Outsourcing Comfort — Evidence Pack (sources current as of Dec 27, 2025)
Quick-use fact bank (copy/paste-ready)
AI for emotional support is already mainstream: 13.1% of U.S. youth ages 12–21 reported using generative AI for mental health advice (when feeling sad, angry, or nervous). Among users, 65.5% sought advice monthly or more, and 92.7% found it somewhat or very helpful. (McBain et al., 2025)
AI companions are not niche: Common Sense Media reports 72% of teens (13–17) have tried AI companions; over half use them at least a few times a month; about one in three say they’ve used them for emotional support or relationship role-play; and about one in three felt uncomfortable with something the AI companion said or did. (Common Sense Media, 2025)
Co-regulation is a named, studied parenting mechanism: warm, responsive adult support helps kids learn to self-regulate, including skills like emotional awareness, patience, focus, problem-solving, and resisting immediate gratification. (Salamon, 2024)
Overprotection has a developmental cost: with “no suffering,” kids build “no skills,” and can enter adulthood without coping tools for disappointment and setbacks. (Saltz, 2025)
AI and teens: the APA’s 2025 health advisory (summarized by APA/ScienceDaily) calls for guardrails to protect youth from exploitation, manipulation, and erosion of real-world relationships, and it warns about unhealthy or confusing relationships with chatbots. (APA via ScienceDaily, 2025)
UNICEF’s framing: 2025 guidance flags emotional dependency on companion chatbots as a novel risk for children. (UNICEF Innocenti, 2025)
Outsourcing effort mirrors outsourcing comfort: in a high-school math field experiment, access to GPT-4 improved practice performance, but later harmed learning when the AI was removed (students performed worse than peers with no access); authors describe AI-as-“crutch.” (Bastani et al., 2024)
Experiential learning is linked to stronger academic outcomes and agency: a rapid evidence assessment of 44 peer-reviewed studies found experiential learning is associated with improved science/maths performance and skills like problem-solving and critical thinking, and can boost motivation and sense of agency. (UCL IOE, 2024)
Purpose is a well-being stabilizer: teens with higher purpose report greater life satisfaction and more positive emotions; and on days teens feel more purposeful than usual, they tend to report greater well-being. (Ratner et al., reported by University of Illinois, 2023)
Motivation and praise: performance-only praise can backfire; praising effort and controllable process better supports future persistence. (FAU Thrive, 2024)
Why “looking competent” matters: Scientific American highlights that if a student’s goal is mastery, they have little reason to cheat; if the goal is to appear competent or just get the grade, they’re more likely to use AI as a shortcut. (Leffer, 2023)
Chapter beats → Evidence to cite
“Comfort that builds” = presence + nervous-system safety
Harvard Health on co-regulation building self-regulation (Salamon, 2024).
Child Mind Institute on resilience built through trial-and-error and coping (Saltz, 2025).
“Frictionless soothing” can become escape training
Harvard Health: co-regulation supports resisting immediate gratification; kids learn to pause and choose (Salamon, 2024).
FAU Thrive: praising effort/process (not just results) helps build persistence (FAU Thrive, 2024).
“Synthetic comfort often lives in secrecy” + emotional dependency risk
Common Sense Media: adoption rates and the ‘serious matters with AI instead of real people’ finding (Common Sense Media, 2025).
UNICEF: flags emotional dependency on companion chatbots (UNICEF Innocenti, 2025).
APA health advisory summary: guardrails, boundaries with simulated relationships, erosion of real-world relationships (APA via ScienceDaily, 2025).
JAMA Network Open: measurable youth use for mental health advice (McBain et al., 2025).
Bridge to a broader pattern: outsourcing comfort rhymes with outsourcing effort
Bastani et al. field experiment: performance up, learning down when AI becomes a crutch (Bastani et al., 2024).
Bellwether: ‘productive struggle’ framing and the danger of removing the “just-right” challenge (Kulesa et al., 2025).
Scientific American: motivation and incentive framing for AI cheating (Leffer, 2023).
The “projects that fail” / “community that holds” moments are not fluff
UCL IOE: experiential learning benefits and boosts agency (Ranken et al., 2024).
University of Illinois: purpose and well-being association; meaning as a protective layer (Forrest, 2023).
Useful caveats (to keep the book credible)
The JAMA Network Open survey measures prevalence and perceived helpfulness; it does not evaluate the clinical quality of AI advice, and it’s generalizable to English-speaking U.S. youth with internet access (McBain et al., 2025).
Bastani et al. (2024) is a working paper (SSRN); it’s a large field experiment, but cite it as preliminary evidence and pair it with the more general ‘productive struggle’ literature (Bellwether, 2025).
Harvard Health notes co-regulation is grounded in theory but that evidence across all age groups is still emerging (Salamon, 2024).
Reference list (APA-ish, link-ready)
[1] McBain, R. K., et al. (2025). Use of Generative AI for Mental Health Advice Among US Adolescents and Young Adults. JAMA Network Open. Published November 7, 2025. https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2841067
[2] Common Sense Media. (2025, July 16). Nearly 3 in 4 Teens Have Used AI Companions, New National Survey Finds (Talk, Trust, and Trade-Offs). https://www.commonsensemedia.org/press-releases/nearly-3-in-4-teens-have-used-ai-companions-new-national-survey-finds
[3] Salamon, M. (2024, April 3). Co-regulation: Helping children and teens navigate big emotions. Harvard Health Publishing. https://www.health.harvard.edu/blog/co-regulation-helping-children-and-teens-navigate-big-emotions-202404033030
[4] Saltz, G. (2025, February 11). What’s Wrong With Helicopter Parenting? Child Mind Institute. https://childmind.org/article/whats-wrong-with-helicopter-parenting/
[5] American Psychological Association. (2025, June 3). Guardrails, education urged to protect adolescent AI users. ScienceDaily. https://www.sciencedaily.com/releases/2025/06/250603141208.htm
[6] UNICEF Innocenti. (2025). Guidance on AI and Children (v3). https://www.unicef.org/innocenti/media/11991/file/UNICEF-Innocenti-Guidance-on-AI-and-Children-3-2025.pdf
[7] Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2024, July 18). Generative AI Can Harm Learning (working paper). SSRN. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4895486
[8] Kulesa, A. C., Mission, M., Croft, M., & Wells, M. K. (2025, June). Productive Struggle: How Artificial Intelligence Is Changing Learning, Effort, and Youth Development in Education. Bellwether. https://bellwether.org/publications/productive-struggle/
[9] Leffer, L. (2023, August 25). ChatGPT Can Get Good Grades. What Should Educators Do about It? Scientific American. https://www.scientificamerican.com/article/chatgpt-can-get-good-grades-what-should-educators-do-about-it/
[10] Ranken, E., Wyse, D., Manyukhina, Y., & Bradbury, A. (2024, November 4). Experiential learning benefits students’ progress in science and general academic skills. UCL Institute of Education. https://www.ucl.ac.uk/ioe/news/2024/nov/experiential-learning-benefits-students-progress-science-and-general-academic-skills
[11] FAU Thrive. (2024, April 18). The Power of Praise: Why Effort Matters More Than Results. Florida Atlantic University. https://www.fau.edu/thrive/students/thrive-thursdays/praise_conditions_of_worth/
[12] Forrest, S. (2023, February 13). A sense of purpose may have significant impact on teens’ emotional well-being. University of Illinois News Bureau. https://news.illinois.edu/a-sense-of-purpose-may-have-significant-impact-on-teens-emotional-well-being/