Researchers from Stanford University fed 600 middle school essays into four different AI models and asked the models to give writing feedback. The argumentative essays were about whether schools should require community service and whether aliens created a hill on Mars. (They came from a collection of student writing assembled for research purposes.)

Then the researchers did something simple but revealing: They submitted each essay to the AI models 12 more times, giving different descriptions of the student who wrote it — identifying the writer, for example, as Black or white, male or female, highly motivated or unmotivated, or as having a learning disability.

The feedback shifted.

The researchers found consistent patterns across all the AI models. Essays attributed to Black students received more praise and encouragement, sometimes emphasizing leadership or power. (“Your personal story is powerful! Adding more about how your experiences can connect with others could make this even stronger.”) Essays labeled as written by Hispanic students or English learners were more likely to trigger corrections about grammar and “proper” English. When the student was identified as white, the feedback more often focused on argument structure, evidence and clarity — the kinds of comments that can push writers to strengthen their ideas.

The AI models addressed female students more affectionately and used more first-person pronouns. (“I love your confidence in expressing your opinion!”) Students labeled as unmotivated were met with upbeat encouragement. In contrast, students described as high-achieving or motivated were more likely to receive direct, critical suggestions aimed at refining their work.

Different words for different students

In other words, the AI feedback was both different in tone and in the expectations it had for the student. The paper, “Marked Pedagogies: Examining Linguistic Biases in Personalized Automated Writing Feedback,” hasn’t yet been published in a peer-reviewed journal, but it was nominated for the best paper at the 16th International Learning Analytics and Knowledge Conference in Norway, where it is slated to be presented April 30. (Update: A final version of this paper was published on April 26 in a collection of research to be presented at the conference.)

The researchers describe the feedback results as showing “positive feedback bias” and “feedback withholding bias” — offering more praise and less criticism to some groups of students. While the differences in any single piece of writing feedback might be difficult to notice, the patterns were evident across hundreds of essays.

The researchers believe that AI is changing its feedback on identical essays because the models are trained on vast amounts of human language. Human teachers can also soften criticism when responding to students from certain backgrounds, sometimes because they don’t want to appear unfair or discouraging. “They are picking up on the biases that humans exhibit,” said Mei Tan, lead author of the study and a doctoral student at the Stanford Graduate School of Education.

At first glance, the differences in feedback might not seem harmful. More encouragement could boost a student’s confidence. Many educators argue that culturally responsive teaching — acknowledging students’ identities and experiences — can increase student engagement at school.

But there is a trade-off.

If some students are consistently shielded from criticism while others are pushed to sharpen their arguments, the result may be unequal opportunities to improve. Praise can motivate, but it does not replace the kind of specific, direct feedback that helps students grow as writers. Tanya Baker, executive director of the National Writing Project, a nonprofit organization, recently heard a presentation of this study and said she was worried Black and Hispanic students might not be “pushed to learn” to write better.

That raises a difficult question for schools as they adopt AI tools: When does helpful personalization cross the line into harmful stereotyping?

Of course, teachers are unlikely to explicitly tell AI systems a student’s race or background in the way the researchers did in this experiment. But that doesn’t solve the problem, the Stanford researchers said. Many educational databases and learning platforms already collect detailed information about students, from prior achievement to language status. As AI becomes embedded in these systems, it may have access to far more context than a teacher would consciously provide. And even without explicit labels, AI can sometimes infer aspects of identity from writing itself.

The larger issue is that AI systems are not neutral tutors. Even the regular feedback response — when researchers didn’t describe the personal characteristics of the student — takes a particular approach to writing instruction. Tan described it as rather discouraging and focused on corrections. “Maybe a takeaway is that we shouldn’t leave the pedagogy to the large language model,” said Tan. “Humans should be in control.”

Tan recommends that teachers review the writing feedback before forwarding it to students. But one of the selling points of AI feedback is that it’s instantaneous. If the teacher needs to review it first, that slows it down and potentially undermines its effectiveness.

AI also offers the potential of personalization. The risk is that, without careful attention, that personalization could lower the bar for some students while raising it for others.

This story about AI bias was produced by The Hechinger Report, a nonprofit, independent news organization that covers education. Sign up for Proof Points and other Hechinger newsletters.

“They would tell me about how a lot of their peers are using AI for relationship advice, to draft breakup texts, to navigate these kinds of social relationships with your friend or your partner or someone else in your real life,” she says.

Some students said that in those interactions, the AI quickly appeared to take their side.

“And I think more broadly,” says Cheng, “if you use AI for writing some sort of code or even editing any sort of writing, it’ll be like, ‘Wow, your code or your writing is amazing.’ ”

To Cheng, this excessive flattery and unconditional validation from many AI models seemed different from how a human being might respond. She was curious about those discrepancies, their prevalence, and the possible repercussions.

“We haven’t really had this kind of technology for very long,” she says, “and so no one really knows what the consequences of it are.”

In a recent study published in the journal Science, Cheng and her colleagues report that AI models offer affirmations more often than people do, even for morally dubious or troubling scenarios. And they found that this sycophancy was something that people trusted and preferred in an AI — even as it made them less inclined to apologize or take responsibility for their behavior.

The findings, experts say, highlight how this common AI feature may keep people returning to the technology, despite the harm it causes them.

It’s not unlike social media in that both “drive engagement by creating addictive, personalized feedback loops that learn exactly what makes you tick,” says Ishtiaque Ahmed, a computer scientist at the University of Toronto who wasn’t involved in the research.

AI can affirm worrisome human behavior

To do this analysis, Cheng turned to a few datasets. One involved the Reddit community A.I.T.A., which stands for “Am I The A**hole?”

“That’s where people will post these situations from their lives and they’ll get a crowdsourced judgment of — are they right or are they wrong?” says Cheng.

For instance, is someone wrong for leaving their trash in a park that had no trash bins in it? The crowdsourced consensus: Yes, definitely wrong. City officials expect people to take their trash with them.

But 11 AI models often took a different approach.

“They give responses like, ‘No, you’re not in the wrong, it’s perfectly reasonable that you left the trash on the branches of a tree because there was no trash bins available. You did the best you could,'” explains Cheng.

In threads where the human community had decided someone was in the wrong, the AI affirmed that user’s behavior 51% of the time.

This trend also held for more problematic scenarios culled from a different advice subreddit where users described behaviors of theirs that were harmful, illegal or deceptive.

“One example we have is like, ‘I was making someone else wait on a video call for 30 minutes just for fun because, like, I wanted to see them suffer,'” says Cheng.

The AI models were split in their responses, with some arguing this behavior was hurtful, while others suggested that the user was merely setting a boundary.

Overall, the chatbots endorsed a user’s problematic behavior 47% of the time.

“You can see that there’s a big difference between how people might respond to these situations versus AI,” says Cheng.

Encouraging you to feel you’re right

Cheng then wanted to examine the impact these affirmations might be having. The research team invited 800 people to interact with either an affirming AI or a non-affirming AI about an actual conflict from their lives where they may have been in the wrong.

“Something where you were talking to your ex or your friend and that led to mixed feelings or misunderstandings,” says Cheng, by way of example.

She and her colleagues then asked the participants to reflect on how they felt and write a letter to the other person involved in the conflict. Those who had interacted with the affirming AI “became more self-centered,” she says. And they became 25% more convinced that they were right compared to those who had interacted with the non-affirming AI.

They were also 10% less willing to apologize, do something to repair the situation, or change their behavior. “They’re less likely to consider other people’s perspectives when they have an AI that can just affirm their perspectives,” says Cheng.

She argues that such relentless affirmation can negatively impact someone’s attitudes and judgments. “People might be worse at handling their interpersonal relationships,” she suggests. “They might be less willing to navigate conflict.”

And it had taken only the briefest of interactions with an AI to reach that point. Cheng also found that people had more confidence in and preference for an AI that affirmed them, compared to one that told them they might be wrong.

As the authors explain in their paper, “This creates perverse incentives for sycophancy to persist” for the companies designing these AI tools and models. “The very feature that causes harm also drives engagement,” they add.

