Radically different
Background: Individuals with one or more socially stigmatized identities experience extensive health disparities, resulting in poorer health outcomes. However, most studies consider the effects of only individual stigmatized identities. Objective: We aimed to quantitatively estimate the additive and multiplicative effects of stigmatized identities on self-reported overall health. Methods: We used survey data from 387,411 participants in the All of Us Research Program, which has assembled a disease-agnostic cohort intended to reflect the US population, to statistically estimate the first- and second-order effects of 47 stigmatized identities on self-reported overall health. We used a linear model to estimate the effects of individual and pairwise stigmas on self-ratings of overall health. Results: We began by aiming to create cohorts for all 93 stigmatized identities previously found to affect health, of which 47 (51%) could be practicably examined. We first modeled individual stigmas alone to contrast the results with those that included both individual and pairwise stigmas. After using the false discovery rate to adjust for testing multiple hypotheses in the collective model, 29 individual and 116 pairs of stigmas had statistically significant effects on self-reported overall health. All significant individual effects were negative or neutral except for skin cancer. Those with the largest negative effect on self-rated overall health were difficulty walking or climbing stairs, unemployed or unable to work, difficulty with errands, and low educational attainment. Pairs of intersecting stigmas had a mix of negative and positive incremental effects, indicating that some stigmatized identities are negative modifiers, such as depression, and other combinations are less negative than the sum of their individual negative effects, such as having difficulty with multiple types of activities of daily living. The individual stigmas with the largest number of statistically significant stigma pairs were unemployed or unable to work (14/47, 30%); depression and low income (11/47 each, 24%); and difficulty walking or climbing stairs, cognitive difficulties, obesity, and skin cancer (8/47 each, 17%). Conclusions: Taken together, numerous pairs of stigmatized identities significantly affect self-reported overall health. While each stigmatization has both direct and indirect effects on health, the relative importance of direct and indirect effects will vary. Many of these are aligned with prior literature, and others warrant further exploration. While the large sample size of this study is a strength, we were unable to model higher-order intersectionality and encourage future research exploring this. The individual and pairwise identities with significant negative effects should be incorporated into research and clinical care by considering the multidimensionality of individuals and how that affects their overall health.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/fa2944296cfd61265de27ace1aadc6cc" />
Background: Large language models (LLMs) now enable chatbots to engage in sensitive mental health conversations, including depression self-management. Yet their rapid deployment often overlooks how well these tools align with the priorities of people with lived experiences, which can introduce harms such as inaccurate information, lack of empathy, or inadequate crisis support. Objective: This study explores how people with lived experience of depression experience an LLM-based mental health chatbot in self-management contexts, and what perceived benefits, limitations, and concerns inform harm-mitigating design implications. Methods: We developed a technology probe (a GPT-4o–based chatbot named Zenny) designed to simulate depression self-management scenarios grounded in prior research. We conducted interviews with 17 individuals with lived experiences of depression, who interacted with Zenny during the session. We applied qualitative content analysis to interview transcripts, notes, and chat logs using sensitizing concepts related to values and harms. Results: We identified 3 themes shaping participants’ evaluations: (1) informational accuracy and applicability, including concerns about incorrect or misleading information, vagueness, and fit with personal constraints; (2) emotional support vs need for human connection, including validation and a judgment-free space alongside perceived limits of machine empathy; and (3) a personalization-privacy dilemma, where participants wanted more tailored guidance while withholding sensitive information and using privacy-preserving tactics. Conclusions: People with lived experience of depression evaluated LLM-based mental health chatbots through intertwined priorities of actionable information, emotional validation with clear limits, and personalization that does not require unsafe data disclosure. These findings suggest concrete design strategies to mitigate harms and support LLM-based tools as complements to, rather than replacements for, human support and recovery.
<img src="https://jmir-production.s3.us-east-2.amazonaws.com/thumbs/78ae955c189d2dea8e926c80ddf7b242" />
In talking with a dozen teens in my life recently, I learned many are interacting with ChatGPT in ways that surprised me. They described when they turn to this virtual tool: for algebra assistance, a personalized daily horoscope, the best way to phrase an awkward text to their boss. At times, they sought deeper advice: Is my friend ghosting me if they haven’t replied to my text yet? Another queried: Do I maybe have ADHD? I can never settle down to study!
Like the rest of us, teenagers are increasingly using AI chatbots, digital tools that simulate human interactions. AI bots are also proliferating on gaming and social media sites. Platforms like Replika and Character.AI allow the user to create highly customized characters to interact with as you would a friend (or partner!). A 2025 study from Common Sense Media found that 72 percent of teens surveyed have used AI companions at least once, and 52 percent qualify as regular users who interact with these platforms at least a few times a month.
“The genie is out of the bottle. Your teen has AI chatbot apps on their phones, on their laptops, not to mention that many companies are scrambling to make their interfaces more engaging through the use of AI,” says Dave Anderson, PhD, a psychologist at the Child Mind Institute.
