One in five students used AI for emotional needs

 Young people leaning on AI chatbots for emotional support or personal advice are significantly more likely to struggle with serious emotional problems, feel isolated, and doubt that they truly matter to others, a new uOttawa study warns.

A new study examined responses from 39,761 students in grades 4 through 12 across four Ontario school boards. These students filled out surveys between May 2025 and March 2026 as part of the Ontario Health and Peer Relations Study.

Out of everyone surveyed, 8,408 students (about 21%) said they had turned to an AI chatbot for help with their feelings or for advice about relationships, whether with friends, family, or others. The tendency to seek emotional support from AI was higher among older students, as well as among racialized and gender-diverse youth compared to their white or male peers.

Kamilla Bonnesen, a postdoctoral fellow at the Brain and Behaviour Lab at the University of Ottawa, explains, “Of almost 40,000 students, those who leaned on AI for emotional support were much more likely to be struggling with clinical-level emotional issues, more loneliness, and a lower sense of mattering.”

Bonnesen also points out: “This doesn’t mean that AI is causing these struggles. Instead, using AI in this way might be a sign that a young person needs more human connection and support.”

Emotional use stands apart from schoolwork

The study found that students who used AI a lot for schoolwork were also more likely to use it for emotional reasons. Among students who sought emotional support from AI, 57.7% met the study’s criteria for serious emotional problems, compared with 29.2% of those who didn’t use AI in this way. These students also reported feeling lonelier and less valued by others.

Students completed anonymous, school-based surveys that covered their AI use, emotional health, loneliness, and sense of mattering. The researchers used advanced statistical models to account for the fact that students were grouped within schools and to handle any missing answers. Even after accounting for things like demographics, loneliness, mattering, and academic use of AI, using AI for emotional support was still linked to a 26% higher prevalence of clinical-level emotional difficulties.

Bonnesen co-conceptualized the study with Tracy Vaillancourt, Canada Research Chair in Youth Mental Health and Violence Prevention and Full Professor in the University of Ottawa’s Faculty of Education. Bonnesen also led the statistical analysis and wrote up the results, including the handling of missing data and interpreting what they found.

An invitation to reconnect

“As chatbots become part of young people’s daily lives, parents, educators and mental health professionals shouldn’t just ask if youth are using AI, but why,” says Professor Vaillancourt.

“When a student is turning to AI for emotional reasons, it’s a good opportunity to start an open, nonjudgmental conversation,” Vaillancourt adds. “It is a chance to reconnect a struggling student with meaningful human support.”

Since the study was cross-sectional, it shows associations but can’t prove whether distress drives youth to AI, or whether using AI might make distress worse. Still, the results suggest that turning to AI for emotional support is a sign worth paying attention to, even if it doesn’t prove cause and effect.

The study, titled “Affective Generative Artificial Intelligence Use and Youth Mental Health,” was published in JAMA Pediatrics. Authorship: Kamilla Bonnesen, PhD; Amanda Krygsman, PhD; Sarah Hobson, MA; Daphne Korczak, MD, MSc; Tracy Vaillancourt, PhD.

---

Journal

JAMA Pediatrics

DOI

10.1001/jamapediatrics.2026.3904

Social media could help parents take actions to prevent teen vaping

  Parents are an important line of defense against teen vaping, yet many lack the knowledge and tools needed to recognize vaping products, understand their risks and prevent their children from using them. A new University of Oklahoma study will examine whether targeted social media messages can give parents the information they need to help prevent their children from vaping nicotine and cannabis.

The research, funded by a five-year, $2.5 million grant from the National Institutes of Health, will be led by Katelyn Romm, Ph.D., and Erin Vogel, Ph.D., members of the TSET Health Promotion Research Center (HPRC) at OU Health Stephenson Cancer Center and assistant professors of pediatrics in the OU College of Medicine. The study will enroll 1,000 parent-adolescent pairs and will begin with surveys to assess what parents know about vaping.

“Data suggest that less than half of parents are aware of whether their kids have vaped and what kinds of substances their kids may be vaping,” Romm said. “Many parents don’t realize vapes contain nicotine or that they can contain cannabis. We know that parents can serve as a key prevention tool when it comes to cigarettes and alcohol, but it’s harder to detect when kids are vaping. Vapes are designed to look like highlighters, lip gloss and other items, and vaping devices continue to evolve.”

Once parental knowledge is assessed, researchers will examine whether parents with greater knowledge about vaping are more likely to use prevention strategies such as setting rules, monitoring their children’s behavior and talking with them about vaping. The research team will also examine whether those parenting behaviors are associated with lower rates of vaping among adolescents.