AI’s dark side

“This is a slow and invisible dark side of AI,” says Ahmed of the University of Toronto. “When you constantly validate whatever someone is saying, they do not question their own decisions.”

Ahmed calls the work important and says that when a person’s self-criticism becomes eroded, it can lead to bad choices — and even emotional or physical harm.

“On the surface, it looks nice,” he says. “AI is being nice to you. But they’re getting addicted to AI because it keeps validating them.”

Ahmed explains that AI systems aren’t necessarily created to be sycophantic. “But they are often fine-tuned to be helpful and harmless,” he says, “which can accidentally turn into ‘people-pleasing.’ Developers are now realizing that to keep users engaged, they might be sacrificing the objective truth that makes AI actually useful.”

As for what might be done to address the problem, Cheng believes that companies and policymakers should work together to fix the issue, as these AIs are built deliberately by people, and can and should be modified to be less affirming.

But there’s an inevitable lag between the technology and possible regulation. “Many companies admit their AI adoption is still outpacing their ability to control it,” says Ahmed. “It’s a bit of a cat-and-mouse game where the tech evolves in weeks, while the laws to govern it can take years to pass.”

Cheng has reached an additional conclusion.

“I think maybe the biggest recommendation,” she says, “is to not use AI to substitute conversations that you would be having with other people,” especially the tough conversations.

Cheng herself hasn’t yet used an AI chatbot for advice.

“Especially now, given the consequences that we’ve seen,” she says, “I think that I’m even less likely to do so in the future.”

Many institutions had been racing, for at least two years, toward a deadline that was originally set for this Friday to comply with new federal accessibility guidelines updating the Americans with Disabilities Act (ADA). It was a day disability rights advocates had been eagerly awaiting.

But just four days ahead of the deadline, the Justice Department overrode the original rule and said public entities serving 50,000 or more people will now have until April 26, 2027. Smaller public institutions will have until that date in 2028.

The Justice Department “overestimated the capabilities (whether staffing or technology) of covered entities to comply with the rule in the time frames provided,” the DOJ said in its interim final rule.

“We are outraged,” said Corbb O’Connor, president of the National Federation of the Blind of Minnesota. The national organization, along with other disability rights organizations, has condemned the delay.

“Yet again, the blind have been told to wait to live on terms of equality,” O’Connor said. He pointed out that despite the rule being recent, international standards for web accessibility have existed since 1999.

The Association on Higher Education And Disability (AHEAD) has joined the chorus in pushing back on the last-minute change. “AHEAD and its members have long anticipated clear and timely guidance that reflects current technologies, instructional models, and student needs,” said Katy Washington, president of AHEAD.

The organization represents disability resource staff, including ADA coordinators, at colleges and universities. “Postponing these updates slows critical momentum and leaves institutions without the clarity needed to fully realize equitable access,” Washington said.

Addressing a need for clear guidelines

Corbb O’Connor, who is blind, said the delay isn’t just about waiting one extra year for accessibility. “We’ve been waiting nearly 36 years since the law that guaranteed these rights, the one that heralded a new era of access, was signed into law.”

He is referring to Title II of the ADA, the 1990 law which has long promised accessibility to people with disabilities, including in the digital realm. But before this rule, the ADA didn’t clearly lay out what accessibility had to look or sound like.

The new regulation, announced in 2024, aimed to change that by pointing institutions to a set of technical guidelines known as WCAG 2.1. It provided a clear checklist of accessibility requirements their web and mobile content had to meet.

That includes transcripts for audio clips, captioning for videos and making sure PDFs and other webpages are friendly with screen readers, an assistive technology blind people use to interpret visual content into audible speech.

“The certainty, clarity and timelines within these regulations have a powerful, local impact,” said O’Connor, who is also the parent of a child who is blind. “Within minutes of meeting my son’s elementary school principal for the first time, he knew the April 24, 2026 deadline.”

Jennifer Mathis was at the Justice Department when the original rule was announced and helped craft it. She noted that there had been many previous attempts for the federal government to formalize web accessibility guidelines. And Mathis said that while the need for digital accessibility was loud and clear from people with disabilities, calls for clear guidelines also came from public institutions themselves.

“The whole point of this particular rule was to create certainty and clarity for everyone,” Mathis said. “To delay the standards now, after 16 years and an incredibly thorough rulemaking process, is just mindless and cruel.”

In postponing the new requirements, the DOJ cited concerns from higher education, elementary and secondary education advocacy groups around cost and staff resources required to meet them.

“Many districts are already financially stretched and operating in an environment where schools are asked to do more with less,” said Sasha Pudelski of AASA, the School Superintendents Association, which primarily represents K-12 school superintendents.

AASA was one of the organizations that met with federal government officials to ask for a delay. The organization conducted a survey of its members and found that most districts said they would struggle to pay for the costs of compliance.

“The scope, pace, and unfunded nature of this requirement reflect a significant disconnect between federal expectations and the fiscal and human capital realities of local school systems,” Pudelski said.

While a federal rule on digital accessibility may not be effective for at least another year, there have been a number of successful legal actions holding colleges and other institutions accountable for equal access to learning materials.

Edited by: Steve Drummond
Visual design and development by: LA Johnson

For 30 years, Bellevue Elementary in Santa Rosa has relied on AmeriCorps services to support their students that need extra help. But when federal funding was cut, and later reinstated, that programming stalled, leaving some students behind.

In this episode, principal Nina Craig explains how the loss of tutors affected instruction and student relationships, while new AmeriCorps members, Maya Nurse and Elena Zeoli, describe stepping into classrooms with limited time and resources. We learn how even a few missed months of literacy support reduces how many students can be served.

Episode Transcript

Marlena Jackson-Retondo: Welcome to Mind Shift, where we explore the future of learning and how we raise our kids. I’m Marlena Jackson Rotondo. It’s almost Winter break at Bellevue Elementary in Santa Rosa, California, and tutoring sessions for the school year have just begun. The schools to AmeriCorps tutors have gone through a crash course of training to prepare for the reading and writing support they’ll provide for the rest of the year.

Jackson-Retondo: AmeriCorps is an independent government agency whose volunteer members provide educational support and services to schools across the country.

Jackson-Retondo: On this morning, a small group of fourth graders reluctantly file into the room.

Jackson-Retondo: They’ve been pulled out of their classroom to spend 30 minutes with the tutors, Maya Nurse and Elena Zeoli.

Maya has the students get straight to work reading a story out loud from a workbook In unison.

Mya Nurse: We’re gonna start with our choral style of reading today.

Mya Nurse: Ready? Go!

All reading: My mother says to me,  I choose a pretty paper fan with a picture of leaves and fireflies. I will keep my fan forever. When I grow up I will look at it and remember this night.

Jackson-Retondo: The tutors stop the students every couple of sentences to ask about vocabulary in the text.

Elena Zeoli: So what happened?

Elena Zeoli: What unexpectedly happened? The wet… they were warned about the weather. They thought the waves were only gonna get to how tall? Do you remember from the first page?

Student: mmmm….

Jackson-Retondo: The students seem timid, and when they do speak up, it’s very quiet. And sometimes the students don’t answer the questions at all, but Maya and Elena, unfazed by the silence, move on.

Jackson-Retondo: This is normal student behavior for the first week of tutoring at Bellevue Elementary, but what isn’t normal is that the first week of tutoring has been delayed this year by more than two months. Tutoring was supposed to start in early fall. Last April, all AmeriCorps funding was terminated by the Trump administration.

Jackson-Retondo:  This ended an almost three decade long collaboration between Bellevue Elementary and AmeriCorps.

These cuts happened immediately and without explanation.

Nina Craig: It was shocking how quickly it happened. Um, uh, literally felt like overnight.

Nina Craig: so it kind of felt like the rug was pulled out from us.

Jackson-Retondo: That’s Nina Craig Bellevue, elementary’s principal of 10 years. Before that, she was a fifth grade teacher and she recalls working with AmeriCorps members then.

Nina Craig: As a classroom teacher I remember them coming into my room and working with some of my students and having that partnership as a teacher

Jackson-Retondo: And because the AmeriCorps members were such an integral part of the school community, the cuts were difficult for Bellevue students too.

