Leadership Leadership Jun 14, 2022
Podcast: What One School District’s Fiasco Says About the Strengths and Limits of AI
On this episode of The Insightful Leader: When Boston Public Schools looked to algorithms to solve equity issues and save money, it ran into a roadblock—the complicated lives of parents and students.
In 2017, Sebastien Martin had spun up an algorithm that could help Boston Public Schools sift through a seemingly endless number of bus routes in search of the most efficient, less expensive option.
But what started as a monumental feat for artificial intelligence turned out to be a lesson in what happens when humans aren’t ready for its computer-generated solutions.
“I just see it as one of these extremely complex problems,” said Martin, an assistant professor of operations at Kellogg.
In this episode of The Insightful Leader, we look at what happened in Boston to better understand how leaders should think about using artificial intelligence to solve some of our thorniest social problems.
NARR: The traveling salesman problem. It’s a known computer science puzzle. And goes like this…ready? Given a list of cities and the distances between each pair of cities, what is the shortest possible route that visits each city exactly once and returns to the origin city? Essentially, what is the most efficient round trip you can take?
I know, it kind of sounds like something you’d do for extra credit in eighth grade math class. But actually….this is something that computer scientists have thought about for more than a century...because...trying to find the shortest round trip for just 15 cities...even...would mean sifting through 87 BILLION possible round-trip scenarios. And here’s the kicker: Every city you add...shoots the number of possible routes up exponentially higher.
Sebastien MARTIN: As the number of cities grow, the time it takes for a computer to find the very best path sort of explodes.
NARR: That’s Sebastien Martin — he’s an assistant professor of operations at Kellogg…and he specializes in using math and AI to make transportation systems as efficient as possible.
It turns out, Martin ended up building a machine that could quickly plow through a pretty incredible number of possible routes.
MARTIN: So yeah, the number of possibilities would be of the order of…ummm…
NARR: That’s the sound of Martin doing the calculations in his head. You see..this machine… could sort through 1 novemtrigintillion different route scenarios. That’s a one followed by 120 zeros.
This piece of technology….which he drew up with a fellow grad student at MIT…attracted attention. It won awards. It certainly caught the eye of Boston Public Schools…a district that had…a really hairy problem on its hands.
Tommy CHANG: We have the second highest transportation cost per student in the country.
NARR: That’s Tommy Chang. He was the superintendent at Boston Public Schools in 2017. And he’s speaking at a kickoff for a competition that BPS held that year. A kickoff competition for computer scientists to help put hundreds of buses on more efficient routes and save the district millions of dollars.
CHANG: Boston, as you all know, is a very difficult city to navigate, and it has a very high cost of living…and so, because of that, Boston Public Schools spends five times more per student than a typical school our size.
NARR: In 2017, when Chang was speaking, the district had more than 600 buses on the road, and having buses on the road cost money. The district was spending more than $116 million — more than 10 percent of its budget — on transportation.
And so…Martin, his MIT colleague Arthur Delarue and their advisor, Dimitris Bertsimas, put their computer to the task…to find more cost-effective routes….and they came up with an option that the district was happy with!
The whole thing was poised to put Boston Public Schools — and honestly, artificial intelligence — at the helm of an entirely new approach to designing transportation systems: one that could transform the way school districts across the country organized their very complex bus routes…
And Martin’s algorithm had another huge benefit: it could help them reveal and correct for racial and economic disparities that had long been baked into the existing bus routes…
PARENT MONTAGE: If I had known it was a possibility for my daughter to be forced out of the house from six 30 in the morning to six at night, I would have never enrolled her in the Boston public school system…Are you serious? My niece has autism. She has she’s autistic. She cannot function early in the morning…..Rest…rest assured…we are upset…and we are not going away [applause]
NARR: It came crashing down.
You’re listening to The Insightful Leader. I’m Laura Pavin. And in this episode, we’re going to do something a little different. We’ll take a closer look at what happened at BPS...and why parents, students and the surrounding community got so upset about a solution that, on paper, would have solved a LOT of pain points for the district .