The researchers will then test whether targeted social media messages can increase parents’ knowledge and help them take action. About 300 parents who initially demonstrate lower levels of vaping knowledge will participate in a five-week online experiment in which they view Facebook messages about vaping devices, health risks, teen vaping and youth-targeted marketing. To create the messages, the research team is partnering with the national organization Parents Against Vaping.

Researchers will test whether exposure to those messages increases parents’ knowledge and use of prevention strategies and whether those changes are associated with lower levels of vaping among their children.

“Nearly 90% of parents report using social media, primarily Facebook, including for parenting advice and support,” Romm said. “That’s why we wanted to leverage social media to expose parents to messages that might help them identify vaping products and understand the harms associated with them. We want to boost parents’ confidence and effectiveness in being able to have conversations with their kids.”

If the messages are effective, the researchers hope to work with public health organizations to develop larger-scale social media campaigns that could influence parents across the country. Because social media campaigns can reach large audiences at relatively low cost, the approach could provide a practical way to support parents in preventing youth vaping.

Although nicotine vaping among youth has declined from its peak in 2019-2020, adolescents who continue to vape are reporting more frequent use.

“Young people who vape are reporting more days of use, and over a third who vape nicotine also report vaping cannabis or THC, which comes with additional health risks,” Romm said.

Vaping is associated with a number of health risks, including lung problems from substances that are not safe to inhale. Other risks include increased blood pressure and heart rate, susceptibility to addiction and an increased likelihood of using other tobacco products like cigarettes.

“The adolescent years are an important time to prevent nicotine and cannabis use from becoming established,” Romm said. “We want to give parents the information and tools they need to recognize vaping and help prevent its use with their children.”

Graduate Student Lending After the Elimination of GradPLUS

 The One Big Beautiful Bill Act (OBBBA, P.L. 119-21) eliminates the federal GradPLUS loan program for new borrowers effective July 1, 2026, replacing nearly two decades of uncapped federal graduate lending with annual limits of $20,500 ($50,000 for eleven professional fields) and new aggregate caps. The resulting financing gap — approximately $8 billion annually across roughly 370,000 affected borrowers — raises the question of whether and how private capital will substitute for displaced federal lending.

This paper analyzes the economics of that transition. Before GradPLUS, private lenders filled the gap between federal Stafford limits and graduate program costs, with nonfederal graduate originations reaching $5.3 billion (constant 2024 dollars) by 2005–06. 

After GradPLUS extended federal borrowing to the full cost of attendance in 2006, private lending collapsed — falling more than 70 percent within a few years and never recovering. The federal program provided not only credit but, through income-driven repayment and Public Service Loan Forgiveness, insurance against earnings risk that no private product could replicate. The same features that crowded out private lending generated compounding distortions: borrowing untethered to repayment capacity, upward pressure on tuition, and a reversal from projected federal surpluses to losses exceeding thirty cents per GradPLUS dollar lent.

Linking institution-by-field federal borrowing data to institution-by-field earnings outcomes, this study examines the relationship between historical borrowing above the new OBBBA caps and post-completion earnings across programs. 

The author finds substantial variation in the extent to which different degree programs would likely support above-cap borrowing from private lenders or other sources. In fields such as law and MBA, earnings rise steeply with borrowing exposure, providing a basis for program-level risk pricing; in fields such as social work, counseling psychology, and physical therapy, the earnings-to-borrowing gradient is essentially flat. 

The study also examines features of the private capital market and institutional responses shaping the transition — including capital constraints on scaling private lending, legal barriers to using program-level outcomes data in underwriting, and the emerging role of institutions as intermediaries between lenders and students.

Private Tutoring Centers and Student Achievement

 Private tutoring chains like Kumon and Sylvan have expanded rapidly in recent decades. Combining data on business locations, household spending, and district test scores, this study looks at who uses these centers and how they affect academic skills. 

Nearly half of tutoring users are from the top income decile, with Asian and White households five and two times as likely to purchase tutoring as Black and Hispanic households. 

A staggered differences-in-differences design shows the opening of a district's first chain center raises district-wide math and reading scores by 0.02-0.03 standard deviations, at a much lower cost than public spending typically producing similar gains. 

Math effects are largest for White and Asian students, so that center entry widens racial achievement gaps, though less so in more racially integrated districts. 

These private investments help their users, deepen educational inequality, and shape outcomes by which state accountability systems and families judge schools.

Can AI replace traditional language learning? A new study says not yet

When university students set out to learn English as an additional language, it’s not just about internalizing a new list of words: they also have to learn common word combinations. 

In order to sound fluent and natural, they’ll need to say “strong wind” but not “muscular wind,” or “reach a conclusion” but not “pull a conclusion” - a task that’s particularly challenging to learners who are stepping into terminology-heavy academic fields from economics to engineering. 