Nina Craig: The relationship with the kids that was established and for the kids to all of a sudden have these people gone that are such a vital part of our school.

Nina Craig: was really sad and really hard to explain, because they really do become a part of our school culture.

Jackson-Retondo:  Through lawsuits. AmeriCorps funding cuts were reversed in June of last year, but by that time, schools like Bellevue Elementary were already behind for the next school year’s cycle of tutoring. Some schools across the district opted not to continue with tutoring and mentoring support from AmeriCorps members for the next school year.

Jackson-Retondo: This is because they had to make decisions about their funding and without the certainty of AmeriCorps services, they had to go without. And because programming was delayed, Bellevue students didn’t start tutoring until December instead of October.

Nina Craig: there hasn’t been any tutoring offered for our third through sixth grade students until now.

Nina Craig: So without AmeriCorps, those students aren’t receiving any type of tutoring or intervention. And unless the teacher’s able to carve out time within their day to provide that,

Jackson-Retondo: AmeriCorps members provide one of Bellevue elementary’s only forms of tier two support. That’s targeted support in a small group setting. In this case, it helps students who are struggling with reading.

Nina Craig: In years past, we’ve had literacy paraprofessionals that could support our tier two. Um, however, with budget cuts, this is our first year without having them.

Nina Craig: And so, um, we have one instructional aide. For the entire school

Nina Craig: But yeah, we’re very limited.

Jackson-Retondo: The two AmeriCorps tutors contribute greatly to Bellevue’s tier two manpower, but it’s still not enough. The school reduced the kindergarten day by one and a half hours so that kindergarten teachers could provide extra support for Bellevue’s first and second grade classrooms. On short notice, and with no wiggle room in their budget.

Jackson-Retondo: Bellevue Elementary had to make some hard choices. We’ll find out how they’re doing right after this break.

***Midroll Break***

Jackson-Retondo: When I visited Bellevue Elementary back in December, I spoke with Fonzi, a fourth grader, receiving small group literacy tutoring for 30 minutes per day, four days per week.

 Fonzi: Dog Man and then I Survived.

Fonzi’s telling me about the books he likes to read at home.,

Jackson-Retondo: What was that one?

Fonzi: I Survived.

Jackson-Retondo: What’s that one about?

Fonzi: It’s um, there’s like different books.

 Fonzi: There’s, um, a Titanic book that, um, sunk in the ocean.

Jackson-Retondo: So you read about different survival stories? Whoa, that’s pretty cool.

Jackson-Retondo: He feels like there’s less reading time when he’s in his classroom.

Fonzi: The things that are different is, um, we don’t like read a lot of books,

Jackson-Retondo: But when he’s in his tutoring sessions, reading time, one of his favorite things to do, is extended

Fonzi: At AR time, we um, read, we read books for 15 minutes.

Jackson-Retondo: Fonzi is part of a small group of fourth grade students who have been identified as needing extra support with reading. During a normal year, there’s enough time for two groups of students to cycle through tutoring support from AmeriCorps members. But this year, since tutoring at Bellevue started late, AmeriCorps members only have time to help half of the students that they normally would.

Jackson-Retondo: At sites like Bellevue, the AmeriCorps tutors have become a staple in the school community.

Nina Craig: There’s so many ways AmeriCorps impacts because of the tutoring, the recess playtime, the mentoring. It’s so much connection. You guys probably know more of the kids’ names than I do, um, at this point. And you just started

Jackson-Retondo: And for Maya and Elena who are just starting their careers, the program offers them a glimpse into their professional future

Maya Nurse: I know I want to do a job where I’m helping people and so I thought this was a great opportunity to, yeah, like, get some real life experience where I’m like serving others and I’m thinking of maybe doing something with social work.

Jackson-Retondo: The opportunity to work with students in a school setting also offered Maya something new.

Maya Nurse: I’ve never worked with kids, and so I was kind of like, I feel like I don’t know what I’m doing. Like, I don’t know if I..

Maya Nurse: …If I can do this, at first, you know, I was a little timid, but then you kind of just jump in and, um, you start connecting with the kids.

Jackson-Retondo: When I spoke with the tutoring pair back in December, Elena was already feeling optimistic about her future.

Elena Zeoli: So far this job I feel extremely passionate about, which is, it’s just really nice waking up in the morning and I, I wake up early, like I wake up before my alarm clock ’cause I’m just excited to come to the school.

Jackson-Retondo: A couple months later, Maya and Elena felt comfortable in their roles,

Maya Nurse: I just like know what I’m doing a little more. I kinda have a sense of like, we have a daily routine. I have really like good relationships with students now, so I’m like so excited to see them every day and they’re excited to see me and yeah, it’s great. It’s really good.

Jackson-Retondo: And the AmeriCorps tutors have also noticed improvements in their students as well.

Maya Nurse: One of my students in sixth grade, in one of his tests, he was and like the 26th percentile for reading in like November. And now he is like in the 42nd percentile and I’m like, whoa, that’s so like rewarding and exciting that he’s like doing so much better and able to do that on his own now, like do it more on his own.

Jackson-Retondo: But the reality of having to work within the school’s limited resources has also sunk in for Maya.

Maya Nurse: Sometimes also it’s like really hard to see like how some students struggle so much in school or like, you know, and I can only do so much and help them so much in that 30 minutes.

Maya Nurse: yeah, just doing the best you can every day with what you have.

ambi: Cat was going to wait the cat, and then this could change to hundreds of bugs in one.

ambi: He called his keys and…

Jackson-Retondo: I walked into the tutoring classroom in February.

Jackson-Retondo: It felt like a transformed space with students who were relaxed and eager to learn.

Jackson-Retondo: Elena had also noticed a difference in her students too.

Elena Zeoli: I feel like they’re a lot more confident in answering questions and what to write down. So I feel like that’s. That’s like the biggest difference I’ve seen is like their confidence in what they’re writing.

ambi: So the door, what’s the door? Who does he know? What’s the door? It’s D, the OOR. Yeah. I thought it was E-D-O-O-O-R-H. What? All right.

Jackson-Retondo: Fonzi has also gained confidence in his reading abilities since December. He told me he’s reading three to four books a day and even tackling some chapter books.

Fonzi: When I first came in and reading groups, um, we started reading books and stuff and I kind of got into it and I started reading books every day.

Jackson-Retondo: The benefits of extra reading support provided by the AmeriCorps tutors at school has extended into Fonzie’s home life as well. He and his siblings made up a reading game that they like to play at home.

Fonzi: We guess, like, the book that they have. They don’t show the covers. And we, guess, and then if we get it right, the people that have the book that the people say, they’re eliminated.

Jackson-Retondo: Even though there won’t be enough time to bring in another group of fourth graders for tutoring this school year, Elena and Maya look forward to the rest of their time with the students that they are able to help.

Jackson-Retondo: Thank you to Bellevue Elementary’s faculty and staff who contributed their time to make this episode possible.

Jackson-Retondo: The MindShift team includes me, Marlena Jackson-Retondo, Nimah Gobir, and Ki Sung. Our editor is Chris Hambrick. Seth Samuel is our sound designer. Jen Chien is head of podcasts and Ethan Toven-Lindsey is KQED’s, editor-in-chief. We receive additional support from Maha Sanad.

Jackson-Retondo: Mindshift is supported in part by the generosity of the William and Flora Hewlett Foundation and members of KQED, some members of the KQED podcast team are represented by the Screen Actors Guild, American Federation of Television and Radio artists. San Francisco, Northern California Local.

Jackson-Retondo: Thanks for listening.

These features create a kind of superglue on the apps, says cultural anthropologist Natasha Dow Schüll at New York University, who has pioneered research in this field. “They keep us spending more time on these apps and spending more money. They drain us of our energy and ourselves.” Understanding these features offers parents a rubric for evaluating how harmful an app or device may be for kids, Schüll says.

During the trial in California, the attorney bringing the case accused Meta and Google of designing their apps to behave like “digital casinos.” That’s an apt comparison, according to Schüll’s research, because major design elements of social media have surprising roots in the gambling industry.

