who have no choice about it.

In general, when not pushed into it by factors at rush-hour, the wealthy will stick to private robotaxis or luxury shared robotaxis populated only by those of similar income.


People will be particularly paranoid about their children. They won’t want them riding with untrusted strangers, certainly not with anybody of negative reputation. This might limit the ride options of the children as much as the people with the negative reputations. Children below a certain age also need adult supervision, performed to one degree or another by the bus driver on things like school buses. This presents another challenging problem.

Race & Pickup

In a reversal of this, the news is generally good for reputable people of minority races. Today, they find it hard to get a taxi in many places. Barack Obama has reported his experience in being unable to get a cab before he was President. One thing minorities love about Uber is they always get a ride. Robotaxis will be even more non-discriminatory. At the same time, passengers may not be, and because some races are poorer than others, on average, the factors listed above could end up disproportionately hurting certain ethnic groups. Of course, discrimination on the basis of race and similar factors is illegal in most places, but that doesn’t stop it from happening.

People also like that Uber comes even in poor neighbourhoods, something that many taxis don’t. Compared to taxis it has been a win for the poor and robotaxis could be, too.

What solutions do you have for providing low cost and equitable transportation to all in the world of robotaxis and shared self-driven vehicles?

I'd recommend you actually

I'd recommend you actually try riding public transport in a "poor" part of the world. Ride it seriously rather than as a pop-in-pop-out tourist. Because the foundation of this article is so vastly at odds with what actually happens that I can't think of where to begin. At the most fundamental level, the people who have a need to get from one place to another and are willing to pay to have that need satisfied have almost nothing in common with your conception of the vandalizing poor. But that's just the tip of the iceberg.

Of course not

And if you got that impression, indeed the article needs some revision. This article starts not with the assumption that the poor are particularly likely to damage or vandalize the cars, but rather with the assumption that the owners of the taxi fleets will ask themselves, “What will I do if a customer damages my vehicle?” Rental car companies of course ask that question all the time. They want your identity (they get a right to ask for driver’s licence, credit card) and they want to sell you grossly overpriced insurance.

So then they ask the question, “OK, if the customer is poor, such that in the rare event they damage the vehicle, I have no real way to get recourse from them, what can I do?” And they will ask that question. We don’t want the answer to be, “don’t let the poor ride” though in fact that is the answer for Uber, which has no riding unless you are identified and have a credit card. We want another answer.

Same on vandalism. Of course very few will vandalize or commit crimes. But as we have seen, that has not stopped large campaigns of publicity around rare sexual assaults and other assaults within Uber. (Mostly by drivers, I have not heard many reports yet on UberPool seat mates, but we will.)

I will revise the article to make that more clear. Yes, obviously when I have ridden transit in poor parts of the world it has been as a tourist, but this has little to do with the issues above.

Free rides

All this is a non-problem in the US. We subsidize the poor at every turn -- welfare, school meals, nutrition programs, public defenders, etc. etc. Now we provide free phones and,soon, free internet. Why not free rides as well? Identity is already provided by the ubiquitous EBT ("food stamps") card. Trash the car and lose value on your EBT card. Of course, at that juncture, various community organizers will be demanding amnesty.

A depressing analysis

A depressing analysis


What you describe sounds a lot like insurance redlining. I am also wary of giving a transportation company pre-authorization to charge individual customers for it's idea of damage or cleaning. We've already seen abuses of this with Uber (e.g. drivers allegedly faking vomit to get the $250 cleaning fee). Pretty soon we'll be being dinged for $5 "mud on floor" charge and $10 "wet seat from raincoat" charge, etc. And don't dare dispute the charge, or risk getting blacklisted. A much more fair solution would be to treat it as ordinary wear and tear, cost of doing business, and build it into the price for everyone.

Wear and tear

It’s an option. In this case, there is no driver to try to defraud you. One approach I describe has a camera that takes pictures before you enter and after you leave the vehicle. If it’s damaged/soiled, there is little doubt on who did it. Mainly this camera will detect when you left something in the car.

As long as vandalism is not a problem, the approach of just charging cleaning/damage to everybody baked into the price could work. If it is not that bad a price, it would be OK. Of course, as noted, one common problem will be people spilling food or throwing up. These things happen more with certain people, and throwing up mostly in the early hours of weekend mornings when picking up for bars. But we definitely want the cars to pick up from bars, though they might send cars with easy to clean interiors.


Why not just outfit a few robo-taxis for bar pickup duty? Already some police cars have hard fiberglass back seats. Build a completely non-porous interior. When necessary the car can swing a company-owned robotic cleaning facility -- wash, steam clean and dry the interior, then send it out for the next episode.

This is relatively not a problem.

If cars can drive themselves, it would be an epic failure if they couldn't simply clean themselves as well. They can drive themselves to a special facility where a robot who would know enough to notice or charge someone for the mess would know enough to clean it up. Would you rather design an environment/system for cleaning vomit or a replacement for the human mind while driving? I think the latter technology implies the former.