So what’s the best way to learn those invaluable combinations, also known as collocations? According to a new paper from the UBC Sauder School of Business, data-driven learning, with a little boost from AI, is still the most reliable option. 

According UBC Sauder lecturer Dr. Déogratias (Deo) Nizonkiza, the author of the study, the development of corpora — large, structured collections of texts and other language data — has been foundational for DDL approaches. It has happened in three major stages. 

In 1961, a groundbreaking collection of texts called the Brown Corpus was developed at Brown University, and comprised 500 text samples from media, religion, fiction, science and other sources, each one containing approximately 2,000 words. At roughly one million words, it represented the first large-scale collection of real-world American English, and it allowed researchers to effectively catalogue and analyze the language and its usage.  

In the 1980s, as technology advanced and computing became more widely available, the concept of data-driven learning, or DDL, was introduced. Instead of memorizing phrases from textbooks, learners could access increasingly massive databases of texts — with word counts eventually exceeding 100 million — to discover everyday language patterns. 

In 2008 came the Corpus of Contemporary American English, or COCA — a one billion-word database that comprises nearly 500,000 texts from 1990 to 2019 and allows students not only to search for specific words, but to look for examples in different contexts and academic disciplines.  

“That really significantly changed how we do things, specifically in the area of English for academic or professional purposes,” explains Dr. Nizonkiza. “But the searches took so much time, and people would get discouraged.”  

In addition to the sluggish searches, another stumbling block was that many language instructors didn’t even know the corpora existed, let alone how to use them, and those who did often treated DDL as a secondary activity — not a core teaching tool. 

Now with the advent of generative AI, or GenAI, says Dr. Nizonkiza, the process of searching for collocations has become far faster and more familiar — but that doesn’t mean educators should toss out the more traditional DDL approach.  

While GenAI tools such as ChatGPT can replicate some of the functions of the existing tools — it can identify and generate examples of common word combinations, for example — it’s often unclear where the examples come from, or how accurate they are.  

For the French-language study, whose title translates to At the Intersection of Corpora and Artificial Intelligence: A Critical Review of DDL Approaches to Teaching Collocations in Higher Education, Dr. Nizonkiza examined empirical evidence, including meta-analyses and systematic reviews, on the effectiveness of DDL.  

The literature shows that a hybrid approach that’s rooted in corpus-based teaching, and gets an added boost from GenAI, is the most effective.  

"We have to be very careful, because GenAI tools aren’t as precise as those traditional tools,” explains Dr. Nizonkiza. "So instead of relying perhaps on GenAI, we need to make sure that there is a combination of GenAI and the traditional corpora." 

The paper also provides a possible roadmap for that hybrid approach. For example, an English instructor could select a particular vocabulary — from the Economics Academic Word List, say — then identify common collocations like “control volatility”, “market volatility,” etc… From there, they can ask corpus-informed GenAI tools to generate examples of those collocations, which can be used to design structured exercises where students first fill in the blanks (e.g., “Portfolio diversification is a key strategy used by fund managers to ______ volatility in emerging markets”), then use AI to verify their answers. 

So rather than relying on popular AI platforms like ChatGPT or Copilot to generate examples, students complete the exercises — and only then use AI to verify and refine their conclusions.  

That process helps students to raise awareness of collocations, improve accuracy, and spur critical interactions with GenAI tools, says Dr. Nizonkiza. 

“It’s when they start university that they have exposure to these new words they haven't necessarily encountered in their everyday lives,” says Dr. Nizonkiza. “That's where having these tools, and being aware of what these tools can help them do, becomes essential.” 

---

Journal

Nouvelles perspectives en sciences sociales

DOI

10.7202/1126572ar 

 

Exposure to suicide attempts within one’s social network = subsequent suicide attempts

 Full report

Also see: https://www.nber.org/papers/w35603?utm_campaign=ntwh&utm_medium=email&utm_source=ntwg14

The mental health of adolescents and young adults has been described as a public health crisis, with escalating rates of depression and anxiety¹ and high rates of suicide thoughts and behaviors.² Research on suicide has documented contagion effects associating exposure to suicide thoughts and behaviors in one’s social network with subsequent suicidal ideation and attempts.³ These effects have been explained using models including media copycat effects,⁴ social learning,⁵ and assortative relating⁶ and have been supported by both cross-sectional and longitudinal studies.^7,8