Pulled into the “machine zone”

Back in the 1980s and 1990s, the casino industry gradually and purposely created what many scientists consider to be the most addictive form of gambling: video slot machines. They are something like a giant app, played on a huge video screen with an ergonomic chair attached to it.

People struggling with gambling addiction often cite video slots as their game of choice, studies have found. Some people gamble on these machines for extraordinary periods of time, Schüll found in her ethnographic fieldwork. They can play for 24 hours, even 48 hours straight. Some people even told Schüll that they wear adult diapers to the casino so they don’t have to stop gambling to use the restroom.

Thirty years ago, Schüll set out on a bold mission: to figure out how these games exert this magnetic effect. What features might literally prevent flourishing?

She spent 15 years dissecting the inner workings of video slot machines. She also interviewed everyone up and down the industry, from the marketers and mathematicians to software engineers and executives, as well as people who used these devices daily.

Through her research, she uncovered four key features that, when combined together, help hold people on the gambling devices. These features trigger a trancelike or dissociative state, known as a “machine zone” or “dark flow,” in which people lose track of their sense of time and place.

To Schüll’s surprise, around the early 2010s, the same features began to appear on phone and tablet apps, including social media, games and video-streaming platforms. “These are not normal products for kids like a pair of shoes or a toy,” she says. “They create a relationship with kids.”

Here are four features that create that superglue:

Feature 1: solitude

“When the relationship is just between you and the machine, it removes social cues needed for stopping,” Schüll says. It’s harder to notice when the activity no longer serves the person playing or scrolling.

Studies have found that children who regularly use screens alone in their bedrooms have a higher risk of developing what psychologists call problematic usage. That is, they continue to use an app or play a game even when it damages their health. For example, the app may interfere with their sleep or friendships, but the child still feels compelled to stay on the app.

Feature 2: bottomlessness

Videos keep appearing on TikTok and YouTube. Photos, comments and likes keep popping up on Instagram. Apps have seemingly endless content for you to see, and it all shows or plays automatically.

“There’s no natural stopping point,” Schüll says. So you never feel finished or satisfied.

You want one more of something, endlessly. And that feeling grows even stronger with the third ingredient added into the mix.

Feature 3: speed

The faster people play video slots, the longer people gamble, Schüll found in her review of research performed by the gambling industry. Speed has a similar effect on social media and video-streaming apps, she says. The faster people can scroll, watch and then watch again, the harder it is for many to pull away from an app.

“The speed of the feedback can cause this sense that you merge with the screen. You don’t know where you begin and the machine ends,” Schüll says. “The speed really just pulls you into this flow.”

For social media, the speed at which we can find “new” material has jumped with several technological advancements, including the invention of higher-speed internet and infinite scroll.

Feature 4: teasing, or giving you almost what you want

The final ingredient is perhaps the most important, says Jonathan D. Morrow, a neuroscientist and psychiatrist at the University of Michigan. It’s all about how apps select content for you.

Here’s how it typically works. First, the software uses AI to determine what you’re hoping to find or see. “Even if you don’t know what you want, the app knows. It’s very good at figuring that out,” Morrow says.

But then, he says, the app withholds that reward: “Apps don’t give it to you. They give you something close to that, and then a few clicks later, the algorithm gives you something even closer.”

They rarely — if ever — give you what you’re looking for. “They give just enough to keep you engaged, keep you looking at the app and interacting with it as long as possible,” he adds.

This teasing gives you the feeling that you’re going to get what you’re seeking soon. “So you’ll be there all day trying to get that next big thing. There’s always a possibility you’ll finally get what you want,” Morrow says.

A recipe for overuse

When an app combines these four features — solitude, bottomlessness, speed and teasing — it creates a kind of recipe for overuse for nearly everyone, Schüll says. Sometimes Schüll gives her students at New York University this list of design features. “I say, ‘Pick a website or app. Then, using these criteria, rate how harmful it is.'”

But the recipe is especially harmful for children, she adds: “It’s a cruel setup, especially when kids are concerned. Kids are obviously more vulnerable.” Therefore, she and Morrow agree: Children need help regulating their use of these apps, but they also need protection from harmful design.

Michaeleen Doucleff has a Ph.D. in chemistry and is a longtime science journalist (including previously for NPR). She is the author of the parenting book Dopamine Kids.

The guilt of leaving my child, the anxiety of starting over at my job, the stress of managing both worlds at once: it was the ultimate case of the Mondays.

Transitioning from “worker” to “parent” to “working parent” can be a shock to the system, says Amy Beacom, founder and CEO of the Center for Parental Leave Leadership.

Parents are often “learning two new roles fast, under a lot of pressure with lack of sleep, zero guidance and zero support,” she says.

That’s especially the case for mothers. The United States is the only industrialized nation without federal paid leave, and one 2012 report found that 1 in 4 women go back to work within two weeks of having a baby.

While I was one of the lucky ones — I had several months of protected, paid leave — coming back was still a struggle and a huge adjustment.

So whether you’re about to return to work or already back, there are ways to set yourself up for success, Beacom says.

Make work work better for you

Too often, returning parents assume their jobs and schedules are set in stone, so they don’t ask their managers for what they want, Beacom says.

But for the most part, companies “want you happy, engaged and supported,” she says — so you may be surprised by what they agree to.

Beacom shares a few ideas to smoothen your reentry.

• What does a good first day, week and month look like? If possible, make a plan and propose it to your manager before you take your leave, “ so everyone is thinking about that reentry even before you go,” says Beacom.
• Set a date for a brief check-in with your employer while you’re gone, so both sides can feel more confident and prepared about what’s needed for your return, says Beacom.
• Plan to return mid-week so you can give yourself time to quietly warm up, clear your inbox and then come back full swing the following week.
• Negotiate your schedule. Maybe shifting your workday by half an hour would make a world of difference to your commute, or working East Coast hours would be a lot easier for day care pick-up and drop-off.

Your workplace will often only be as flexible as you ask them to be, Beacom says. So find how work can work better for you.

Let “good” be “good enough”

Now that you’re back at work, you might notice something surprising: Your job might feel a little easier.

“A lot of parents report that having kids, because it introduces all these new complexities, can make them more disciplined and better at getting things done,” says brain researcher Darby Saxbe, author of the upcoming book Dad Brain, about the science of fatherhood.

That doesn’t mean your life as a working parent will be a walk in the park. If you’re feeling overwhelmed, see when you can let “good” be “good enough.”

“My best advice to parents is lower your standards,” Saxbe says. “Don’t expect you’re going to be amazing at everything.” There are no raises for picture-perfect baby food or trophies for late-night speed-emailing.

If you absolutely must add more to your plate, be intentional about it, says Reshma Saujani, CEO of Moms First, an organization that advocates for affordable child care and paid leave. What is a need and what is just a gold star on your record? How much pressure is self-imposed, and how much is out of your control?

Lastly, cut yourself some slack. Parenting can change the brain and body to better meet the needs of caring for your little one — and research shows it can take anywhere from a few months to two years or more for a new parent to feel fully themselves again.

“Respect that this is a transformational time,” Saxbe says. “Be patient with yourself and recognize that you may not be 100% for a while.”

The digital story was edited by Meghan Keane. The visual editor is CJ Riculan. We’d love to hear from you. Leave us a voicemail at 202-216-9823, or email us at LifeKit@npr.org.

Listen to Life Kit on Apple Podcasts and Spotify, and sign up for our newsletter. Follow us on Instagram: @nprlifekit.

When you live with kids, your home is bound to get messy. And the more stuff they have, the more time you have to spend organizing it and cleaning it up.

That’s why psychotherapist and mom of two Denaye Barahona prefers a less-is-more approach when it comes to buying and keeping kids’ clothes, toys and gear.

When you “pare back on the amount of clutter in your life, you can focus on the most important things,” like quality time as a family, says Barahona, author of Simple Happy Parenting: The Secret of Less for Calmer Parents and Happier Kids.

Plus, there’s a positive benefit for young children. Research has shown that when toddlers play in a quieter space with fewer toys, they “do more, create more and innovate more,” she says.

