एआईले मान्छेलाई पटमुर्ख बनायो, स्मरणशक्ति खराब पारिदियो || Nepal Times

Brief summary

AI tools that people increasingly rely on for writing, memory, and decision-making can reduce brain activity, impair memory and creativity, and encourage unquestioning acceptance of AI output. Several small studies and related research report short-term drops in neural activation and longer-term signs of weaker neural connectivity after using generative AI as a cognitive shortcut. Researchers warn this “cognitive offloading” or “cognitive surrender” could, if widespread and long-term, increase risk factors associated with cognitive decline.

Concern: using AI as a cognitive shortcut may reduce mental exercise and, over time, raise risks linked to diminished memory, creativity, and possibly dementia-related factors.

Key scientific concepts and phenomena

• Cognitive offloading / cognitive outsourcing Delegating thinking and memory tasks to external tools (search engines, AI), thereby reducing internal cognitive effort.

• Google effect Reduced ability to remember details because the internet functions as an external memory store.

• Cognitive surrender Accepting AI output without scrutiny and ignoring one’s own intuition or critical thinking.

• Measured neural activity changes Lower EEG activation (including lower gamma-band activity) in people who rely on AI for answers; some participants showed reduced neural connectivity months later.

• Creativity and originality loss AI-generated outputs tend to be similar and “soulless,” reducing practice in generating original work.

• Potential long-term risk Reduced cognitive exercise may raise risk factors linked to dementia (analogy to concerns about GPS reliance and spatial-navigation decline).

• Hybrid intelligence (proposed) A model where humans think first and then use AI as a collaborator or tool, rather than a replacement for thinking.

Studies, methods, and findings

• MIT Media Lab (Natalia Kosmina) — EEG essay study

  • Design: 54 students split into three groups writing short essays on open-ended prompts (e.g., loyalty, happiness).
    • Group A: used ChatGPT (generative AI)
    • Group B: used Google search (non-AI)
    • Group C: no technology
  • Measurements: brain waves (EEG) recorded during the task.
  • Findings (preprint/unpublished):
    • AI group showed up to ~5× lower brain activity in areas tied to creativity and information processing.
    • Search group showed strong visual-area activity.
    • No-tech group showed widespread brain activation.
    • Memory effect: AI users later could not recall or cite their own essays.
    • Follow-up after four months: lower neural connectivity in participants who previously used AI.

• University of Pennsylvania (summary)

  • Describes people entering a state of cognitive surrender when using AI — accepting AI answers without scrutiny.

• Multinational clinical study (summary)

  • Clinicians who used an AI tool to detect colon cancer for three months performed worse at tumor detection without the tool afterward.

• Vivian Ming (UC Berkeley–related work)

  • Experiment: students predicted real-world outcomes (e.g., oil prices).
  • Observations:
    • Many participants asked AI directly and copied its answers; measured gamma waves (indicator of cognitive effort) were low when they copied AI.
    • Participants who used AI as a data source and performed their own analysis showed higher gamma activity and more accurate predictions.
    • Those who took AI’s raw answer tended to predict less accurately than those who interrogated/analyzed AI output.

• Correlational research on GPS/Google Maps

  • Heavy GPS use is associated with worsening spatial memory over time; poor spatial navigation can be an early sign of Alzheimer’s disease.

Practical recommendations and mitigation strategies

• Adopt hybrid intelligence: think first, then use AI to augment, test, or refine your ideas rather than replace your thinking.

• Build foundational understanding before using AI: learn topics without AI assistance, then use AI to expand or refine that knowledge.

• Use a “Nemesis Prompt” (Vivian Ming’s suggestion): ask the AI to act as an opponent/critic that explains why your ideas are wrong and how to improve them — forces defensive, critical thinking.

• Apply friction-based production:

  • Configure prompts so AI provides context but not full answers initially.
  • Ask the AI to pose questions to you or require incremental steps that demand user deliberation.

• Configure AI to withhold full answers in early interactions to prompt user engagement and reasoning.

Researchers and sources featured

• Natalia Kosmina — research scientist, MIT Media Lab (54-student EEG essay study)
• Vivian Ming — computational neuroscientist; author of Robot Proof; ran UC Berkeley–associated experiments and recommends mitigation techniques
• Researchers at the University of Pennsylvania — work describing “cognitive surrender”
• Multinational research team — clinical study on AI use in colon cancer detection
• Prior research on the Google effect and GPS/Google Maps and spatial navigation studies

Tools mentioned

• ChatGPT (generative AI)
• Google search / Google Gemini (search and AI tools)
• GPS / Google Maps (navigation tools)