So why are we talking about school bus routes in a podcast about strong leadership? Well, this particular story hits on a really big question…and that’s…whether AI alone is capable of solving our most complex human problems. Because—beyond school districts—companies…leaders…governments…police departments…are becoming more and more interested in using algorithms to help them find answers to really complex problems….like…workflow inefficiencies, economic inequality, and public safety. And that’s because, its proponents say, computers are unbiased and work faster to sift through tons of possible solutions that humans alone could not find in the same amount of time…or even in their own lifetimes.
The problem, though, is that a lot of the leaders looking to AI for answers don’t always do enough due diligence to consider those people they’re supposed to serve.
Today, we’ll look at what happened in Boston….we’ll look at how Martin’s algorithm was used to attempt to solve the district’s pain points…what leaders did right…and where they might have fallen short. Within this story…are lessons for leaders on HOW to think about using artificial intelligence to solve some of our thorniest social problems.
MARTIN: Okay. Can you see? Okay. So what you have here on the right is the map of the school district.
NARR: That’s Sebastien Martin again. We’re on a Zoom call, and he’s showing me something that looks like a Google Map with more than a hundred little colored dots that represent schools. The map is situated on a busy excel sheet that displays various start times, the number of buses available to take kids to and from school, and how much money the school would be saving for each efficiency found for its routes.
Martin shows me how….if you change one thing, the whole thing recalculates how it would affect the transportation to other schools and the money either saved or…not saved…as a RESULT of that change.
MARTIN: …this solution…needs this number of buses…which would mean the district would save $4 million a year…and this is an average course change of 33 minutes…[fades out]
NARR: It does all of this REALLY quickly…in just seconds. This is a computing system that Martin and his colleague—Arthur Delarue—built to help schools quickly come up with a solution for …any number of issues a school district might face.
And it was an earlier version of this system that the researchers used to tackle the many issues that the Boston Public School District wanted Martin and Delarue to address. Because, like I said, they wanted to cut transportation costs, but there was more to it than that. A LOT more.
MARTIN: The elephant in the room is start times and school schedules.
NARR: One of the key changes BPS wanted to make was to start more high schoolers later in the day….after 8 a.m. Nearby schools were trying out later start times…too…and the reception was positive. It was gaining traction because some research had come out showing just how much teens needed their sleep to lead happier, healthier, more academically successful lives. And studies were showing that…biologically….adequate sleep for high school students tended to be 9 to 10 hours.
Unfortunately, at BPS…
MARTIN: Most high schoolers were starting as early as 7:15 in the morning. If you add one hour of transportation, you can imagine…
NARR: It was really hard for BPS students to make that work.
On top of moving high schools later…BPS also wanted to distribute start times more equitably across all of its schools. Because, as it stood, there was a clear trend in who did and didn’t get to start at the most desirable time slot — which according to parent surveys…was between 8 and 9 a.m. The ones with the so-called “good” times tended to include those where the population was whiter and more resourced. The ones who wound up with less-ideal times — which tended to be too early or too late for many people’s tastes…tended to be schools in poorer neighborhoods with higher concentrations of black and brown students.
BPS said it had lots of conversations with the community and leaders about what they wanted and didn’t want…out of this start time plan. And everyone seemed to agree with what the school committee was trying to do.
So Martin, Delarue and Bertsimas took all those priorities…and landed on a plan that balanced…all of this….the cost-cutting, the better routes…later high school start times…and a more equitable distribution of the good start times. It cut 120 buses from the district’s fleet and gave 85 percent of the district new start times. It was an epic, groundbreaking feat.
Then-superintendent Tommy Chang mused to the Boston Globe, before a big school committee vote, that quote “this is a problem that nobody thought we could solve, and we are going to solve it tonight.”
NARR: Jane Miller remembers when she found out BPS was about to change its school times.
Jane MILLER: I couldn’t sleep that night.
NARR: Miller is a parent at Boston Public Schools. And at the time that all of this was supposed to take effect, she had three kids going to JP Manning elementary school in Boston’s Jamaica Plain neighborhood…which is about 3 miles southwest of Fenway Park.