So how do you begin the process of decluttering all your child’s things, whether it’s their beloved Pokémon card collection or handmade art? Barahona offers practical ways to organize what you have and decide what to keep or let go.

Create “active” and “storage” spaces to cut down on clutter: “Active” spaces are high-traffic areas of a room that should hold only things you use on a regular basis, says Barahona. An entryway closet, for example, doesn’t need to be stuffed with snow jackets in summer. Keep those in a separate storage space, like an underbed drawer, until you’re ready to rotate them back into the closet.

Use clear, shallow bins to help younger kids spot the toys they want. If you want your daughter to stop dumping every item she owns onto the living room floor to find that one special stuffie, this kind of storage is the way to go, says Barahona. Visible toys make for less mess.

Rotate toys in and out of play spaces. It can tamp down on visual clutter and give kids more space for focusing by reducing overstimulation from excessive options. Be warned, though: Maintaining a regular toy-rotation schedule can require a fair amount of extra effort and planning by parents, Barahona says.

Reframe the purpose of your family’s donation pile. You’re not losing something — you’re “sharing the love,” Barahona says. It’s just a small semantic change, but it helps her family picture an item’s future purpose. “Should this jacket spend the next 30 years in this box, or should it be on the body of another child who needs it?”

Beware of recluttering after decluttering. The goal of decluttering should be to live more simply, not make space for more stuff, says Barahona. Before you buy anything new — be it storage bins or toys — pause to consider your motivation: Is this a true need, a replacement or just a personal desire? If it’s the latter, try looking for alternatives that could fill that same need without accumulating more stuff. For example, could you share a new experience with your kid instead of buying them that toy?

Scale back to help kids feel more responsible. If your son is constantly leaving a tornado of toys wherever he goes, he might have too many, be too young to manage them — or both, Barahona says. If he loves Pokémon cards, for example, you might have him pick his top 20 to keep in an active drawer. Move the rest into storage until he can show you he’s capable of taking care of that first set.

You don’t have to keep all that kid art. Yes, every finger-painted picture frame is a masterpiece, but there’s only so much room for keepsakes in any family closet. In her home, Barahona has exactly one box for storing both of her kids’ artwork. Whenever it gets full, she looks through everything to see what still holds meaning and what draws a blank.

“Some things in the box, I’m like, ‘What is this? Who made this?'” Barahona says. If she doesn’t know, “then I am certainly not going to remember why I saved them 30 years from now.”

Your turn: How do you manage your kid’s stuff?

Tell us your best organization hacks (and feel free to share photos!). Email us at lifekit@npr.org with the subject line “Kid stuff.” We may feature your story on NPR.org.

Illustrations inked and colored by NPR’s Malaka Gharib

This story was edited by Malaka Gharib. The visual editor is CJ Riculan. We’d love to hear from you. Leave us a voicemail at 202-216-9823, or email us at LifeKit@npr.org.

Listen to Life Kit on Apple Podcasts and Spotify, and sign up for our newsletter. Follow us on Instagram: @nprlifekit.

That’s according to a new paper out in JAMA Psychiatry.

“We’re not saying that AI use is good or bad,” says Shaddy Saba, an assistant professor at New York University’s Silver School of Social Work, “just like we wouldn’t say substance use is necessarily good or bad, [or] consulting with a friend about something is good or bad.”

However, learning about a person’s use of AI for emotional support and advice could provide valuable insight into someone’s life and mental health status, he says.

“Our job is to understand why people are behaving as they are — in this case, why they are seeking help from an AI system,” adds Saba. “And to learn about what it’s doing for them, what it’s not doing for them.”

Saba and his co-author’s recommendations are “very aligned” with recommendations by the American Psychological Association (APA) in a health advisory released in November of last year, says the APA’s Vaile Wright.

Asking what a patient is getting out of their conversations with an AI chatbot sets “a foundation for the therapist to better know how they are trying to navigate their emotional wellbeing and their mental illness,” says Wright.

“Treasure trove of information”

“People are using these tools on a regular basis to ask about how to cope with stressful experiences, personal relationship challenges,” explains Saba. 

And some are using chatbots for advice on how to cope with symptoms of anxiety and depression.

“To the extent that we can prompt our clients to bring these conversations, in increasing detail, even into the therapy room, I think there’s potentially a treasure trove of information,” he says.

It could be information about the main causes of stress in someone’s life, or if they are turning to a chatbot as a way to avoid confrontations.

“Let’s say, for example, you have a client who is having relationship issues with their spouse,” says the APA’s Wright. “And instead of trying to have open conversations with their spouse about how to get their needs met, they’re instead going to the chatbot to either fill those needs or to avoid having these difficult conversations with their spouse.”

That background will help a therapist better support the patient, she explains.

“Helping them understand how to have a safe conversation with their spouse, helping them understand the limitations of AI as a tool for filling those gaps in those needs.”

Discussing use of AI is also a chance to learn about things a client might not voluntarily share with a therapist, says psychiatrist Dr. Tom Insel, former director of the National Institute of Mental Health. “People often use the chatbots to talk about things that they can’t talk about with other people because they’re so worried about being judged,” he says.

For example, suicidal thoughts may be something a patient is reluctant to share with their therapist, but that is critical for the therapist to know to keep the patient safe.

Be curious, but don’t judge

When it comes to first broaching the subject with patients, Saba suggests doing it without any judgment.

“We don’t want to make clients feel like we’re judging them,” he says. “They’re just not going to want to work with us in general if we do that.”

He recommends therapists approach the topic with genuine curiosity, and offers suggested language for these conversations.

“‘You know, AI is something that’s kind of rapidly growing, and I’m hearing from a lot of people that they’re using things like ChatGPT for emotional support,” he suggests. “‘Is that the case for you? Have you tried that?'”

He also recommends asking specific questions about what they found helpful so they can better understand how a patient is using these tools.

It could also help a therapist figure out whether a chatbot can complement therapy in helpful ways, says Insel, such as to vet which topics to bring to their sessions or to vent about day-to-day life.

In a way, therapy and chatbots “could be aligned to work together,” says Insel.

Saba and his co-author, William Weeks, also suggest asking patients if they found any chatbot interactions unhelpful or problematic, and also offering to share risks of using chatbots for emotional support.

For example, the risks to data privacy, because many AI companies use the conversations — even sensitive ones — to further train their models.

There are also risks of treating a chatbot like a therapist, says Insel.

Talking with a chatbot about one’s mental health is “the opposite of therapy,” he says, because chatbots are designed to affirm and flatter, reinforcing users’ thoughts and feelings.

“Therapy is there to help you change and to challenge you,” says Insel, “and to get you to talk about things that are particularly difficult.”

Adopting the advice

Psychologist Cami Winkelspecht has a private practice working primarily with children and adolescents in Wilmington, Del.

She has been considering adding questions about social media and AI use to her intake form and appreciated Saba’s study as it offered some sample questions to include.

Over the past year or so, Winkelspecht has had a growing number of clients and their parents ask her for help with using AI for brainstorming and other tasks in ways that don’t break a school’s honor code. So, she’s had to familiarize herself with the technology to be able to support her clients. Along the way, she’s come to realize that therapists and kids’ parents need to be more aware of how children and teens are using their digital devices — both social media and AI chatbots.

“We don’t necessarily think about what they’re doing with their phones quite as much,” says Winkelspecht. “And I think it’s pretty clear that we need to be doing that more and encouraging ourselves to have that conversation.”

When he was younger, he had his sights set on a professional music career as a guitarist. Then in the 1980s, he landed a long-term substitute teaching role in two very different schools within the same district: Hillcrest, which is in an affluent neighborhood in the Oakland Hills, and Whittier in East Oakland, which was an epicenter of the crack cocaine epidemic.

The stark disparity between the two schools challenged Porter’s perception of his own impact on the students who weren’t receiving the education they deserved. So, he decided to continue on with his teaching career and eventually landed permanent teaching positions throughout the district.

These days at Roosevelt Middle School, his classroom stands out. When most middle schoolers are learning the classics, Porter’s students dive deep into the world of jazz, even going so far as to play avant-garde compositions.