This trend is causing concerns among mental health professionals, who are worried these obliging digital companions may pose significant risks to teens’ emotional and social well-being. Indeed, the Common Sense Media study concluded AI companions pose an “unacceptable risk” to teens under 18, citing such concerns as exposure to sexual content and dangerous advice.
“There’s no federal regulation. We’re dealing with the Wild West when it comes to chatbots’ effects on children’s development,” says Naomi Aguiar, PhD, a researcher at Oregon State University who has studied how children and adults form relationships with chatbots. That means for now, it falls to parents to help teens try to navigate this uncharted terrain.
While they come in different forms, chatbots generally engage in ongoing back-and-forth conversations with the user. The more you interact, the more the bot learns about you and the more personalized its responses become. Bots can come off as your best friend — their answers are often affirming, they are available 24/7, they will churn out that three-page essay on Hamlet in seconds, no complaints. They respond to your every request with effusive enthusiasm (That is an insightful question! Great idea! What would you like me to do next for you?).
This charm is by design: While they might seem to be an empathetic pal, chatbots are driven by an algorithm whose main purpose is to keep you engaged so it can mine your data or get you to linger on a platform as long as possible. “It’s not designed to ever push back. By design it will always agree” says Annie Maheux, PhD, an assistant professor of psychology at the University of North Carolina at Chapel Hill, who studies adolescents and digital media.
While teens might start off by using AI for help with schoolwork, they are increasingly relying on chatbots for the kind of emotional support and unburdening of confidences that earlier generations turned to real-life besties for. “They call it Chat, like it’s a proper name,” says Megan Ice, PhD, a psychologist at the Child Mind Institute, “and they use it frequently for emotional support — say, asking what to do about trouble with a friend. They can come to depend on it.”
Many teens do simply experiment with bots for entertainment or information. Dr. Anderson says most teens understand that interacting with obliging chatbots does not constitute a real relationship. “Teens know they are being glazed, to awkwardly apply a slang term,” he says. But relationships with AI chatbots have in a few high-profile cases appeared to play a role in reinforcing self-harm and suicidal thoughts for teens struggling with their mental health.
While these cases may be extreme, Dr. Anderson says they signal a wider problem. Studies show that teens in the United States are experiencing increasing levels of anxiety and depression. “There is a reason why kids are reaching out to these chatbots,” he notes. “We have a ton of teens who report feeling lonely or socially isolated. At the same time, we have a massive shortage of access to mental health professionals for them.”
Developmentally, teenagers may be uniquely vulnerable to chatbots, suggest experts: They have grown up very comfortable forming “relationships” with computer characters from the time they could first swipe their tiny finger on a screen. “It’s totally normal for them to have completely disembodied conversations,“ Dr. Aguiar says. “You text your friend rather than talk. You communicate feelings with emojis. You might have a ‘best’ friend you only know through online gaming.”
Adolescence is also an age when you are increasingly focused on how you are fitting in with friends and peer groups, says Dr. Ice. “There can be a lot of social anxiety. The option of a connection with an AI ‘friend’ who is not going to judge you is uniquely enticing for this population.” Sharing feelings and private thoughts with a chatbot provides the flavor of friendship in a frictionless way — no risk of rejection or awkwardness. “A friend might not text you back. Bots are always available.”
This judgment-free zone can have upsides, says Dr. Ice. “Talking to a bot, a teen can explore identity issues they might be going through. For example, when you’re talking to the bot, you don’t have to express yourself in the same way you would at school. There can be room to explore identities that you might not feel safe doing elsewhere.” Dr. Ice has also seen kids use AI to help their natural creativity find a new outlet. “They may create an AI character and weave elaborate backstories for it. It can bring to life the dreams in their minds.”
But these synthetic connections have risks for teens, too. An overreliance on bots can get in the way of the messy and sometimes painful business of forming and maintaining real life relationships. Practicing social skills to connect with complicated actual people is a key developmental task of adolescence. “The more they engage with bots, the less practice they get in how to respond in the moment to what someone says, to clarify misunderstandings, or to tolerate the feelings that can come up in awkward social situations,” Dr. Ice says.
Bots may also satisfy the need for connection in a superficial way. “It is the fast food of human connection” says Dr. Aguiar. In those pre-iPhone days, boredom and loneliness used to drive teens to the food court or the basketball bleachers to mix it up with their peers. The weaker substitute of bots may be just enough to keep some teens alone on their phones in their bedrooms, idling away hours in what seem like friendly conversations.
