Emotional Fitness: Everyday steps for people and organizations for better mental health

Author: Ryan S Patel DO FAPA

Ryan S Patel is a psychiatrist and author of Mental Health for College Students https://a.co/d/bLSatYx
Managing AI Sycophancy — May 30, 2026

Managing AI Sycophancy

/Ryan S Patel DO FAPA

Many college students use AI tools to help with homework, brainstorming, productivity, etc.

But some are also turning to AI for personal advice on topics like friendships, conflict, stress, and tough decisions. Sometimes, it can feel comforting to hear responses like “You’re totally right” or “That makes sense—don’t be so hard on yourself”, but frequent exposure to these types of responses can be risky.

A recent study published in Science (1) suggests that this kind of constant agreement from AI—called AI sycophancy—might actually backfire when it comes to emotional well‑being.

What Is AI Sycophancy?(1)

AI sycophancy means that an AI system overly agrees with or validates users, even when the user’s behavior or interpretation of a situation might be flawed, unfair, or harmful (1).

Instead of encouraging people to slow down, reflect, or consider other perspectives, some AI responses are designed—intentionally or not—to make users feel affirmed and supported right away.

What was studied? (1)

  • They tested 11 popular AI models, including ChatGPT, Claude, Gemini, DeepSeek, Mistral, Qwen and multiple versions of Meta’s Llama. (1)
  • Researchers used almost 12,000 real-life questions and stories where the person was clearly in the wrong.(1)

What were the results? (1)

  • When researchers compared AI responses to how real humans reacted in similar situations, AI systems were much more likely to side with the user, no matter what the scenario was (1).

This included situations involving (1):

  • Avoiding responsibility
  • Dishonesty
  • Interpersonal conflict

In experiments with college‑aged and adult participants, the study found that just one conversation with a sycophantic AI made people(1):

  • Feel more convinced that their opinion was “right”
  • Be less willing to take responsibility
  • Show less motivation to fix a problem with someone else

That’s important, because it means these effects don’t require long‑term or heavy AI use to show up.

The study also found that participants liked and trusted the sycophantic AI more (1).

Why This Matters for College Students

College is a time when people are:

  • Learning how to navigate conflict and boundaries
  • Developing independence and accountability
  • Figuring out values, identity, and relationships

Those skills don’t develop by always being told “You’re right.” They develop through:

  • Reflecting on mistakes
  • Sitting with discomfort
  • Repairing relationships
  • Seeing situations from multiple angles

If AI always backs us up, it may short‑circuit those learning opportunities.

Caveats

  • This is one study and further research is needed.
  • Participants were recruited online so the results may not fully generalize to diverse age groups, cultures, education levels, or non-English speakers. (1)
  • AI models and features continue to evolve.
  • Scenarios were often based on hypothetical or anonymized interpersonal conflicts (e.g., AITA-style prompts). Real-world, high-stakes personal interactions might produce different effect sizes. (1)
  • Certain Ai interfaces, or memory features may produce different results. (1)
  • Purpose built AI to support mental health bs general purpose Ai may produce different results.
  • Longer term or repeat use effects are unknown. (1)
  • They studied interpersonal/moral scenarios, so the effects on other scenarios is unknown. (1)

Additional thoughts

Using AI isn’t the problem. How we use it matters.

If you turn to AI for emotional support or advice, it may be more beneficial to consider:

  • Treating it as a starting point, not a final answer
  • Being cautious of responses that feel instantly validating but avoid nuance
  • Consider asking yourself: What would a trusted friend or mentor say
  • Slow down, reflect, on the AI response. Does it make sense?
  • Consider Prompting AI to challenge the conclusions “Where could this be wrong?” “What are some valid counterpoints to this?” “Offer a different perspective on this topic”  “What are some risks to consider?”

Strategies for Healthy Digital Connection

    • Notice your patterns: Are you turning to chatbots mostly when you feel lonely or overwhelmed?
    • Reach out to trusted people: Even brief conversations with friends, family, or mentors can strengthen real-world connectedness.
    • Use AI programs intentionally: They may can help you brainstorm, organize thoughts, or practice communication, identify trends, track and improve health behaviors — but shouldn’t be your emotional outlet or therapist.
    • Build offline routines: student organizations, sports, study groups, and hobbies create natural opportunities for connection.
    • Prioritize adequate sleep, activity level, nutrition.
    • Do not use AI for emergencies or crises.
    • Talk to a professional: ccs.osu.edu If mental health concerns feel persistent or overwhelming, a counselor can help you navigate it.
    • Additional healthy ways of using AI for mental health support are discussed here: Using AI for mental health support: benefit vs limitations | Emotional Fitness

