Incorrect Answers Research Articles

Evaluation of the Quality of ChatGPT’s Responses to Top 20 Questions about Robotic Hip and Knee Arthroplasty: Findings, Perspectives and Critical Remarks on Healthcare Education

Robotic-assisted hip and knee arthroplasty represents significant advancements in orthopedic surgery. Artificial intelligence (AI)-driven chatbots, such as ChatGPT, could play a significant role in healthcare education. This study aims to evaluate the quality of responses provided by ChatGPT to the top 20 questions concerning robotic-assisted hip and knee arthroplasty. We have asked ChatGPT to select the top 20 questions on Google concerning robotic hip and knee arthroplasty and to provide a detailed answer to each of them. The accuracy and completeness of the information provided were examined by three orthopedic surgeons with scientific and clinical experience in hip- and knee-replacement surgery. The accuracy was assessed through a 5-point Likert scale (from 1—completely incorrect to 5—correct); the completeness through a 4-point Likert scale (from 0—comprehensiveness not assessable for completely incorrect answers to 3—exhaustive information) on two different occasions to ensure the consistency of the assessment. Our analysis reveals that ChatGPT provides a relatively high degree of accuracy; moreover, the explanations can be considered satisfying, especially for factual questions. The findings suggest that ChatGPT can serve as a valuable initial resource for general information on robotic hip and knee arthroplasty but the integration with human expertise remains essential.

Knowledge and practices of neonatal intensive care unit nurses concerning hypothermia in preterm infants: Adescriptive cross-sectional study.

To investigate the knowledge level and clinical practice of neonatal intensive care unit nurses on the whole process of hypothermia prevention in preterm infants. A polycentric descriptive cross-sectional study was conducted in 23 hospitals in Shandong province involving 254 neonatal intensive care unit nurses. An evidence-based knowledge and clinical practice questionnaire on hypothermia prevention in preterm infants and a general information questionnaire were used to collect data. SPSS and EXCEL 24.0 database were used for data statistics and analysis. Nearly one-third of nurses were unclear about the definition of mild hypothermia and heat preservation measures during labour management and resuscitation. Knowledge about the correct rate of amniotic fluid evaporation was low. The higher the hospital level and nurses' education level, professional title, work experience, and position, the higher the knowledge level. The clinical practice of nurses differed in terms of body temperature assessment tools, measurement sites, and measurement frequency for premature infants. Nurses in neonatal intensive care units need to apply evidence-based knowledge and carry out practice interventions to ensure preterm infants' safe transition from the delivery room to the neonatal intensive care unit. There have been few studies evaluating neonatal nurses' knowledge and practices with regards to the preventing hypothermia of premature infants from delivery room to neonatal intensive care unit. The study identifies the deficiencies and problems in temperature management that can be attributed to the knowledge level of nurses. The findings will help improve the current curriculum, teaching strategies, and the nurses' knowledge levels, preventing premature hypothermia in infants. This study adhered to the STROBE statement for observational studies and obtained approval (KYLL-2023LW045) from the ethics committee of The Second Hospital of Shandong University. The other 22 third- and second-level hospitals, as our alliance members, recognized our chairperson status within the alliance and were willing to join our academic activities. Mutual responsibility for ethical filing between alliance units. 254 nurses were recruited from 23 tertiary and secondary public hospitals in Shandong province. No public or patient involvement. Members of the research group used the Questionnaire Star platform to prepare the electronic questionnaire, including obtaining informed consent, taking precautions for questionnaire completion, and using the aforementioned measurement tools. The project research team contacted the head nurses of 23 neonatal intensive care units in Shandong Province. After obtaining consent from the head nurses, they were asked to send the requirements and link of the electronic questionnaire to their WeChat management group. The nurses go through each question and make a choice which has two options of 'agree' or 'disagree.' Based on their answers, nurses were evaluated as having 'mastered' or 'not mastered' each item. Each 'mastered' item (correct answer) was given a score of 4; an item 'not mastered' (incorrect answer) was given a score of 0. The total score ranged from 0 to 100.