Yes, they will

Deeper in the links you will find this exact concept. But what I’m talking about is the same pressures that make car rental companies not readily rent to the poor (or even those without credit cards) and which make taxis not go to poor areas when called — that those pressures might cause robotaxi companies to not serve the poor at the same prices the wealthy get.

Maybe fixed by more data

Perhaps the global cost of cleaning up after the small percent of riders which leave problems will be just factored into the cost of the ride. Maybe riders from bars, for example, incur more cleaning problems and that could very dynamically be structured into the quoted price, i.e. a premium based on something like a bar district at closing time. There should be a lot of data to find such connections. Of course if travelling to/from a poor part of town sees the same kind of liability, it's back to the point you were making.

It is true

That with a rental car, where we see obscenely priced insurance and requirement for a credit card with a decent credit limit, you can do a lot of damage — you can total it. A rider can’t do nearly as much.

On the other hand, if I were a fleet owner I would see a the argument that if people are not responsible, ie. “cleaning is included in the price,” that they will make more messes and other damage — ripping upholstery, permanent stains etc. So there is an argument for including cleaning in the price, and another argument for not including it. This will vary with the type of interior. If it’s like a bus interior, you would expect cleaning to be easy and in the price. If it’s a fancy Mercedes with leather or nice cloth, it could be a different story.


Thanks Brad, although it is a bit unclear whether you are primarily considering the poor in developing countries (as suggested by the picture) or low income areas in developed countries. Either way, a couple points.

While it would be wrong to photograph people while they are in the vehicles, it may be acceptable for an external camera over the door to record a picture as they approach and enter while still in public. This is what dash cams are already doing. It seems a fact of life that anonymity is going to steadily decrease in society at the commercial level as payment systems continue to head towards identifying the purchaser as part of the cashless society. We may have to accept that much of what we do is recorded and put privacy efforts into restricting who has access to the information and insisting on an audit trail being left by all those trusted to view it.

It may also be possible that individual vehicles themselves could be based in certain locations so the neighbourhood they primarily serve feels a degree of "ownership" over them. This could apply especially in developing countries where people living in relative poverty do look out for each other; and arguably more than those in wealthy areas. They would put natural pressure on anyone vandalising their neighborhood vehicles and tend to often know who the troublemakers are.

Also Brad: There currently is much debate whether robo taxis will reduce congestion or make it worse? Perhaps in a future article you could give your thoughts on this and whether government incentives could be used to encourage road pricing to advantage shorter travel distances (my own thought bubble). It seems that much of our traffic problems stem from the choices people make when choosing jobs, restaurants, etc that involve traveling well outside their local area.

By all means the heavy hand

By all means the heavy hand of government should be employed to keep people from "traveling well outside their local area".

Poor Bus

"An option to mitigate this might be cars aimed at carrying those who are higher risk."


a special "higher risk" car showing up is a giant signal to everyone in the universe that DIS BROTHA BE PO AS FUCK and probably will not go over well with potential users of your service.


This is what we want to find a way to avoid.

The challenges of RoboTaxi services for the poor

Brad, we worked on these issues for Denver's Smart City Challenge application ($40M US DOT grant). A key objective was enabling services for low-income neighborhoods since for many, reliable transportation is a gateway to education as well as a better-paying job. Part of our solution was neighborhood kiosks and free Wi-Fi availability since you can buy a no-contract basic smartphone for $10 these days at some big box retailers. But a key part is expanding city-issued MyDenver cards used for Rec centers, etc, so that it can be used for public transit, Denver Bike Share, Lyft, Car2Go, ZipCar, Flitways, and ParkWhiz. We'll use Xerox's GoDenver app for multimodal scheduling. All this will be done with an eye on building the infrastructure and business models needed for RoboTaxi services as soon as they are available.
For those with no credit, Denver will partner with PayNearMe (available in 468 Denver storefronts) to load cash credits. Until a better solution can be found, Denver will pay the transaction fees. While this is the "strawman" plan, the final plan will be the product of an extensive series of open meetings with the communities of interest. This won't be a case of "we're from Denver government, and we know what you want and need." No doubt the plan will evolve after engaging the folks we are trying to help.
We certainly haven't addressed all the issues you raise, Brad, but it's a start, and will need to evolve. But if we don't try, we'll never learn.
And thanks for your insightful blog. It must consume a lot of your time.
BTW, great job on Dateline. You obviously never learned how to pull punches... which is good!

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

Post new comment

His name is Brad Templeton. You figure it out.
Please make up a name if you do not wish to give your real one.
The content of this field is kept private and will not be shown publicly.
Personal home pages only. Posts with biz home pages get deleted and search engines ignore all links
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd>
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.

More information about formatting options