Tuesday, November 4, 2025

Is your PG's AI usage that of a Centaur or a Cyborg: Journal club "NEJM review"

Article in NEJM that shows how PG residents can be classified into Centaurs and Cyborgs in terms of their AI usage!




CC licence: https://en.wikipedia.org/wiki/Liminal_being#/media/File%3APalais_Bourbon%2C_Malerei_in_der_Kuppel_der_Poesie%2C_Szene-_Erziehung_des_Achill_(Eug%C3%A8ne_Delacroix).jpg



Unfortunately the article is closed access and I hope they don't mind my sharing their figure 4




https://www.nejm.org/doi/full/10.1056/NEJMra2503232?trk=feed-detail_main-feed-card_feed-article-content


Here's another article that "offers an in-depth analysis of a groundbreaking study, "Navigating the Jagged Technological Frontier," conducted by leading academic and corporate institutions. They explore key findings, reveal real-world implications, and ponder ethical considerations, ultimately aiming to shed light on AI's role in the future of work including 

Methodology: How Was the Study Conducted?

The research engaged 758 consultants from Boston Consulting Group and employed a comprehensive stratified random assignment method. Participants were assessed on various tasks ranging from creativity to analytical thinking, with or without AI support and the article linked below unpacks these fascinating results."

Unquoted from: https://www.linkedin.com/pulse/decoding-jagged-frontier-ai-centaurs-cyborgs-future-work-dan-martines#:~:text=Unlike%20the%20Centaur%20model%2C%20where,complex%20challenges%20in%20real%2Dtime.

The name Centaur for humans who are supposedly superior in intelligence may have stemmed from the fact that centaurs are deemed to be liminal beings "combining two distinct states of simultaneous existence within one physical body. This unique perspective may provide the liminal being with wisdom and the ability to instruct, making them suitable mentors.

https://en.wikipedia.org/wiki/Liminal_being



The term cyborg applies to a living organism that has restored function or enhanced abilities due to the integration of some artificial component or technology that relies on feedback. https://en.wikipedia.org/wiki/Cyborg

CC licence: https://en.wikipedia.org/wiki/Cyborg#/media/File%3ASerengeti_Cyborg%2C_by_Fanuel_Leul.jpg


Conversational learning self citations:


[08/12, 07:00]hu1 sharing from a public WhatsApp channel named AI ideas with 1400 members: *Recent Studies & Findings on AI-Driven Search Behaviour*


Recent research shows that generative-AI tools are transforming how people look up information, shifting from traditional link-based searching toward conversational, dialogue-style retrieval. Studies across 2024–2025 consistently show that users value *speed* and *convenience*, but this ease introduces challenges around *trust, accuracy, and verification*.


A 2025 study [1] found that users often trust chatbot answers because they feel natural and easy, though this can create misplaced confidence when the AI is wrong. Another study focused on interdisciplinary learning [2] shows that chatbots help once learners understand their field, but can hinder early exploration when background knowledge is weak.


Trust dynamics are changing too. Research comparing Google and ChatGPT for health information [3] shows users often trust ChatGPT more, even where accuracy is uncertain. A theoretical model of AI search behaviour [4] suggests adoption is driven by perceived ease and reduced effort.


Other work identifies major behavioural shifts. A 2024 analysis [5] documents 45 ways users adapt old search habits into conversational interactions. Experimental comparisons of ChatGPT-style tools and Google [6] indicate that AI chat enables faster task completion but may weaken verification.


Large-scale survey data confirms rapid adoption. Pew Research Center reports that 34% of U.S. adults have used ChatGPT, up from ~17% in 2023 [7]. Market-intelligence reports show a sharp rise in “search-like prompts,” suggesting growing dependence on AI for everyday queries [8].


However, risks remain. Scholars warn that generative search may reduce transparency, embed hallucinations, and erode critical evaluation skills [9]. Performance is also uneven: AI is helpful in simple or familiar domains but unreliable in technical or high-stakes fields.


*In short*: AI-driven search is becoming the new default, but reliability, verification, and digital literacy now matter more than ever.


*References:*  

[1] https://arxiv.org/abs/2504.05156  

[2] https://arxiv.org/abs/2507.21490  

[3] https://arxiv.org/abs/2403.09987  

[4] https://journals.sagepub.com/doi/10.1177/21582440241300007  

[5] https://www.sciencedirect.com/science/article/pii/S0740818824000525  

[6] https://www.researchgate.net/publication/372075166  

[7] https://www.pewresearch.org/short-reads/2025/06/25/34-of-us-adults-have-used-chatgpt-about-double-the-share-in-2023/  

[8] https://www.bain.com/insights/how-customers-are-using-ai-search/  

[9] https://jevinwest.org/papers/Memon2024searchenginescipblog.pdf

[08/12, 08:06]hu2: 👏

[08/12, 08:08]hu3: Important conclusions

[08/12, 08:10]hu3: How a student can learn in an ecosystem of ready made Generative AI Agents or LLMs like ChatGPT or Gemini or Grok or Meta or Claude  is an open question given the fact that they are readily available and will be used by the students

[08/12, 08:28]hu2: An interesting exploration , around a journal club, on how post graduate medical residents learn with AI was identified and modeled after two mythical creatures 👇


https://medicinedepartment.blogspot.com/2025/11/is-your-pgs-ai-usage-that-of-centaur-or.html?m=1


Essentially centaurs are dominant human learners with AI in the loop while cyborgs are humans who let AI take the upper hand and are content to be a part of the AI loop

No comments:

Post a Comment