MILLER: Our family is a late family…as I just said…my kids are rolling out of bed…um..hahah…We chose a late start school because of our schedules. So my husband doesn’t get home until midnight…so he didn’t want to be dropping off the kids at 7am. I work days, he works nights…that way there’s always a drop off and a pickup … and we wanted to have some quality of life with him sleeping.
NARR: At BPS, you have a say in which school you get to go to…through a ranking system. Miller’s kids got Manning Elementary…and it worked perfectly for them because kids started school at 9:30 a.m. and ended at 4:10 p.m…so her husband could easily do the drop off…and she could easily do the pick-up.
But then BPS released its list of changed school times…and, under the change, Manning elementary would be starting at 7:15 in the morning and ending at 1:55 in the afternoon.
MILLER: Which is the middle of the day.
NARR: This was a FAR cry from the 9:30 am start time and 4:10 p.m. end time they were accustomed to. Hundreds of families were facing the same nearly-two-hour time shift. And while, yes, the school did offer before and after school care, Miller said it was expensive and filled up really quickly.
MILLER: I had just started a new career and the cost of three of my kids to go to after school was going to be more than I earned. And I felt like once again, it’s like a woman who’s going to have to figure out like, okay, you know, I either don’t eat my pay or I give up a career that had waited a really long time to get into. Um, and that was my first year.
NARR: She knew many other families in other elementary and middle schools who were afraid they couldn’t swing their new start times either. They were all dealing with all kinds of different realities. Like…having kids at multiple schools with drastically different start times they hadn’t planned for. And there were families where the teens, if they got off of school too late, couldn’t work jobs that they needed to help support their families.
So Miller did something.
Laura PAVIN: When did you decide to get a petition going?
MILLER: Oh, I didn’t really decide. I felt really like…do I have to quit my job? And, um, maybe at like 3:00 AM, I wrote the petition and posted it and then finally fell asleep. And when I woke up in the morning, a whole bunch of people had signed it.
NARR: Thousands of people ended up signing the petition…which asked the district to stop the school start time changes immediately. Many of those same parents packed the school board meeting the following week and spoke for hours during public comment...about it.
It was deeply unpopular. The committee dropped the whole thing…months later, former Mayor Marty Walsh asked Superintendent Tommy Chang to resign from his role. The district’s spending on transportation still remains among the highest in the nation…compared to other districts.
NARR: Martin told me that…the outcome for BPS and its start times…stayed with him.
MARTIN: It was really hard on me. And what you heard, like all of these testimonies in, in, um, in the committee meetings are all true, very hard stories.
NARR: Sure, Martin IS the computers and math guy…but he doesn’t operate in a vacuum. He knows that using algorithms to change policies of any kind…will affect real people.
So…Martin wondered…at what point in this process did things go wrong? Was there anything the leaders of this project could have done different…to make this more palatable for parents?
I asked him if there was a communication breakdown between the parents and the school district. And he didn’t think so.
MARTIN: I’ve been through dozens of community engagement meetings with the school district before the change. It would be hard for me to say that there has not been committee engagement.
NARR: BPS WAS, indeed, clear on its intent: to cut costs, improve its routes, distribute start times more equitably, and start high school students later in the day. And it told parents about the start time changes that were on the table. They asked them to rank start times between 7 AM and 9:30 AM on a scale of one to seven — one being the least preferable and seven being the most preferable.
BPS operations chief…John Hanlon…presented the results of this district-wide survey at a school committee meeting leading up to the change. And he was blunt: the 8:00 - 8:30 time frame WAS the most popular with parents….but not by a huge margin. Because NONE of the start times were universally popular across the whole district.
John HANLON: If you notice here, not a single start time on this graph has an average score of even five out of seven…and what that amounts to, is there’s, quite frankly, indifference, across the district.
NARR: During his presentation, Hanlon even showed the committee a graph of parental preferences at just one school. And the preferences were scattered all over the place.
HANLON: And ultimately, what this shows us, is that within schools, it is impossible to make everyone happy…and as part of our engagement efforts after the last several months, we made it very clear to everybody, this is a classic example of we will not be able to make everyone happy.