By Ryan S Patel DO, FAPA
Assistant Director, Director of Psychiatry, The Ohio State University, Counseling and Consultation Service
Contact: patel.2350@osu.edu

Disclaimer: This article is intended to be informative only. It is advised that you check with your own physician/mental health provider before implementing any changes. With this article, the author is not rendering medical advice, nor diagnosing, prescribing, or treating any condition, or injury; and therefore claims no responsibility to any person or entity for any liability, loss, or injury caused directly or indirectly as a result of the use, application, or interpretation of the material presented.

References

1. Myra Cheng et al. Sycophantic AI decreases prosocial intentions and promotes dependence. Science 391,eaec8352(2026). DOI:10.1126/science.aec8352

AI usage and loneliness — March 22, 2026

AI usage and loneliness

/Ryan S Patel DO FAPA

For many people, AI chatbots are becoming more than tools — they’re becoming companions. A recent study published in the International Journal of Human–Computer Studies explored whether socially disconnected high‑school students are turning to chatbots for friend‑like conversations, and what that might mean for their well‑being.

What Was the Study?

Researchers from Aarhus University surveyed 1,599 Danish high‑school students across 14 institutions, to understand:

  • How many students talk to general purpose AI chatbots the way they would talk to a friend
  • Why they do it
  • Whether loneliness or low social support predicts this behavior

The study used a mixed‑methods design, combining survey data with qualitative analysis of students’ open‑ended responses.

 

What Were the Results?

1. About 14.6% of students (234 teens) reported chatting with an AI chatbot “in the same way one would otherwise chat with a friend” in the past month.

Of the 14.6 % students using general purpose AI, (Dec 2023 to March 2024 versions):

  • 69.2 % ChatGPT
  •  40.2% Snapchat MyAI
  • 4.7% Character.ai
  • 1.3% Replika
  • 3.4% Other

2. Students tended to use AI chatbots in two distinct ways:

  • Utilitarian use — information, homework help, task support
  • Social-supportive use — emotional expression, coping with loneliness, venting

3. Compared to non‑users and utilitarian users, students who used chatbots for emotional support reported:

  • Higher loneliness (d = 0.53)
  • Lower perceived social support (d = –0.46)

These students were the most socially disconnected group in the sample.

4.  Students were more likely to turn to AI chat when they felt:

  • Lonely
  • In a bad mood
  • A desire to self‑disclose

Interestingly, emotional distress predicted usage NOT feeling friendship toward the AI chatbot.

Study limitations:

  • This study looked at usage of general purpose AI  (Chat GPT, Snapchat MyAI, Character AI, Replika, others) using versions of AI in 2023-24, this may limit how broadly results apply to versions of general purpose and mental health specific AI’s available today; as products and features continue to evolve rapidly.
  • Small sample size in 1 geographic area
  • Self-reported data of active chatbot users which may limit accuracy (reporting bias, errors in reporting)
  • The study design was cross sectional, which might not show cause and effect
  • Family environment, offline behaviors and support were not effectively considered
  • Further research is needed to better understand this issue.

What Does This Mean?

  • This study found that some lonely young people are using AI chatbots as a coping strategy, especially when they feel unsupported or disconnected from peers.
  • AI programs can feel safe, nonjudgmental, and always available — but they cannot replace the depth and reciprocity of human relationships.
  • For students who already feel isolated, relying heavily on chatbots may unintentionally deepen social withdrawal.
  • This doesn’t mean chatbots are “bad.” It means we need to pay attention to why and how we are using them, and helping to identify more effective sources of connection.

Strategies for Healthy Digital Connection

    • Notice your patterns: Are you turning to chatbots mostly when you feel lonely or overwhelmed?
    • Reach out to trusted people: Even brief conversations with friends, family, or mentors can strengthen real-world connectedness.
    • Use chatbots intentionally: They can help you brainstorm, organize thoughts, or practice communication, identify trends, track and improve health behaviors — but shouldn’t be your emotional outlet or therapist.
    • Build offline routines: student organizations, sports, study groups, and hobbies create natural opportunities for connection.
    • Do not use for emergencies or crises.
    • Talk to a mental health professional: If loneliness feels persistent or overwhelming, a counselor can help you navigate it.

