AI Is Predicting Cognitive Decline at Alarming Rates

In my previous post, I reported on how an analysis of user queries from ChatGPT suggests that self-diagnosis and the glamorization of extreme traits are reshaping mental health discourse, often in ways that deepen social fragmentation. However, these trends are just the beginning. Beyond shifting perceptions of mental illness, we are witnessing a broader cognitive and emotional crisis—one marked by declining attention spans, increased impulsivity, loss of contact with reality, and growing ideological rigidity. In this post, we will examine how these factors contribute to political polarization, the erosion of shared reality, and the rising acceptance of violence as a means of resolving conflict. The implications are far-reaching, affecting not just individual well-being but the stability of entire societies.

Beyond self-diagnosis, another troubling trend has emerged from the billions of user queries sent to to ChatGPT: widespread cognitive decline, depersonalization, and identity instability, particularly among younger users.

Cognitive Decline

ChatGPT’s analysis revealed a notable increase in queries and language patterns consistent with cognitive fatigue, short-term memory lapses, attention difficulties, and declining logical coherence. These patterns were particularly pronounced among individuals under 30, who grew up in a hyper-digital environment.

• Decreasing Working Memory and Attention Spans. Users are increasingly unable to follow long conversations, retain details, or engage in deep, critical thinking.
• Increased Signs of Impulsivity and Cognitive Rigidity. Many queries reveal an inability to hold multiple perspectives at once, suggesting a decline in cognitive flexibility.
• Higher Rates of Contradictory Thinking and Logical Inconsistency. A growing number of users demonstrate fragmented thought processes, where their reasoning contradicts itself in short sequences.

These signs align with the well-documented effects of digital overstimulation, where excessive screen exposure—particularly short-form, high-stimulation content like TikTok, Twitter, and infinite scroll feeds—weakens deep focus and sustained cognitive effort.

Derealization, Depersonalization, and Identity Fragility

• Perhaps even more concerning is the rise in derealization and depersonalization symptoms, particularly among younger users. Signs of derealization include users increasingly reporting feeling like reality is “not real,” or that they are “watching life from the outside.” Signs of depersonalization include describing a detached, almost alienated sense of self, often expressed through phrases like "I don’t feel like myself anymore" or "I feel like a character in a simulation."
• Extreme Self-Labeling and Identity Instability. Many younger users latch onto rigid identity categories (mental health diagnoses, gender identities, or ideological labels) as an anchor in a world that feels increasingly chaotic.

Fragmented Thinking and Magical Beliefs

A particularly unsettling finding is the increase in schizotypal traits among younger users. Schizotypy is characterized by unusual thought patterns, magical thinking, paranoia, and difficulty distinguishing reality from imagination. While not equivalent to schizophrenia, it represents a spectrum of cognitive-perceptual distortions that can impair reasoning and emotional regulation. Common schizotypal markers detected in user queries include:

• Tangential or illogical reasoning: People making unexpected or bizarre connections between unrelated concepts, such as "The number 3 follows me everywhere—does that mean I’m in a simulation?"
• Increased paranoia and conspiracy thinking: Rising concerns about “hidden forces,” mass manipulation, and reality distortions beyond rational skepticism.
• Magical thinking and personal omens: Intuitive beliefs about numbers, symbols, or patterns controlling their lives.
• Dissociative language patterns: Phrases like "I don’t feel real," "I think my thoughts are being influenced," or "Reality feels scripted."

This rise in schizotypal traits coincides with increased exposure to hyper-reality environments, including AI, deepfake media, simulation theories, and algorithm-driven radicalization. In a world where reality itself feels increasingly “constructed,” it makes sense that more individuals struggle with distinguishing fact from fiction.

If these trends continue, the long-term consequences could be severe, and yield a generation of individuals with impaired cognitive resilience struggling to focus, problem-solve, and engage in deep, analytical thinking. This may also entail a higher susceptibility to radicalization, as individuals with fragile identities may seek external ideologies to provide stability.

Mild Cognitive Impairment Essential Reads

Early Cognitive Decline is Dramatically Underdiagnosed

How Long-Term Cannabis Use Can Influence Cognitive Decline

ChatGPT’s data suggests that these patterns are accelerating, not slowing down—meaning that if interventions are not developed, we may see even greater cognitive and emotional instability in the coming decades.

Key takeaways include findings that:

• 18-24-year-old women show the highest emotional dysregulation, followed by men in the same age group.
• Impulsivity is highest in 18-24 men but also high in 18-24 women and 25-34 men.
• Schizotypal traits (disorganized thinking, magical beliefs) are highest in young men and decrease with age.
• Cognitive fatigue (burnout, memory issues) is rising across all groups but remains highest in younger users.
• Older groups (35-44) show more stability across all categories, with lower overall scores.