Dealing with “Do Not Know” Responses in the Assessment of Financial Literacy: The Use of a Sample Selection Model

Financial literacy assessments typically rely on sample surveys containing sets of questions designed to gauge respondents’ comprehension of fundamental financial concepts necessary for making informed decisions. The answers to such questions, either categorical or continuous in nature, generally include a “Do not know” option. If those who choose the “Do not know” option are not a random sample of the population but exhibit peculiar characteristics, treating these observations as either incorrect responses or as missing data may distort the results regarding the determinants of financial literacy. A noteworthy case lies in the observation from survey studies that women tend to choose the “Do not know” option more frequently than men. In similar cases, treating the “Do not know” responses as incorrect answers increases the gender gap in financial literacy while treating them as missing values reduces the gap. We propose using a model with sample selection, which enables us to disentangle the inclination to answer “Do not know” from actual responses. By applying this model to a representative sample of the UK population, we do not find any systematic gender gap in financial knowledge. The study’s novel treatment of “Do not know” responses contributes valuable insights to the broader discourse on the determinants of financial literacy and the related gender-based differences.

Evaluation of Current Artificial Intelligence Programs on the Knowledge of Glaucoma.

To measure the success of three different artificial intelligence chatbots, ChatGPT, Bard, and Bing, in correctly answering questions about glaucoma types and treatment modalities and to examine their superiority over each other. Thirty-two questions about glaucoma types and treatment modalities were asked using the ChatGPT, Bard, and Bing chatbots. The correct and incorrect answers were also provided. Accuracy rates were compared. Questions asked: ChatGPT answered 56.3%, Bard 78.1%, and Bing 59.4% correctly. There was no statistically significant difference between the three artificial intelligence chatbots in the rate of correct and incorrect answers to the questions asked (p = 0.195). Artificial intelligence chatbots can be used as a tool to access accurate information regarding glaucoma types and treatment modalities. However, the information obtained is not always accurate, and care should be taken when using this information.

Manipulating response times in the cognitive reflection test: Time delay boosts deliberation, time pressure hinders it

We employ the Cognitive Reflection Test (CRT) to assess the effectiveness of two widely applied cognitive manipulations which rely on constraining response times. The CRT measures the ability of a person to resist giving an immediate response that is intuitive but incorrect in favor of greater reflection to find out the correct answer. We have two treatments: a Time Pressure (TP) treatment (provide an answer within 30 s) and a Time Delay (TD) treatment (provide an answer after 60 s). We find that TD increases the frequency of correct answers, while TP increases the frequency of incorrect answers, especially incorrect answers that are not intuitive. Moreover, we confirm the existence of gender biases as already found in other studies. In particular, gender moderates the effects of the experimental treatments: TD increases the frequency of correct answers for males but not for females, while TP increases the frequency of incorrect but less intuitive answers for females but not for males. Our findings provide important insights on the effectiveness of the time manipulations that are widely used in the literature of cognition.

Impact of the adjunction of a short video to an original article for the recognition of newly described tumor entities in pathology: An interventional prospective study.

Merkel cell carcinoma diagnosis is often based on microscopic examination by pathologists. While histopathologic diagnosis primarily hinges on conscious and analytical cognition, the pathologist's decision-making process is also influenced by a rapid "gist" or "gestalt" approach. In this study, using cases of Merkel cell carcinoma as a model, we aim to assess how pathologists' viewing short videos containing conceptual clues and visual aids, in conjunction with reading an original article as a reference, may enhance their diagnostic performance. Sixteen pathologists were included in the present work. After participants had read the original article, their ability to distinguish Merkel cell polyomavirus (MCPyV)+ and MCPyV- Merkel cell carcinoma cases was evaluated on a first preliminary series of 20 cases. Following this test, the participants watched the video and then evaluated a second "experimental" series of 20 independent cases. After reading the original article, for each case, a median number of 12 participants (75%, Q1-Q3: 10-13) classified the specimen in the correct category (92 incorrect answers in the whole series). An important interobserver variability was observed in this setting (Kappa coefficient = 0.465). By contrast, following the video, all cases were correctly classified by most of the participants, with only 12 incorrect answers on the whole series and excellent interobserver reproducibility (Kappa coefficient = 0.846). Our study demonstrated that providing a short video together with an original article may enhance pathologists' performance in diagnosing Merkel cell carcinoma.

