Testimony to joint committee on transportation on S.2379, “An Act relative to providing multimodal transportation technologies” (AVs)
Thank you to Chairs Arciero and Crighton and members of the committee for holding this hearing today. My name is Brendan Kearney and I am the Executive Director at WalkMassachusetts, a statewide pedestrian advocacy organization founded in 1990 as WalkBoston. I’m also sharing comments on S.2379, “An Act relative to providing multimodal transportation technologies,” which would authorize the use of autonomous vehicles in the Commonwealth.
Separately, I’ll be submitting written testimony against the sidewalk robot bills, H.3773/S.2372 (“An Act relative to mobile carrying devices”), and in favor of Rep. Vitolo’s H3804 to give pedestrians more walk time, and Senator Brownsberger’s S.2343 expanding truck safety requirements.
In 2024, there were 369 reported total traffic deaths in Massachusetts. At least 78 pedestrians lost their lives as a result of traffic crashes, accounting for just over 21% of the total.
It is possible that self-driving vehicle technology may help reduce that number, but there is no guarantee. This must not be rolled out without proper safeguards for our communities.
Since there are many unknowns with autonomous vehicle systems, we caution you to err on the side of more data sharing, local control, and consultation so that MassDOT, regional planning agencies, and local transportation departments will know more about vehicles operating on streets across the Commonwealth, and ensure we’re adapting them to our communities instead of the other way around – and so that municipal staff are able to respond to safety issues and resident concerns.
Incomplete crash data is a known issue. I encourage you to require data to be collected and shared not only during any testing phases, but also during deployment of autonomous vehicles. Please lean on the team at the MassDOT Safety Division who maintains the IMPACT Crash Portal to get their feedback on what information would be most useful to be collected and shared.
A few data points AVs could provide beyond traditional crash data might include:
- When and where vehicles are deployed to potentially understand the weather and road conditions the vehicles are driving in.
- Close calls. These might include hard braking incidents (when the AV had to suddenly stop), disengagements (when the AV requires a safety driver to take over) and immobilization (when the vehicle does not have a safety driver and is stuck).
I’ll share two concerns to keep in mind with the software:
- 1st/ Intentional lawbreaking written into the code. In a Washington Post article in December titled “On roads teeming with robotaxis, crossing the street can be harrowing” their tech columnist in San Francisco captured videos of Waymo self-driving cars failing to stop for him at a crosswalk. He sought to answer the question, “How does an AI learn how to break the law?” A spokesperson for Waymo told him that its car might decide not to stop if adjacent cars don’t yield. “So is it possible that Waymo’s AI is learning from the human drivers on the road who also act like jerks?” One of the theories posited is that Waymo cars were intentionally getting more aggressive to help shed a market reputation for being slower than competitors.
It is unclear to me how this behavior – violating the law to yield to pedestrians at a crosswalk – should be allowed, or how it would be creating a safer street than human drivers. The intentional lawbreaking that has been on display in California has also included short term parking in crosswalks and bike lanes – not pulling over to pick someone up, but remaining there for upwards of 5-10min. Communities and the state need to be able to collect data on that and have the ability to regulate it to hold companies accountable for these software choices. - 2nd/ Bias in the software itself. A study a few years ago from Georgia Tech found that “the facial and body recognition technology built into many pedestrian detection systems does not recognize and react to darker-skinned people as consistently as it does lighter-skinned people.” Already, people of color are disproportionately represented in fatal crashes involving people walking. The technology should level the playing field, not reinforce existing shortcomings and biases.
Thank you for your time, I appreciate the opportunity to comment.
This testimony was shared as part of a panel w/ Pete Wilson of Transportation for Massachusetts (T4MA) and Seth Gadbois of Conservation Law Foundation (CLF) at a Joint Committee on Transportation Hearing on June 24, 2025.