Skip to content
Free updates Login
SSCRailwayUPSC & State PSC

AI Chatbots Show Cognitive Decline in BMJ MoCA Study

18 February 20251 min read
SCIENCE & TECHNOLOGYAI Chatbots ShowCognitive Declinein BMJ MoCA Study18 February 2025safalsetu.com

Why in the news

A study in the BMJ tested top AI chatbots with a human cognition test and found signs of decline in older model versions, casting doubt on using them for medical diagnosis.

Key facts

  • Models tested: ChatGPT, Claude, Gemini (large language models).
  • Tool: the MoCA test, normally used on people, adapted to probe attention, memory and executive function.
  • Older versions scored lower, like human cognitive decline.
  • Scores: Gemini 1.0 = 16 (lowest); ChatGPT 4 = 26; full marks are 30, and nobody got them.
  • Weak spots: visuospatial skills and executive tasks; most scored under the 26 cut-off.

Significance

  • Such deficits challenge the reliability of AI for medical diagnostics and the claim that AI will soon replace doctors.
  • AI can help, but may not replace human expertise in areas such as neurology.

Exam angle

  • Journal: BMJ; test: MoCA; pass mark: 26 out of 30.

Test yourself

1. Which journal published the study testing AI chatbots on the MoCA cognitive test?

The study was published in the BMJ.

2. In the MoCA study, which chatbot version had the lowest score of 16 points?

Gemini 1.0 scored lowest with 16.

3. What is the full score on the MoCA test used in the study?

None of the bots scored the full 30; most were below the 26 cut-off.

This topic is part of Daily Current Affairs 18 February 2025 (10 topics).

All topics of the dayDay quizFebruary 2025

Read more