Journal of Yeungnam Medical Science

Original article

Medical Informatics
Why large language models cannot possess consciousness: an integrated information theory perspective: Dong Ah Shin, Pyung Goo Cho, Gyu Yeul Ji, Sang Hyuk Park, Soo Heon Kim, Yoo Jin Choo, Min Cheol Chang; J Yeungnam Med Sci. 2025;42:79. Published online December 1, 2025; DOI: https://doi.org/10.12701/jyms.2025.42.79

1,822 View
73 Download

Abstract PDF: Background
The question of whether large language models (LLMs) possess consciousness has been increasingly debated. Integrated information theory (IIT) offers a quantitative framework for assessing consciousness through a measure of integrated information.
Methods
This study applied IIT principles to the architecture of transformer-based LLMs, focusing on causal integration, temporal persistence, and system irreducibility. Ablation experiments on Generative Pretrained Transformer 2 (GPT-2) were performed, selectively removing individual attention heads and measuring changes in perplexity as a behavioral proxy for integrated information to empirically approximate the measure of integrated information.
Results
The ablation study of a single attention head produced minimal or negative changes in perplexity in four out of five representative sentences, indicating redundancy or noise. Only one sentence revealed a significant increase in perplexity change (ΔPPL +11.29), reflecting a localized but nonessential contribution. A comparison with biological systems demonstrated that LLMs meet the IIT criterion of differentiation, but fail to meet the criteria of integration, causal closure, and temporal persistence. These findings confirm that LLMs are architecturally decomposable, lack persistent internal states, and do not sustain global causal irreducibility. Philosophical considerations, including Searle’s Chinese Room argument, further support the idea that the linguistic fluency of LLMs arises from syntactic manipulation rather than semantic understanding.
Conclusion
Current LLMs do not satisfy the structural and informational requirements of consciousness under IIT. Although capable of simulating intelligent language, LLMs remain unconscious systems with a negligible amount of integrated information, underscoring the distinction between linguistic competence and conscious experience.

Case reports

Neurology
Recovery of the ascending reticular activating system and consciousness following comprehensive management in a patient with traumatic brain injury: a case report: Sung Ho Jang, Young Hyeon Kwon; J Yeungnam Med Sci. 2022;39(4):332-335. Published online August 12, 2021; DOI: https://doi.org/10.12701/yujm.2021.01172

12,697 View
133 Download

Abstract PDF: We report on changes in the ascending reticular activating system (ARAS) concurrent with the recovery of impaired consciousness following rehabilitation and cranioplasty in a patient with traumatic brain injury (TBI), which were demonstrated on diffusion tensor tractography (DTT). A 34-year-old male patient was diagnosed with a traumatic intracerebral hemorrhage after falling from a height of approximately 7 m and underwent a right frontoparietotemporal decompressive craniectomy and hematoma removal. At 5 months after onset, when starting rehabilitation, the patient showed impaired consciousness, with a Glasgow Coma Scale (GCS) score of 4. Comprehensive rehabilitative therapy was provided until 14 months after onset, and his GCS score improved to 8. Cranioplasty was performed using auto-bone at 14 months after onset. One month after cranioplasty, his GCS score improved to 12. On the 15-month DTT, the deviated lower dorsal ARAS was restored on both sides, and the right side had become thicker. The right lower ventral ARAS was reconstructed, and increased neural connectivity of the upper ARAS was detected in both the prefrontal cortices. Thus, changes in the ARAS were demonstrated in a patient with TBI during recovery of consciousness following rehabilitation and cranioplasty.

Anesthesiology and Pain Medicine
Awareness during general anesthesia despite simultaneous bispectral index and end-tidal anesthetic gas concentration monitoring: Jungwon Lee, Chorong Park, Saeyoung Kim; Yeungnam Univ J Med. 2019;36(1):50-53. Published online December 14, 2018; DOI: https://doi.org/10.12701/yujm.2019.00010

12,414 View
155 Download
5 Crossref

Abstract PDF

Awareness during general anesthesia occurs in approximately 0.1–0.2% of cases; nevertheless, particular attention is required because it can lead to critical complications including insomnia, depression, anxiety, and post-traumatic stress disorder. To prevent these complications, bispectral index (BIS) and end-tidal anesthetic gas (ETAG) concentration monitoring are commonly used to examine patient consciousness during surgery. In the present case, an 80-year-old man was scheduled for total gastrectomy. Anesthesia was maintained using desflurane 4.0–5.0% vol, oxygen, and nitrous oxide. The authors simultaneously monitored BIS, which was maintained between 37 and 43, and ETAG, which was maintained between 0.9 and 1.2 minimum alveolar concentration (MAC). After the operation, however, the authors were surprised to learn that the patient complained of awareness during anesthesia. Although BIS and ETAG concentration monitoring are useful in preventing awareness during anesthesia, they cannot be completely trusted. Even though BIS was maintained at approximately 40 and ETAG at 0.7–1.3 MAC, awareness during anesthesia occurred.

Citations

Citations to this article as recorded by

Effect of Opioid-Sparing Multimodal Anesthesia Within an Enhanced Recovery After Surgery (ERAS) Protocol Pathway on Postoperative Recovery in Gynecologic Oncologic Surgery: A Randomized Clinical Study
Maria Bourazani, Nikolaos Fyrfiris, Petros Galanis , Chrysseida Maglari, Antonios Anagnostopoulos, Dimitrios C Papatheodorou, Martha Kelesi, Sofia Poulopoulou
Cureus.2026;[Epub] CrossRef
Bispectral Index Versus Minimum Alveolar Concentration Guided Anesthesia for Assessment of Intraoperative Awareness in Patients Undegoing Laparascopic Abdominal Surgery
Shreya Garg, Vinod Bala Dhir, Jyoti Gupta, Rupesh Yadav, Deepak Verma
Indonesian Journal of Anesthesiology and Reanimation.2025; 7(1): 12. CrossRef
Depth of anaesthesia monitoring: updated evidence. Comment on Br J Anaesth 2023; 131: 196–9
Paul S. Myles
British Journal of Anaesthesia.2023; 131(5): e145. CrossRef
Impact of bispectral index monitoring on critical incidents rate in high-risk patients: a randomised controlled trial
N. V. Trembach
Kuban Scientific Medical Bulletin.2022; 29(1): 48. CrossRef
A Crossover Comparison of the Sensitivity and the Specificity between BIS and AEP in Predicting Unconsciousness in General Anesthesia
Haitao Yang, Guan Wang, Jinxia Gao, Jie Liu, Liang Zhao
Scientific Programming.2020; 2020: 1. CrossRef

First
Prev
Page of 1
Next
Last

Search