A systematic review examined research on exposure to suicide and subsequent suicidal behavior, with the overall evidence suggesting elevated risk among exposed individuals, although findings varied across study designs and outcomes.⁹ However, findings have not always been consistent regarding whether exposure is mainly associated with the emergence of suicidal thoughts, with the transition from thoughts to behaviors, or both. For example, Bearman and Moody¹⁰ found that exposure to friends’ suicide attempts was associated with suicidal ideation among adolescents, whereas evidence regarding progression to suicide attempts has been less consistent. Studies have also varied considerably in design, developmental period examined, exposure measurement, and the extent to which they accounted for individual mental health risk factors and broader social contexts.¹¹ Accordingly, reviews have emphasized the need for longitudinal studies that distinguish between suicidal ideation and suicide attempt outcomes and examine how social network characteristics shape risk over time.⁹

Suicide prevention practices, especially in communities of young people, often mobilize strategies to minimize contagion.¹² However, contagion effects are not distributed equally across young people. Communities facing higher levels of risky social determinants of health and young people identifying with groups that experience disproportionate adversities, including discrimination, also have greater contagion rates.¹³ Although prior studies have documented associations between exposure and suicidality,^9,11 fewer have examined whether exposure predicts suicide attempt onset independent of depressive symptoms or whether exposure is associated with changes in mental health over time. In addition, much of the suicide contagion literature predates contemporary digital social environments or does not account for online social engagement,^4,14 despite the increasingly central role of digital networks in adolescents’ and young adults’ social lives. Furthermore, adolescence and young adulthood are developmental periods characterized by heightened sensitivity to peer influence,¹⁵ making social network exposures particularly relevant for understanding suicide risk.

This longitudinal cohort study examined whether exposure to suicide attempts in respondents’ social networks was prospectively associated with suicide attempt onset, suicidal ideation onset, and depressive symptoms across 2 waves of follow-up. 

Result of study: Exposure to suicide attempts within one’s social network was associated with subsequent suicide attempt onset and higher depressive symptoms over time.

State social media laws don’t always align with evidence-based strategies to support youth mental health

From TikTok bans and age verification requirements to media literacy lessons in schools, U.S. states are exploring a wide range of strategies to protect young people online. 

The flurry of legislation reflects growing concern about social media's role in youth mental health, as rates of depression and anxiety have increased in recent years. But policy approaches have varied considerably, and the scientific evidence backing their effectiveness is mixed. 

To better understand how states are responding, a research team led by social work researcher Melissa Villodas of George Mason University’s College of Public Health analyzed 16 enacted laws across 10 states. The study, published in The Journal of Policy Practice and Research, offers one of the first frameworks for organizing and assessing the growing patchwork of state social media laws.  

Researchers found the policies generally fell into three broad categories: 

1. Limiting access through measures like parental consent requirements, age verification, internet restrictions in school, and, in one case in Montana, an attempted ban on TikTok. This was by far the most common approach, despite relatively limited evidence that many of these measures improve mental health for young people. 
2. Reducing harmful exposure by restricting targeted advertising to minors, limiting potentially addictive platform features, and expanding parental controls. Existing research offers somewhat stronger, though still mixed, support for these approaches. 
3. Building media literacy by teaching students about topics such as misinformation, cyberbullying, online safety, and social media's effects on mental health. This was the least common approach, even though existing evidence suggests it may be a promising strategy, particularly when it addresses both the risks and benefits of social media use. 

Why does this matter? 

"Most research on this topic has focused directly on how social media affects young people's mental health," said Villodas, associate professor in the Department of Social Work and the study's lead author. "We wanted to understand what state governments are actually doing in response, whether those approaches are supported by existing evidence, and where additional research is needed.” She also emphasized the study’s “person-in-environment” perspective, a fundamental approach in social work that considers how broader forces, including public policy, can influence a young person’s experiences and mental health. 

The study uncovered one glaring disconnect: the policy approach used most often, limiting young people's access, had the least supporting evidence of its effectiveness. Meanwhile, media literacy training, included in just one enacted law in Florida, appeared to have some of the strongest evidence behind it. 

"Many states have focused on measures like parental consent and age verification, but the evidence for those strategies is still limited," Villodas said. "Media literacy has received much less attention, even though it may be one of the more promising approaches when it teaches young people about both the risks and the benefits of social media." 

The study captures the policy landscape through November 2024, but Villodas said its findings are designed to help policymakers evaluate future proposals as the evidence continues to evolve. 

In addition to Villodas, the study was coauthored by Michael Wheeler and Natalia Acevedo, both recent graduates of George Mason’s Master of Social Work program; Fadi Hamati of Columbia Cornell Child and Adolescent Psychiatry at New York-Presbyterian; and Ishita Kapur of the University of Tennessee, Knoxville.

---

Journal

Journal of Policy Practice and Research

DOI

10.1007/s42972-026-00155-5