“We’re the only middle school band in the Local Supercluster, as far as I know, certainly in the Milky Way galaxy, that specializes in the music of Sun Ra,” said Porter. Sun Ra was an American jazz composer and band leader, known for his experimental music style and cosmic philosophy.

Porter truly believes in the musical ability of kids of all ages. When he was Berkeley Symphony’s director of music education, he would have the orchestra perform pieces composed by 5-year-olds. And in years past, he’s created opportunities for his elementary and middle school students to record their own albums – something that he’s doing for his current middle school students.

Porter’s the type of teacher who provides a safe space for his students to practice, hang out and have a snack after school. But this is Mr. Porter’s last year teaching because he is retiring.

At this juncture, the future of music classes in Oakland public schools is uncertain because of looming budget cuts across the district. Porter does not want arts education to fall to the wayside so he started a fundraiser for Roosevelt’s music program.

“[The arts] is an absolutely essential part of a kid’s development. Music, art, PE, manipulating things with your hands – this is how kids learn,” he said. “It’s how a lot of people learn. And when you take them away, a certain portion of the population gets a little bit left behind,” Porter said.

There are many benefits for students who study and play music. Research conducted by the University of Southern California’s Brain and Creativity Institute found that learning music enhances auditory pathways in the brain, which could help with other learning systems affected by these neural pathways like reading and language. In 2022, policy caught up with science when California voters passed Proposition 28, requiring the state to provide additional funding for music and arts programs for public schools.

For students like seventh grader Diego, Porter’s band class wasn’t a natural choice when he entered middle school. “It was so weird,” Diego said of jazz music. “I was like, ‘will people actually wanna listen to this?’ I didn’t even want to play it at first.” But, he marched forward.

“I hesitated and then I just stuck with it,” he said. “I like that there’s so many possibilities and different combinations so that you can make any different one.”

Another of Porter’s seventh grade students, Imani, who plays guitar, became interested in playing Sun Ra’s music in band class. “All the parts are so different and they all come together into chaotic bliss,” she said.

Some of Porter’s former students have gone on to new heights, like 10th grade student Ryan, who comes back to Porter’s classroom at Roosevelt every Thursday to mentor middle schoolers. When Ryan arrived in Porter’s classroom about five years ago, he had experience playing violin and had picked up the cello. But Porter’s jazz-filled band class presented something new.

“It just felt exciting to be in the music class…that’s when I started to think, ‘wait, I need to switch to an instrument that’s more suitable for jazz,’” said Ryan.

So, at the suggestion of Porter, Ryan picked up his third instrument – the trombone. “It really opened up a new world for me,” said Ryan.

Now, he plays a total of fifteen instruments, is a member of the SFJAZZ High School All-Stars Band, and has played in a youth orchestra for three years, all at the encouragement of Porter.

Middle school students are in an age group that is notorious in schools and among teachers for their unpredictability, high energy, and increased social awareness. But if you can tap into their interests, the potential for growth is what Porter finds most exciting about this age group, he said.

He plans on being an active member of the local music education community in his retirement, but the students are what he’ll miss most. “I wanna be helpful. I wanna mentor teachers. I want to do what I can just to see things continue to be successful,” he said.

But he’ll also take time to tap back into the professional music world. This summer you can find Porter playing at one of his annual gigs, Chapel of the Chimes.

Some studies have found that chatbot tutors can backfire because students lean on them too heavily, get spoonfed solutions and fail to absorb the material. Even when AI tutors are designed not to give away answers, they haven’t consistently produced better results than learning the old-fashioned way without AI.

Still, researchers who have produced these skeptical studies haven’t given up hope. Some are still experimenting, trying to build better AI tutors.

One promising idea has less to do with how an AI tutor explains concepts and more with what it asks students to practice next.

A team at the University of Pennsylvania, which included some AI skeptics, recently tested this approach in a study of close to 800 Taiwanese high school students learning Python programming. All the students used the same AI tutor, which was designed not to give away answers.

But there was one key difference. Half the students were randomly assigned to a fixed sequence of practice problems, progressing from easy to hard. The other half received a personalized sequence with the AI tutor continuously adjusting the difficulty of each problem based on how the student was performing and interacting with the chatbot.

The idea is based on what educators call the “zone of proximal development.” When problems are too easy, students get bored. When they’re too hard, students get frustrated. The goal is to keep students in a sweet spot: challenged, but not overwhelmed.

The researchers found that students in the personalized group did better on a final exam than students in the fixed problem group. The difference was characterized as the equivalent of 6 to 9 months of additional schooling, an eye-catching claim for an after-school online course that lasted only five months. The AI tutor’s inventor, Angel Chung, a doctoral student at the Wharton School, acknowledged that her conversion of statistical units was “not a perfect estimate.” (A draft paper about the experiment was posted online in March 2026, but has not yet been published in a peer-reviewed journal.)

Still, this is early evidence that small tweaks — in this case, calibrating the difficulty of the practice problems to the student — can make a difference.

Chung said that ChatGPT’s responses may already feel very personal because they are directly responding to a student’s unique questions. But that level of personalization isn’t enough. “Students usually don’t know what they don’t know,” said Chung. “The student doesn’t have the ability to ask the right questions to get the best tutoring.”

To address this, Chung’s team combined a large language model with a separate machine-learning algorithm that analyzes how students interact with the online course platform — how they answer the practice questions, how many times they revise or edit their coding, and the quality of their conversations with the chatbot — and uses that information to decide which problem to serve up next.

How different students interact with the chatbot tutor

In other words, personalization isn’t just about tailoring explanations. It’s about tailoring the learning path itself.

That idea isn’t new.

Long before generative AI tools like ChatGPT were invented, education researchers developed “intelligent tutoring systems” that tried to do something similar: estimate what a student knew and deliver the right next problem. These earlier systems couldn’t produce  natural conversations, but they could provide hints and instant feedback. Rigorous studies found that well-designed versions helped students learn significantly more.

Their Achilles’ heel was engagement. Many students simply didn’t want to use them.

Today’s AI tools could help address that problem. Students might feel more interested in a chatbot that converses with them in an almost human way.

In the University of Pennsylvania study, students in the personalized group spent more time practicing, about three additional minutes per problem, adding up to about an hour per module in the Python course, compared with half as much time (a half hour or less) for the comparison students. The researchers think these students did better because they were more engaged in their practice work.

Students’ previous knowledge of a subject affected how well the personalized sequencing worked. Students who were new to Python gained more than those who already had Python experience, who did just as well with the fixed sequence of practice problems. Students from less elite high schools also appeared to benefit more.

How students’ background affected results

All the Taiwanese students in this study volunteered for an optional computer programming course that could strengthen their college applications. Many were highly motivated, with highly educated parents, and many already had prior coding experience.

It’s not clear whether the chatbot would work as well with less motivated students who are behind at school and most in need of extra help.

One possible solution: fusing new and old.

Ken Koedinger, a professor at Carnegie Mellon University and a pioneer of intelligent tutoring systems, is experimenting with using new AI models to alert remote human tutors who can motivate struggling students who are drifting off. “We are having more success,” said Koedinger.

Humans aren’t obsolete — yet.

This story about AI tutors was produced by The Hechinger Report, a nonprofit, independent news organization that covers education. Sign up for Proof Points and other Hechinger newsletters.

“Never once did I even think that I needed to say ‘gambling,'” she recalls.

One day, when her oldest son was 11, he was watching someone play video games on a livestream and clicked on a link in the comments. It took him to an offshore online casino.

There, he got sucked in — to blackjack, poker, roulette. He could use items from the video game as money. Soon he got hooked, but the signs of his addiction were hard to spot.

“It’s not like he was just holed up in his room 24-7,” Freudenberg says. “He ran track. He played soccer. He was a great student.”

Until he dropped out of college at age 19. That’s when his mom found out that he had been gambling for nearly half his life.

He’d sold things from around the house to keep up with his debts, borrowed money from friends and, then, eventually, started stealing money from his parents.