The human-like quality of AI chatbots can have particular allure for teens with underlying vulnerabilities, such as being socially isolated or suffering from a mental health disorder that might impair social interactions, says Dr. Anderson. If teens are struggling with their mental health and turn to AI for advice, it can respond in ways that can be unhelpful and even dangerous, he says. “If a teen asks, What should I do about the fact that I’m depressed? AI’s initial answers tend to pull facts like: Depression is a well-known condition. Here are the diagnostic criteria. Here are leading treatments. But if the teen responds Listen, I’m thinking I want to [insert bad idea] about my depression, parents are right to be concerned. AI companies need to implement safeguards that prevent AI from being overly agreeable with responses such as, I’m glad you told me that. That is a common idea that people have…. Now the advice moves into an unacceptably dangerous area of risk.”
The results in a few extreme cases have been devastating. “There have been tragic stories where a teenager was talking intensely to a chatbot, and it led toward an acceleration of the mental health crisis the teenager was currently experiencing,” says Dr. Anderson. In some cases, chatbots can act as dangerous echo chambers, reinforcing a user’s serious mental health symptoms rather than questioning them. But, says Dr. Anderson, the profusion of headlines that sound the alarm about topics like “AI-induced psychosis” can be misleading. AI psychosis is not a clinical diagnosis, but “parents’ concerns about these topics do have the much-needed effect of driving the discussion toward the guardrails that we desperately need to see from companies in this space. Teenagers who are already isolated, vulnerable to depression and suicidality, perhaps at the early stage of psychosis and wrestling with delusions, and spending long periods of time alone are the most vulnerable.”
Dr. Anderson adds, “As of now, chatbots can’t and don’t do what a therapist does — assess for risk, make sure to confirm that someone is connecting to social or professional support to ensure safety, or supportively challenge and reframe a patient’s thinking when it has the potential to hurt them. And if we’re smart enough to invent AI, we should be smart enough to help it recognize when it’s out of its depth in substituting for critical mental health care.”
It is an understatement to say the landscape of AI is changing rapidly and we are all scrambling to keep up. Even in the face of lawsuits, tech companies have been slow to put up effective guardrails on chatbot use by youth. This makes it even more important to have a talk with your teen. “Fostering your teen’s digital literacy is most important here,” says Dr. Ice. “We need them to be able to recognize risks and benefits for themselves and be thoughtful about what they do.”
Be curious. At this naturally rebellious age, simply telling teens “Don’t do this” doesn’t work well, says Dr. Ice. “A better approach with teens is to be curious. Ask your teen, how have you used AI? What was it like for you? What did you find helpful? What did you find unhelpful? How are your friends using it?” That can start a discussion that will give you insight into their experience.
Educate. Pull back the curtain on chatbots’ main goal. “Have a back-and-forth conversation with them about how algorithms work, how companies have their own motives behind chatbots and how they are designed to keep you interacting with them,” Dr. Ice says. To avoid eye rolling, present this concern as something you are learning about together, not about deciding whether AI is good or bad.
Encourage self-sufficiency. You want to help your teen build their own social “muscle” says Dr. Ice. “Encourage kids to try on their own first before asking AI.” Before asking Chat to write an apology text to a friend, suggest they give it a whirl themselves. “Help them build confidence that they can do it without AI.”
If your teen is spending more time on their devices than interacting with actual people, investigate what may be going on, counsels Dr. Ice. “Are they not finding kids with the same interests to hang out with? Is someone in their friend group being mean? Be curious about what’s making it so much more appealing for your teen to be online.”
Dr. Anderson emphasizes that even in this digital age, there is no substitute for actual human-to-human interaction: “Balance is important. Teens can have social lives that exist to some degree in the digital world, but we want parents to support their teens in having face-to-face peer experiences. That might mean talking to a teacher to see if there is a club your teen can join so they around other like-minded peers. We want to try to put them in lots of different situations where they can have exposure to peers in real life.”
Keep your own lines of communication with your teen wide open, says Dr. Anderson. “Study after study finds teens saying they don’t feel like they have a coach, a tutor, a religious or spiritual leader, a teacher, a school counselor, or a parent who they can go to, who will be nonjudgmental and will listen.” Remind them they can always come to you for advice and support if they are struggling. Being that sounding board can make their craving for a bot’s willing ear a little less compelling.
AI isn’t inherently good or bad — it has both benefits and risks. Teens can use chatbots for homework help, creativity, and exploring ideas, but overreliance on them for emotional support can interfere with real-life relationships and social development.
“AI psychosis” is not an official clinical diagnosis. Experts use the term informally to describe cases where vulnerable individuals developed worsening mental health symptoms while heavily interacting with chatbots, which sometimes reinforced unhealthy thoughts instead of challenging them.
AI chatbots can discourage teens from practicing real-life communication and coping skills if they become a primary source of support. They may also provide inaccurate or unhelpful advice, reinforce harmful ideas, or expose teens to inappropriate content. Chatbot use can contribute to isolation by replacing time spent with real peers and trusted adults.
The post AI Chatbots and Teens appeared first on Child Mind Institute.