Draft: Waymo shows Autonomous Taxis Are Real. Let's Chat About Policy to Incentivize Carpooling.

Like many, I wish the transit options in the US were better, but now autonomous vehicles offer a solution if we can manage it carefully.

Here we're going to explore a few ideas.

• I'll first try to establish that autonomous vehicles (AVs) are no longer scifi, but are truly ready to be part of planning and discussion (Section 1).
• I'll lay out what I view as the key challenge of this new AV world: how to seize the opportunity to vanquish the single occupancy vehicle in a way everyone can love. (Section 4)
• We have the resources for interesting policy here. But it might take some reckoning and compromise on whether projects realistically make sense with AVs in the equation. (Section 4)

Some of this is a bit contentious, but I think it's interesting and important to think about.

Autonomous Vehicles are Here Now

How real today automous vehicles seem greatly depens on where you live. In certain places like in San Francisco or Phoenix, Arizona, you can just download an app, tap a button, and a car without a driver will appear. Just one leading one leading company, Waymo, from the same parent company as Google, gives about million rides a month.

In late 2025, Waymo began rolling out highway driving, which dramatically opens up the kinds of routes possible^{[citation needed]}(image of highway new map).

Even in these places though, it's still easy to forget this is even possible. If you already own a car, you will likely still drive. If you live next to a transit stop and your destination is next to a different transit stop, you'll might still transit. In other cases, an Uber or Lyft is often still cheaper -- at current supply, Waymo can still charge a slight premium for the novelty, safety, and better experience. However, costs of AVs are likely to fall.

Autonomous vehicles can dramatically improve transportation for the better. However, a possible version of the AV deployment is we end up with something too similar to today, with heavy traffic and high costs leaving much of the vehicle vision on the table.

What Do We Really Want From Transit?

Before discussing policy it can be useful to reflect on what research says about what people want from transit.

Fast

We want our transportation to conveniently get to our destination quickly and without struggle. However, not all parts of the journey are created equal. Research results vary here, but loosely waiting is roughly 1.5x-2x worse than riding in a vehicle ^{[citation needed]}. Walking is estimated about 1.5-2.0x worse. More detailed research will also show the various costs people place on transfers (they're bad) or reliability (the variance in arrival time. Also bad.).

Under this model we can reason about why people frequently choose to own a car or Uber/Lyft. A trip that has a 10 minute walk, a 10 minute wait, and a 10 minute bus ride is in total a 30 minute journey, but with the waiting and walking it might not be preferred over an instant 45 minute vehicle ride right at your front door. A single 45 minute ride is predictable and makes progress towards your destination the whole time.

Additionally that 10 minute wait might hard to predict. A key piece of jargon here is "headway", which is the time between buses/trains on the line. A headway of 5 minutes you can just show up without planning or worry. A 1hr headway is challenging to use.

An autonomous vehicle that can quickly appear nearby can simply be the fastest transpertation possible without dramatic infrastructure change.

Cheap

How much we are willing to pay for various amounts of speed and convenience depends on a lot of factors. People value work time, more than commute time, and more than leisure time.

Broadly car cost is more expensive than many realize as costs get split between purchase price, maintaince, insurance, and fuel. In the US, a car owner spends about $X on transportation. About200billion is spent on travel on

Meanwhile rail can be increadibly challenging. Typical rail projects in the US cost over $100 million per mile^{[citation needed]}.

Safe

Green, Equitable, Pleasant, etc

We won't get to every factor here, so group a couple of factors to gesture that there is more we want from transportation than just fast, cheap, safe. We want new transit options to be green but as consumer electric vehicles show, green won't be a deciding factor for many, we want them to provide equitable access across the community, and we want them to be pleasant, among other things. AVs can have significant advantages here compared to current transportation. They can be efficient electric vehicles, can go anywhere without unequitably deciding something like where to put the tracks, and can be enjoyable to ride. Still there's nuance here which influence policy.

Driving is Fun, but Traffic Sucks.

What this Means for Policy

"Bus good, Train Bad?"

There's a maxim usually traced to Harvard economist Edward Glaeser of "bus good, train bad"which he refers to as a slight characture of a few Harvard economists views (Glaeser, 2009). Long before AVs, there has been a view that if we can put aside some of the romanticism of the train and the stigma of the bus, the economics and route flexibility of buses beat out trains.

This is hard to swallow. I like trains. I really do. In the Bay Area, rail is frequently part of my transporation (via BART, CalTrain, or MUNI), and I've enjoyed light and long distance rail in the US and around the world. Meanwhile, my time with buses feels more pragmatic. There were childhood school buses and then about a two year time where for my commute in Austin, TX was either a bus or a bike ride depending on the weather/mood. Buses got me somewhere, but didn't fill me with wonder like going to a grand century-old train station to be smoothly whisked away above (or perhaps below) the streetside commotion.

Yet, autonomous vehicles make the "bus good" argument too overwhelming to ignore. Investments made in bus routes and roadways now are prime to leverage a future change to autonomous buses. The greatest cost of most bus routes is often not the bus itself, or the fuel, but instead the driver^{[citation needed]}. AV progress can be a massive value multipler on bus infrastructure while giving less extra value to rail. Additionally AVs are a new paradigm where we can likely replicate some of the pleasant joys of train travel.

Beyond the Single Occupancy Robotaxi

As discussed in Section 3, traffic is not a problem for others.

Roughly 70% of car trips are single occupancy with just the driver (Federal Highway Administration, 2022). There is tramendous opportunity to nearly double the capacity of our road networks and half traffic levels if those rides could be converted to carpools of two or more.

So far the numbers for autonomous vehicels are looking to be no better. Similarly 70% of autonomous vehicle rides were single occupancy, and in a new addition 5% had no driver at all^{[citation needed]}. Tesla has embraced this statistic, designing their robotaxis to only two seats.

We should sieze o

Conjestion Pricing

What if we just paid people to carpool?

Engineering for Shared Rides

Thinking Harder on Megaprojects

Embracing New Infrastructure Needs

• Pickup and dropoff bays
• Lockers or storage belts
• Street charging

Interactions with Labor Policy

Autonomous vehicle policy might seem to interact labor policy. There has been cases of labor unions fought to add regulations to AVs that would slow their deployment^{[citation needed]}. However, in many ways these are fairly seperate issues.

The problem of job displacement is deeply troubling which I don't have clear thoughts on. Instead offering answers, I'll paradoxically point out the problem is actually much worse than it might seem. This isn't exactly solice, but does recontextualize the problem into something we can work on independentally on the question of of how to regulate robotaxis.

Using just American data here,about X million work as drivers. Roughly X million workin jobs connected to cars (car dealers, car insurance, gas stations, and many more), which will look very different in a shift away from individual car ownership. This is multiple percentage points of the workforce that gets disrupted by AVs.

How ever the problems of artifical intelligence and automation is far broader than just vehicles. Estimates very, but likely X% of jobs at risk. On a macro level, automating driving is a comparative safe thing for society. Compared to very risky forms of automation (compared to say the more general creators or deciders, autonomous military commander, biologist, or computer hacker). Additionally, driving is not necessarily an idealistically human endeavour in the same way as say artist, musician, or pastor, which we might more clearly want to normatively say shouldn't be automated, even if machines can do tasks well.

We need options for job displacement for providing public support and options. But this can be thought of as a critical broader and parallel problem to also celebrating our opporunities to limit single occupancy vehicles and dramatically lower the costs of transportation.

Conclusions