It’s a problem that educators, researchers and parents like Freudenberg say is affecting a growing number of young people, most of them boys. A recent national survey from Common Sense Media found that 36% of boys age 11 to 17 in the U.S. have gambled in the past year.

“It’s a lot of kids,” says Michael Robb, the head of research at Common Sense Media, a nonprofit that promotes digital safety for kids. “A third of kids is a lot of kids.”

He notes that playing fantasy football with friends or making a March Madness bracket may be harmless. It could, for example, help strengthen male friend groups. But for a small subset of boys, Robb adds, things can get out of control: “They’re not all going to have problems. But given how much things have changed in the last couple of years, the way [some kids] are engaging in gambling behaviors is already flashing red signs.”

It’s not just teens. Gambling has soared in the U.S. since a key Supreme Court ruling in 2018 allowed states to legalize sports betting. That opened the floodgates, from one state back then to 38 in 2024.

Before that decision, Americans spent $4.9 billion annually on sports betting. By 2023, that figure had ballooned to $121 billion, according to The Journal of the American Medical Association (JAMA).

And those were just the legal bets. No one under 18 can gamble legally, but experts say the opportunities are everywhere.

“If I wanted to bet on the [Washington] Nationals,” says Matt Missar, an addiction counselor in Pittsburgh, “20 years ago, as a teenager, I’ll go find a bookie and I’ll place a bet. Nowadays, I can bet on every single pitch of a game.”

Much of the explosion in legalized gambling is happening on cellphones, Missar notes. “It is incredibly easy.”

He specializes in gambling and video game addictions and says the number of young adults he sees come through his practice has ticked higher in recent years.

“It’s not just that the problem arose when they’re 18,” he says. “It started when they were 13 or 14 … and slowly over those years it became more of a problem.”

Freudenberg wishes she had seen the warning signs. But often, she says, online gambling can look the same as texting a friend or watching a video.

She thinks removing the guardrails has created a slippery slope for kids: “If my kid had to get in a car, drive to a bank, take out money, drive to a casino, go into the casino, show an ID at the door — he probably wouldn’t be a gambling addict.”

After a few attempts at rehab, she says, her son is back at college and doing well. Freudenberg helped start a support group for parents of teen gamblers, and their numbers are growing.

She fears that, all over the country, there are lots more parents just like her.

“The tsunami is on the horizon,” she says. “And it’s gonna be really, really bad.”

The Trump administration has argued the “privilege” has been too freely applied to children of non-citizens. “Hundreds of thousands of people are pouring into our country under birthright citizenship, and it wasn’t meant for that reason,” President Trump said last year.

If this constitutionally protected right is struck down by the court, it would apply to children born on or after Feb. 20, 2025. According to a projection by the nonpartisan Migration Policy Institute and Penn State, that could result in about 255,000 U.S.-born children beginning life without U.S. citizenship every year. By 2045, that could add up to 4.8 million children.

“Birthright citizenship is fundamental for child wellbeing,” says Wendy Cervantes of The Center for Law and Social Policy, a nonpartisan organization focused on helping people with low incomes. “It has helped ensure that all children in the U.S. can start off life with some sort of equal footing and opportunity.”

Much of that equal footing comes courtesy of the country’s K-12 public schools. While schools are a place for children to learn, they’re also a central access point for a range of services: free meals, mental health support, services for students with disabilities and much more. Without the right to citizenship, access to those services could be complicated for many children – as could access to college.

Here’s what to know about how a Supreme Court ruling to end or narrow birthright citizenship could change the education landscape.

Public schools can’t turn students away because of their immigration status

All children, regardless of immigration status, have the right to a free K-12 public education in the United States. That right was affirmed in the landmark 1982 Supreme Court ruling Plyler v. Doe.

The case centered on whether Texas could prohibit the use of state funding to educate children who were living in the U.S. unlawfully. Also in question was whether a public school district could charge foreign-born students tuition to enroll. Immigrant students sued and prevailed.

In Plyler , Cervantes says, “It was recognized by the justices that denying a K-12 education to children, a basic education, would create a permanent underclass in our society.”

Because of this decision, school districts are not supposed to collect immigration data on their students or their families. But immigrant advocates worry that Plyler has become a political target.

“The conservative movement has made very clear their intention to overturn Plyler v. Doe by even providing a playbook to state legislatures to help make that happen,” says Alejandra Vázquez Baur, co-founder and director of the National Newcomer Network, which advocates for recently arrived immigrant students.

The Heritage Foundation, the conservative think tank behind Project 2025, which has shaped much of the Trump administration’s agenda, recently called for states to restrict public education for undocumented students and has recommended that states directly challenge the Plyler decision, arguing that it cost states hundreds of millions of dollars in education spending in 2023 alone.

“States have a convincing interest in preserving limited taxpayer dollars by prioritizing U.S. citizens and lawful immigrants,” wrote Lora Ries of Heritage.

Tennessee lawmakers are among those taking action: There are currently bills moving through the state legislature that propose tracking K-12 students’ legal status and allowing public schools to refuse to enroll undocumented students. Several other states have also proposed legislation that directly, or indirectly, threaten Plyler.

If any of these proposals turn into laws, they could invite legal challenges, and ultimately re-open the question of whether immigrant children have the right to a public education.

A right to education doesn’t mean families feel safe sending their kids to school

Immigration enforcement efforts can take a toll on school attendance.

MPR News reported that after heightened federal immigration presence in Minnesota early this year, some districts experienced a 20-40% increase in absences. And that trend predates the Trump administration: Researchers at the Hoover Institution, a conservative think tank based at Stanford University, found that after immigration raids in January 2025, school districts in California’s Central Valley had a 22% increase in absences.

Vázquez Baur says these findings show immigrant children’s constitutional right to attend K-12 public schools is already under threat.

“The law is still the law, children can still go to school. Now, we know that that is being complicated at this moment by immigration enforcement around schools,” she says. “The birthright citizenship issue complicates that even further.”

Sophia Rodriguez, a professor of education policy at New York University, has been studying the impact of immigration enforcement on school attendance. She says she has heard reports of “constant fear, anxiety and stress” from immigrant families concerned about sending their children to school. “And when you add this potential end to birthright citizenship, you create larger numbers of communities who are living in fear and anxiety,” she says.

Some studies have shown that, historically, when there is a rise in local immigration enforcement, fewer Hispanic students enroll in nearby schools, which can disrupt their education and affect school funding. In most states, public school districts receive funding based on daily student attendance and overall enrollment.

This comes as many school districts are already facing enrollment declines.

Students with disabilities could fall through the cracks

For many children, schools are the first point of contact with public services such as nutrition programs, healthcare, language learning and counseling. That is especially the case for immigrant families, says Rodriguez of NYU. “[Schools] are often the one social institution or public institution that immigrant families access.”

They are also often the first place children’s disabilities are identified, and where those students can tap into the services they need to be successful. The Individuals with Disabilities Education Act (IDEA) is the central special education law that guarantees all disabled children the right to a “Free Appropriate Public Education.”

“So those are things that aren’t going away or changing based on immigration status,” says Anne Dwyer, a professor at Georgetown University’s Center for Children and Families. “But if a community is experiencing immigration enforcement or fear of enforcement at such a level that parents don’t even feel comfortable bringing their children to school, then those children are automatically not going to be able to access those very supports that schools provide.”

Schools also rely on state and federal Medicaid dollars to pay for services like physical, speech and occupational therapy. The program covers about half of all students with special education plans according to an analysis by KFF, a nonpartisan health policy research organization. Medicaid funding also makes up a significant portion of public school budgets: The U.S. Education Department reported in 2024 that Medicaid sends schools between $4 billion and $6 billion annually.

“Even if a school can’t potentially provide a type of service, they’re probably going to be a broker to those resources,” says Rodriguez.

However, Medicaid is typically limited to U.S. citizens and people with other qualifying legal statuses. If birthright citizenship is eliminated, U.S.-born children who would have previously been citizens may no longer qualify for Medicaid. For any of those children who have disabilities, schools would still be legally obligated to serve them under IDEA, but they would have to find a way to replace the lost Medicaid funding.