NARR: So it’s not like BPS kept the high likelihood of parental dissatisfaction a secret. Their engagement went beyond surveys, too…they’d held committee conversations…sessions with the community and phone conversations in multiple languages …which they set up a call center to do.
But Miller still felt like they fell short.
MILLER: There was never an idea of what the impact would be to individual schools. We knew, they said it was okay to change the start times, but nobody knew what it would mean to your individual school, um, until a list rolled out.
NARR: The WAY that the new schedules were announced…wasn’t ideal.
The committee voted on the framework for the new bell times just one day before they were released to the public. This was in December of 2017…and the new start times were set to take effect the following school year. So yeah…that was pretty quick.
I should say that Miller is white…and so were a lot of the parents that spoke out over the start times. And that wasn’t lost on some spectators…who concluded that the plan was being dropped to satisfy this vocal, well-connected few…at the expense of more under-resourced communities.
It’s a narrative that would make sense, given Boston’s history with race and busing.
Back in the 1970s, a U.S. District Court had required BPS to desegregate the city’s schools by busing students between white and black areas of the city. The opposition was violent. And eventually, much of the white population fled the city and its schools.
And so, a lot of people with this historical context were quick to blame the crumbling of BPS’s start time plan …on the fewer but vocal white families that remained in the school district.
But…the thing about that is that…later, after that raucous meeting about start times with the mostly-white parents…the NAACP and the Lawyers’ Committee for Civil Rights and Economic Justice ALSO spoke out against the plan. This didn’t surprise Miren Uriarte. She was a school committee member when the conversations about school start times were happening. I talked to her over the phone about it.
URIARTE: Black and brown enrollment are working class and, and below, you know, that have jobs that are, they do not give them a lot of flexibility. And, you know, people have choices and parents make choices on the basis of their needs.
NARR: In other words, many parents with fewer resources had already opted into a school with start time that they favored–and were likely to find it even HARDER to deal with a sudden schedule change than parents with more resources.
So…maybe the district should have rolled out the change more slowly…done some beta testing…given families more time to understand, truly, what the schedule change would mean for their lives. And whether it would just be too onerous for people.
That could have been part of it. But Martin wondered if there was something deeper to the pushback over start time changes.
MARTIN: I think there is something that is a little bit impossible in there…which is…when it’s your children, it’s extremely hard to think about the group.
NARR: Which is to say…EVEN if these new start times would be best for many of the district’s sleep-deprived teens…people tend to think about what’s best for their own kids…and not someone else’s.
MARTIN: There is nothing more humane than this. This is why I really want to insist that. I mean, on my side, at least there is no hard feeling. I just see it as one of these extremely complex problems.
NARR: At least in Boston, this complexity still seems to be preventing any solutions. Here’s former school committee member Miren Uriarte again.
URIARTE: To this day, we have not been able to change the start times because of that incident.
NARR: Are we ready…truly ready…to let artificial intelligence help us solve problems that affect so many lives…in so many ways?
Martin still thinks so.
Dozens upon dozens of school districts reached out to him and his MIT colleagues to help them solve transportation problems of their own…even after everything that happened with BPS. They even created a startup called AlphaRoute to tackle the demand.
Recently, Martin actually did get to see the fruits of his algorithm’s labor go to work at San Francisco Unified School District. The district DID actually start its high schools later and alter other start times for the 21-22 school year. And it’s set to save the district $5.5 million each year.
MARTIN: I think it’s maybe the first time a school district—a major school district—managed to change their start time in a major way…especially using an algorithm. Of course it was hard. Changing start times is never a happy thing for many parents. Like, it’s very hard for a lot of parents.
NARR: What went differently this time? Well, a state law.
In 2019, California Governor Gavin Newson signed a bill that required high schools to push start times back to 8:30 a.m. So, the school district and the community it served…didn’t really have a choice.
Now, thanks to these legally mandated changes…we finally can see what an AI-optimized schedule can really do …when it’s left to play out over the years.
MARTIN: I’m convinced that it will be so much better for the district, for the teachers, for the students’ health.
NARR: So…are laws and mandates the answer here? Is short-term resistance all that’s really holding AI back from making our organizations and societies better?