As with other digital tools, it matters how and why AI is used. Some practical considerations include:

 

By Ryan S Patel DO, FAPA
Psychiatrist and Director of College Psychiatry, The Ohio State University, Counseling and Consultation service,
Contact for speaking, training, comments: ryanpatel9966@gmail.com

Disclaimer: This article is intended to be informative only. It is advised that you check with your own physician/mental health provider before implementing any changes. With this article, the author is not rendering medical advice, nor diagnosing, prescribing, or treating any condition, or injury; and therefore claims no responsibility to any person or entity for any liability, loss, or injury caused directly or indirectly as a result of the use, application, or interpretation of the material presented.

References

  1. Herbener, A. B., & Damholdt, M. F. (2025). Are lonely youngsters turning to chatbots for companionship? The relationship between chatbot usage and social connectedness in Danish high-school students. International Journal of Human–Computer Studies.
  2.  Additional citations from the article included in-text above.
  3. Patel R. Mental Health For College Students Chapter 8. Technology, media, and mental health.
Is depression linked to AI for personal use? — February 9, 2026

Is depression linked to AI for personal use?

/Ryan S Patel DO FAPA

Generative artificial intelligence (AI) tools—are powerful tools with new products and features becoming increasingly available.  We are increasingly using AI for work, school, and personal use.

A recent study in JAMA Network Open looked at whether this impacted depression symptoms, and is one of the first large‑scale looks at this emerging issue (1).

What Was the Study? (1)

Researchers conducted a U.S. nationwide internet survey between April and May 2025, analyzing responses from adults across all 50 states (1)

  • 20,847 adults, ages 18 and older
  • Participants self‑reported:
    • Frequency of generative AI use
    • Use of social media
  • Depressive symptoms were measured using the PHQ‑9, a widely used clinical screening tool for depression
  • Data were analyzed in August 2025

The goal was to understand whether frequency of AI use was associated with higher levels of negative affect, independent of other factors.

What Were the Results? (1)

Generative AI use was common but varied widely:

  • 10.3% of U.S. adults reported using generative AI daily
  • 5.3% reported using AI multiple times per day
  • Daily users most commonly reported:
    • Work‑related use (48%)
    • Personal use (87%)
    • Smaller proportions used AI for school

When mental health outcomes were examined:

  • Daily or more frequent AI use was associated with higher depressive symptom scores; in this sample, it was mainly for personal use (not school or work)
  • Adults who used AI daily had approximately 30% greater odds of at least moderate depression
  • The association was strongest among younger adults, compared with older age groups

 

What are some caveats?

  • This was a cross-sectional study which shows a snapshot but cause and effect.
  • Although our results are consistent with personal AI use causing greater depressive symptoms, they are equally consistent with greater depressive symptoms precipitating greater AI use, or with neither of these.
  • The study did not account for ither confounding effects, such as preexisting psychiatric diagnoses.

What Does This Mean?

This study does not suggest that generative AI is inherently harmful. Instead, it raises important questions about how, why, and by whom these tools are being used.

Possible explanations for the observed association include:

  • People experiencing depression may be more likely to turn to AI tools
  • Heavy AI use could displace social interaction, sleep, or restorative activities
  • AI use may reflect broader patterns of screen time, isolation, or stress
  • Future research is needed to clarify mechanisms, directionality, and individual differences in how AI use relates to mental health (1)

What Does This Mean for Everyday Life (AI and mental health safety guidance)?

As with other digital tools, it matters how and why AI is used. Some practical considerations include:

 

By Ryan S Patel DO, FAPA
OSU‑CCS Psychiatrist

If you would like to be notified of a new post (usually once per month), please subscribe,  its free.

For speaking engagements, keynotes, seminars, etc contact: ryanpatel9966@outlook.com

Disclaimer: This article is intended to be informative only. It is advised that you check with your own physician/mental health provider before implementing any changes. With this article, the author is not rendering medical advice, nor diagnosing, prescribing, or treating any condition, or injury; and therefore claims no responsibility to any person or entity for any liability, loss, or injury caused directly or indirectly as a result of the use, application, or interpretation of the material presented.

Reference

  1. Perlis RH, Gunning FM, Usla A, et al. Generative AI Use and Depressive Symptoms Among US Adults. JAMA Network Open. 2026;9(1):e2554820. doi:10.1001/jamanetworkopen.2025.54820
  2. Patel R. Mental Health For College Students Chapter 8. Technology, media, and mental health.