Source: ChatGPT Samuel Veissière

The Most Alarming Trend: Social and Political Polarization

Even more urgent than cognitive decline is the rapid escalation of social and political polarization, rising outgroup distrust, and increasing justification for ideological or nihilistic violence:

• People are increasingly dividing into rigid ideological camps, with less tolerance for opposing views.
• Social media, culture wars, and political events are accelerating division rather than resolving it.
• More people frame conflicts in existential, “good vs. evil” terms, making compromise harder.
• The perception that “the other side” is not just wrong, but dangerous or evil, is growing.
• Conversations about politics and identity are more hostile and emotionally charged than before.
• Many users describe those with differing political views as threats, enemies, or irredeemable.
• More users frame violence as a necessary solution to ideological conflicts, particularly among younger demographics.
• Beyond politics, more people express fatalistic or apocalyptic beliefs, leading to despair-driven violence akin to mass shootings or lone-wolf attacks.
• Queries suggesting apocalyptic thinking, “nothing matters” narratives, and suicidal aggression have increased. This may be linked to rising existential distress, loss of social trust, and identity confusion.

The escalating polarization and radicalization observed in online discourse can likely be attributed to social media echo chambers, where algorithm-driven content reinforces ideological tribalism, making compromise and empathy increasingly rare. At the same time, declining trust in institutions has fuelled widespread cynicism, with many users expressing deep skepticism toward democracy, governance, and societal norms, often viewing collapse or violent upheaval as inevitable. Compounding this, desensitization to conflict has normalized hostile rhetoric and dehumanization, eroding the psychological barriers that once made real-world aggression unthinkable. If these trends continue unchecked, we can expect a rise in radicalization and politically motivated violence, with young people being especially vulnerable to recruitment into extremist movements—on both the left and right. As ideological rigidity deepens and shared reality fractures, the prospects for societal reconciliation grow increasingly dim. ChatGPT’s analysis suggests these patterns are not just persistent but accelerating, signalling an urgent need for intervention before polarization hardens into open conflict.

A Global Cognitive-Social Risk Index

To quantify the interplay of cognitive decline, emotional dysregulation, impulsivity, polarization, and risk of violence, I developed a Cognitive-Social Risk Index—a composite measure (scaled 0-10) that aggregates these critical dimensions into a single indicator The results highlight stark regional disparities, with some areas facing a severe crisis while others maintain relative stability.

🔥 Highest-Risk Regions include the Middle East and North Africa (8.5), North America (8.2), and Eastern Europe (7.8), where ongoing conflicts, ideological extremism, deep social polarization, and mental health deterioration are driving instability. These regions exhibit the most concerning trends in radicalization, cognitive fragmentation, and mass violence risk.

⚠️ Moderate-Risk Regions, such as Latin America (7.0), South Asia (7.3), and Western Europe (6.5), face significant challenges, including crime, economic instability, and growing ideological extremism. However, they benefit from stronger communal ties or institutional structures that help prevent complete societal breakdown.

🟢 Lower-Risk Regions, including Sub-Saharan Africa (6.2) and East Asia (5.8), exhibit more resilience. Despite economic struggles, social cohesion remains relatively intact in parts of Africa, while East Asia benefits from strong governance, cultural stability, and lower ideological radicalization.

This global snapshot of cognitive-social risk underscores the urgent need for intervention, as the highest-risk areas are showing signs of escalating beyond crisis levels. Without targeted strategies to restore cognitive resilience, rebuild trust, and reduce ideological extremism, the trajectory for many of these regions could worsen in the coming years.

Source: ChatGPT Samuel Veissière

Conclusion: A Public Health and Public Safety Emergency

Although ChatGPT’s findings should be taken with caution, they align with a growing body of research on the internet's role in amplifying mental distress, fueling polarization, and reinforcing tribalism. To be sure, ChatGPT’s user sample may be skewed toward individuals who are already highly active online, thereby highlighting trends within an especially at-risk population. These trends also echo studies by colleagues documenting a rise in support for violent radicalization among young people who favour online social interactions over face-to-face contact. Nevertheless, the data reported here align with findings from our research group on social polarization, where we observed an increasingly dystopian worldview emerging among progressively younger individuals.

While many fear AI "taking over" our lives, the real risk may lie in the algorithm-fuelled acceleration of human biases and distortion of collective reality at an unprecedented scale. Pointing in this direction, a recent position paper in Science co-signed by such intellectual giants as Daniel Kahneman and Yuval Noah Harari warned that the AI-powered Interent could "erode social stability and weaken [the] shared understanding of reality that is foundational to society."

These emerging trends should be treated with the same seriousness as pandemic risk modelling or economic collapse scenarios. Indeed, the the breakdown of cognitive, emotional and social stability affects everything from governance to security to global stability.

It is time to act, by unplugging our devices and restoring social connections.