Not necessarily…Martin says there is still validity to the criticism that leaders can be a little too quick to wield algorithms as a cure-all–particularly when it treats people as statistics and not as individuals, with their own uniquely idiosyncratic lives and preferences.
It’s hard to know for sure what Boston could have done differently. What Martin DOES feel more certain about is that…there aren’t enough leaders out there who really know how to use public policy and algorithms in tandem with each other.
MARTIN: You need translators, you need people who know both worlds.
NARR: In other words, he thinks we need more technically-informed leaders, leaders who can help us to build a world that’s optimized by computers…FOR humans.
NARR: This episode of The Insightful Leader was written by me—Laura Pavin. It was produced by Jessica Love, Fred Schmalz, Emily Stone, Maja Kos, Andrew Meriwether and Laura Pavin. Special thanks to Sebastien Martin. As a reminder, you can find us on iTunes, Google Play, or our website. If you like this show, please leave us a review or rating—that helps new listeners find us. We’ll be back in a couple weeks with another episode of The Insightful Leader.
What Went Wrong at Silicon Valley Bank?And how can it be avoided next time? A new analysis sheds light on vulnerabilities within the U.S. banking industry.
How Are Black–White Biracial People Perceived in Terms of Race?Understanding the answer—and why black and white Americans may percieve biracial people differently—is increasingly important in a multiracial society.
Podcast: Workers Are Stressed Out. Here’s How Leaders Can Help.On this episode of The Insightful Leader: You can’t always control what happens at work. But reframing setbacks, and instituting some serious calendar discipline, can go a long way toward reducing stress.
3 Things to Keep in Mind When Delivering Negative FeedbackFirst, understand the purpose of the conversation, which is trickier than it sounds.
Will AI Eventually Replace Doctors?Maybe not entirely. But the doctor–patient relationship is likely to change dramatically.
What Went Wrong at AIG?Unpacking the insurance giant's collapse during the 2008 financial crisis.
Which Form of Government Is Best?Democracies may not outlast dictatorships, but they adapt better.
What Happens to Worker Productivity after a Minimum Wage Increase?A pay raise boosts productivity for some—but the impact on the bottom line is more complicated.
Why Do Some People Succeed after Failing, While Others Continue to Flounder?A new study dispels some of the mystery behind success after failure.
Why Well-Meaning NGOs Sometimes Do More Harm than GoodStudies of aid groups in Ghana and Uganda show why it’s so important to coordinate with local governments and institutions.
How Much Do Campaign Ads Matter?Tone is key, according to new research, which found that a change in TV ad strategy could have altered the results of the 2000 presidential election.
How Peer Pressure Can Lead Teens to Underachieve—Even in Schools Where It’s “Cool to Be Smart”New research offers lessons for administrators hoping to improve student performance.
Marketers, Don’t Be Too Hasty to Act on DataDon’t like the trends you’re seeing? It’s tempting to take immediate action. Instead, consider a hypothesis-driven approach to solving your problems.
When You’re Hot, You’re Hot: Career Successes Come in ClustersBursts of brilliance happen for almost everyone. Explore the “hot streaks” of thousands of directors, artists and scientists in our graphic.
How Has Marketing Changed over the Past Half-Century?Phil Kotler’s groundbreaking textbook came out 55 years ago. Sixteen editions later, he and coauthor Alexander Chernev discuss how big data, social media, and purpose-driven branding are moving the field forward.
Podcast: "It's Hard to Regulate U.S. Banks!"Silicon Valley Bank spectacularly collapsed—and a new analysis suggests that its precarious situation is not as much of an outlier as we’d hope. On this episode of The Insightful Leader, we learn what went wrong and what should happen next.
At Their Best, Self-Learning Algorithms Can Be a “Win-Win-Win”Lyft is using ”reinforcement learning” to match customers to drivers—leading to higher profits for the company, more work for drivers, and happier customers.
Immigrants to the U.S. Create More Jobs than They TakeA new study finds that immigrants are far more likely to found companies—both large and small—than native-born Americans.
Leaders, Don’t Be Afraid to Admit Your FlawsWe prefer to work for people who can make themselves vulnerable, a new study finds. But there are limits.