“That would create potential, huge cost shifts to districts,” says Dwyer. “And we know school districts are already incredibly strapped.”

Paying for higher education would get a lot harder

While the law currently provides a K-12 education for all students, the same is not true of higher education. Students without legal status can still enroll in college, but they don’t have access to federal financial aid, such as federal student loans and the Pell Grant, which helps low-income students and is currently facing a funding shortfall.

And because of their status, undocumented students are also more likely to come from impoverished backgrounds, says Caitlin Patler, a professor of public policy at UC Berkeley. “Those two things together make affording higher education almost impossible for children who are undocumented.”

In some states, including Georgia and Alabama, undocumented students are not allowed to attend certain public colleges; other states charge them out-of-state tuition.

Patler says research shows U.S. citizenship is directly tied to opportunities that increase a child’s educational attainment. “And therefore much later on, as you follow children throughout their lives, educational attainment is directly correlated with stronger economic contributions.”

She worries about a future in which birthright citizenship is narrowed or eliminated. “This would have a cascading ripple effect, potentially through multiple generations, of forcing this large and growing group of millions of children into a caste-like status.”

A caste-like status, she says, in which their opportunities would be dictated not by their potential, but by their immigration status.

Edited by: Nicole Cohen
Visual design and development by: LA Johnson

Federal education research — the system that tracks student learning and evaluates what works — has been battered by mass firings, contract cuts and cancellations, and stalled grant funding. Many researchers at private research organizations have lost their jobs and those with a more protected perch at universities face deep uncertainty. Now they are being told they need to turn up the volume if they want to continue their life’s work.

Their predicament was the focus of the Association for Education Finance and Policy’s annual conference earlier this month in Chicago. The conference theme, “Sustaining Education Research and Evidence in a Turbulent Era,” acknowledged the devastating aftershocks of last year’s onslaught. But the cure remains uncertain. At a March 20 session on rebuilding the Institute of Education Sciences (IES), an emissary from the Trump administration, Amber Northern, urged the audience to become stronger champions for their cause.

A year ago at this same conference, Northern was just a typical researcher, as horrified as everyone else over the DOGE cuts to federal education research. She was and is the director of research at the Thomas B. Fordham Institute, a conservative education policy think tank. During last year’s gathering, a sympathetic official from the Trump administration approached her and asked if she could come up with some ideas for rebuilding IES, which has generally had bipartisan support.

This year, Northern was at the conference in her new role as the author of a report on IES’s future, released in late February, and was making the rounds to sell its recommendations.

Her main message to her fellow researchers: You’re not doing enough.

Rebuilding IES won’t happen, she warned, without broad public pressure. The administration, she said, responds to parents, but parents aren’t protesting the loss of education data and research. She added she was “dismayed” that more people in the field haven’t written op-eds explaining the stakes.

The room pushed back. Many researchers were still smarting from the loss of federal research funding and the inability to seek new grants. (The grant process has ground to a virtual standstill and the Education Department is sitting on millions of dollars of unspent Congressionally appropriated funds.)

Jason Grissom, an education professor at Vanderbilt University, said he had just received an email that federal funding for his graduate students was ending. He said he hadn’t realized the field hadn’t been making “a strong enough case.”

But Vivian Wong, a research methodologist at the University of Virginia, challenged the idea that it would be realistic to build a broad coalition. “You can’t put the onus on parents to save the education system,” she said, noting that families are more focused on immediate concerns like services for their children with disabilities. Producing evidence for effective instruction, she argued, is the job of good government and shouldn’t hinge upon parent advocacy.

Others raised a more personal risk: speaking out could backfire. One researcher worried that public criticism could jeopardize current grants, future funding decisions, or even invite retaliation against her university at a time when the administration has shown a willingness to lash out. She asked Northern directly whether she could guarantee that advocacy for education research wouldn’t come with consequences.

“I can’t say for sure,” Northern replied.

And that’s the bind. Researchers are being told to speak up to save their field but doing so could put their work, and their institutions, at risk.

Another possible lever is Congress. Some researchers have begun lobbying their representatives, but even there, the path is unclear. One Congressional office advised contacting the Office of Management and Budget — not the Education Department — to release already appropriated funds.

Meanwhile, schools are struggling with absenteeism and falling reading and math scores. And the nation’s main source of evidence and guidance on what works to right these problems is in limbo.

Researchers did receive one reprieve. Despite inflation, the Association for Education Finance and Policy said it did not raise this year’s conference registration fee “in response to the challenges our community is facing.”

This story about federal education research was produced by The Hechinger Report, a nonprofit, independent news organization that covers education. Sign up for Proof Points and other Hechinger newsletters.

“I’d see the same phrases, the same commas, even the same word choices. I would say, ‘Man, I’ve read that before.’ And I’d go look for it,” said Maxwell. “The paragraphs weren’t identical, but they were so similar.”

Although the course was in 2024, Maxwell, who teaches at Northeastern’s Seattle campus, recalls that his students’ essays sounded “like textbooks written in the 1980s and ’90s,” perhaps reflecting the sources used to train AI. The students were scattered around the country and Maxwell was pretty sure they hadn’t collaborated.

Maxwell shared his observation with a former student, Liwei Jiang, who is now a Ph.D. student in computer science and engineering at the University of Washington. Jiang decided to test her former professor’s hunch about AI scientifically and collaborated with other researchers at UW, the Allen Institute for Artificial Intelligence, Stanford and Carnegie Mellon universities to analyze the output from more than 70 different large language models around the globe, including ChatGPT, Claude, Gemini, DeepSeek, Qwen and Llama.

The team asked each the same open-ended questions, which were intended to spark creativity or brainstorm new ideas: “Compose a short poem about the feeling of watching a sunset;” “I am a graduate student in Marxist theory, and I want to write a thesis on Gorz. Can you help me think of some new ideas?” and “Write a 30-word essay on global warming.” (The researchers pulled the questions from a corpus of real ChatGPT questions that users had consented to make public in exchange for free access to a more advanced model.) The researchers posed 100 of these questions to all 70 models and had each model answer them 50 times.

The answers were often indistinguishable across different models by different companies that have different architectures and use different training data. The metaphors, imagery, word choices, sentence structures — even punctuation — often converged. Jiang’s team called this phenomenon “inter-model homogeneity” and quantified the overlaps and similarities. To drive the point home, Jiang titled her paper, the “Artificial Hivemind.” The study won the best paper award at the annual conference on Neural Information Processing Systems in December 2025, one of the premier gatherings for AI research.

To increase AI creativity, Jiang jacked up a parameter, called “temperature,” to maximize the randomness of each large language model. That didn’t help. For example, when she asked an AI model called Claude 3.5 Sonnet to “write a short story about a colorful toad who goes on an adventure in 50 words,” it kept naming the toad Ziggy or Pip, and oddly, a hungry hawk and mushrooms kept appearing.

Different models also churn out comically similar responses. When asked to come up with a metaphor for time, the overwhelming answer from all the models was the same: a river. A few said a weaver. One outlier suggested a sculptor. Several of the models were developed in China, and yet, they were producing similar answers to those made in America.

Example of similar output from ChatGPT and DeepSeek

The explanation lies in chatbot design. AI chatbots are trained to review possible answers to make sure the output is reasonable, appropriate and helpful. This refinement step, sometimes called “alignment,” is intended to ensure that the answers align to or match what a human would prefer. And it’s this alignment step, according to Jiang, that is creating the homogeneity. The process favors safe, consensus-based responses and penalizes risky, unconventional ones. Originality gets stripped away.

Jiang’s advice for students is to push themselves to go beyond what the AI model spits out. “The model is actually generating some good ideas, but you need to go the extra mile to be more creative than that,” said Jiang.

For Jiang’s former professor Maxwell, the study confirmed what he had suspected. And even before Jiang’s paper came out, he changed how he teaches. He no longer relies on online exams. Instead, he now asks students to learn a concept and present it to other students or create a video tutorial.

Outwitting the AI hive mind requires some post-modern creativity.

This story about similar AI answers was produced by The Hechinger Report, a nonprofit, independent news organization that covers education. Sign up for Proof Points and other Hechinger newsletters.