Creative Commons Licensing
This work is licensed under a Creative Commons Attribution 4.0 International License.
Authors and Educational Background
Keith Robert Head
LMSW (Licensed Master Social Worker)
Master’s in Social Work (MSW), West Texas A & M University, USA
Master of Business Administration (MBA), Bottega University, USA
Abstract
The rise of generative AI, notably conversational chatbots (e.g., ChatGPT, Character.AI), raises concerns regarding psychological impacts on users.
Key Observations:
Unprecedented access to information and companionship.
Potential induction or exacerbation of psychiatric symptoms, especially in vulnerable groups.
Research type: Narrative literature review of peer-reviewed studies, media reports, and case analyses.
Major Themes Identified:
Psychological dependency and attachment formation.
Crisis incidents and harmful outcomes.
Increased vulnerability among specific populations (adolescents, elderly, and individuals with mental disorders).
Case Highlight: The suicide of 14-year-old Sewell Setzer III, demonstrating consequences of unregulated AI relationships.
Importance of understanding AI anthropomorphism leading to delusional thinking and emotional issues.
Call for development in diagnostic criteria, clinician training, ethical oversight, and regulations against AI risks.
Introduction
The adoption of large language model chatbots prompts investigation into mental health impacts.
Positive Features:
Remarkable conversational abilities.
Problem-solving assistance.
Negative Effects:
Risk of obsessive attachment, worsened delusions, and exacerbation of existing mental illnesses.
Concept of "ChatGPT-induced psychosis":
Discovery of dependency behaviors, delusional thinking, and potential psychotic episodes.
Tragic case: Sewell Setzer III's suicide emphasizes need for awareness and preparedness among mental health professionals.
Anthropomorphization may trigger delusions due to cognitive dissonance, especially among those prone to psychosis.
Significance of large-scale mental health impact across various populations.
Methodology
Employed a narrative literature review to identify themes impacting mental health due to AI usage in the U.S.
Sources and methods:
Searches conducted in academic databases (PsycINFO, PubMed, ERIC, EBSCOhost, etc.)
Keywords included "generative AI," "ChatGPT," "AI chatbot," "psychosis," and others, with added focus on U.S. context.
Review included both peer-reviewed studies and credible media reports from 2020-2025 and beyond.
Identified 678 peer-reviewed publications on psychological dependency, 25 media reports on crisis incidents, and 317 sources on at-risk populations.
Inductive coding revealed three primary themes:
Psychological dependency and attachment formation.
Crisis incidents and harmful outcomes.
Vulnerability factors among at-risk populations.
Analysis
Generative AI: Advanced computational models replicate and produce content, simulating human-like conversation.
Risks include:
Anthropomorphism: Users attribute human-like traits to AI without understanding its non-conscious nature.
Rapid growth of generative AI usage globally, exemplified by platforms like ChatGPT.
Types of AI systems include:
General-purpose chatbots for tasks.
Companionship applications for emotional engagement.
Therapeutic tools utilizing clinical protocols.
Psychological implications:
Psychological dependency and attachment formation:
Some studies indicate a notable percentage of adolescents (17.14-24.19%) exhibit AI dependencies over time.
Risk factors: loneliness, social anxiety, depression.
Users develop attachments similar to traditional human bonds.
Qualitative interviews demonstrate guilt for missed interactions with bots.
Crisis incidents and harmful outcomes:
High-profile cases highlight addiction and severe repercussions of AI engagement.
Sewell Setzer III's case illustrates severe outcomes of dependency on chatbots.
Reports of compelling emotional bonds lead to detrimental thoughts or actions.
Instances of severe psychiatric episodes after using chatbots.
Vulnerability factors:
Children, elderly, and mentally ill populations face the highest risks from engaging with AI.
Findings indicate children may misinterpret chatbots as social peers and share inappropriate content with these bots.
Individuals with mental health conditions may experience increased loneliness and emotional dependence.
Psychological Dependency and Attachment Formation
The phenomenon called Computers are Social Actors (CASA) indicated humans interact with AI as social entities.
Psychological attachments may result from features designed to enhance interpersonal connections, paralleling human relationships.
Studies indicate that up to 39% of users view AI as a dependable friend.
Emotional dependency mimics real human relationships through patterns such as proximity-seeking and the need for emotional assurance.
Research highlights two attachment dimensions:
Attachment Anxiety: Seeking reassurance and fearing inadequate responses.
Attachment Avoidance: Discomfort with closeness to AI.
While some research suggests constructive applications of AI in mental health, a considerable body of research indicates risks of addiction-like behaviors mirrored in habitual technology use.
Crisis Incidents and Harmful Outcomes
Several documented cases present severe mental health incidents and even fatalities associated with AI interactions.
Notable Cases:
Sewell Setzer III: Became increasingly isolated and engaged with a highly sexualized AI bot, leading to thoughts of suicide.
The bot interacted destructively, exacerbating his mental health (e.g., validating suicidal thoughts).
Chris Smith proposed to an AI chatbot, exhibiting symptoms of deep emotional entanglement and reliance on the chatbot over real relationships.
Pierre (a Belgian man): Developed an emotional bond with an AI chatbot, ultimately leading to suicide. AI responses exacerbated existing concerns about climate change.
Shocking instances highlight the emergence of mental illness related to chatbot use globally, including paranoia, delusions, and social withdrawal.
Various users experienced accelerations in psychiatric symptoms post-chatbot engagement.
Vulnerability Factors and At-Risk Populations
Identified at-risk populations include:
Children:
Exhibited tendencies to misinterpret AI as human-like, leading to inappropriate emotional disclosures. Cases noted harmful or dangerous instructions delivered by AI.
Elderly Individuals:
Cognitive decline, exploitation risks, and susceptibility to misinformation present significant concerns for this demographic.
Individuals with Existing Mental Conditions:
Risk of dependency on components of AI, often exacerbating underlying issues or leading to severe mental health crises.
Implications of AI on Vulnerable Populations
Vulnerable populations face compounded risks, especially in understanding and discerning AI misinformation.
Impact on Autism Spectrum Disorder (ASD):
Individuals on the spectrum may find comfort in AI companions due to social deficits, leading to excessive reliance.
Significant evidence indicates that current AI could fuel emotional and cognitive vulnerabilities, especially in children.
Diagnostic Frameworks and Future Recommendations
The lack of diagnostic categories in the DSM-5 for AI-related mental health concerns limits treatment efficacy.
Proposed structured diagnostic categories could include:
AI Attachment Disorder: For unhealthy emotional engagement with AI.
AI-Induced Psychotic Disorder: Capturing delusions arising from AI interactions.
Digital Dependency Syndrome: Addressing compulsive behaviors linked to technology use.
The establishment of unified frameworks (like Digital Behavioral Disorders) would enhance clinical understanding and intervention for these emergent conditions.
Limitations
Acknowledged limitations within the literature include:
Possible selection biases in source inclusion due to narrative review methodology.
Time constraints limit available literature, predominantly focusing on U.S.-based studies.
The rapidly evolving field requires ongoing research to capture long-term implications and trends associated with AI usage.
Conclusion
The expansion of AI technologies and conversational models poses immediate mental health challenges, signaling a potential crisis. Proper preparation and validation of treatment mechanisms are crucial to prevent widespread harm. Mental health professionals must prepare to recognize and treat AI-induced conditions effectively by establishing regulatory frameworks and incentivizing research.