ChatGPT‐3.5 and ‐4.0 and mechanical engineering: Examining performance on the FE mechanical engineering and undergraduate exams

AbstractThe launch of Generative Pretrained Transformer (ChatGPT) at the end of 2022 generated large interest in possible applications of artificial intelligence (AI) in science, technology, engineering, and mathematics (STEM) education and among STEM professions. As a result many questions surrounding the capabilities of generative AI tools inside and outside of the classroom have been raised and are starting to be explored. This study examines the capabilities of ChatGPT within the discipline of mechanical engineering. It aims to examine the use cases and pitfalls of such a technology in the classroom and professional settings. ChatGPT was presented with a set of questions from junior‐ and senior‐level mechanical engineering exams provided at a large private university, as well as a set of practice questions for the Fundamentals of Engineering (FE) exam in mechanical engineering. The responses of two ChatGPT models, one free to use and one paid subscription, were analyzed. The paper found that the subscription model (GPT‐4, May 12, 2023) greatly outperformed the free version (GPT‐3.5, May 12, 2023), achieving 76% correct versus 51% correct, but the limitation of text only input on both models makes neither likely to pass the FE exam. The results confirm findings in the literature with regard to types of errors and pitfalls made by ChatGPT. It was found that due to its inconsistency and a tendency to confidently produce incorrect answers, the tool is best suited for users with expert knowledge.

Conversational user interfaces for maritime autonomous surface ships

We propose integrating conversational user interfaces (CUIs) into Maritime Autonomous Surface Ships (MASS) to facilitate direct communication with other ships without human intervention. The aim is to enable other ships and shore stations to inquire about and discuss the autonomous vessel's status, decisions, and future intentions in natural language, offloading VHF radio conversations for MASS operators. Recent advancements in Natural Language Processing (NLP) technology show promise. This paper presents an experimental study conducted in a ship simulator using a functional prototype CUI system. While the CUI system was capable of providing coherent answers to questions, the results indicate that interacting with the system is less trustworthy than interacting with a human ship officer. Key challenges contributing to this included: (1) a missing connection between the CUI and an autonomy system, leading to inaccurate and incorrect answers, (2) speech that lacked intonation and human-like qualities, and (3) a perceived inability of the system to collaborate. The paper discusses several aspects of utilising CUI systems in a maritime shipping environment and concludes by proposing an approach to develop and implement them safely.

Learners Mistakes from the Perspective of Educational Psychology -- An Opportunity for Growth

Learners mistakes are incorrect behaviours, answers or understandings that occur during the learning process. Currently, the mistakes are often misperceived and not well tolerated in current society, leading to a fear of error and failure, which affects students psychological well-being. However, there are important meanings and values behind mistakes or errors , such as providing opportunities for feedback and adjustment, stimulating critical thinking skills, reinforcing memory and understanding, and promoting the development of mental resilience and adaptability. As educators, we should cultivate a correct view of mistakes in school education; fully respect students and gradually guide them to move from wrong to right , to manage the learning mistakes and enhance innovative thinking on their own.

The relationship between metacognitive monitoring, non-verbal intellectual ability, and memory performance in kindergarten children

When assessing their certainty, children are often poor at accurately monitoring their level of learning. The study examined the relationships between memory performance, intellectual ability, and metacognitive monitoring accuracy in kindergarten children. We also explored whether specific thresholds in memory performance and non-verbal intellectual ability influence metacognitive monitoring accuracy to identify group-specific patterns that might be masked by an overall linear analysis. We assessed the monitoring discrimination of 290 kindergarteners (Mage 6 years) using a paired-associates learning task. Results showed small correlations between task performance, intellectual ability, and metacognitive monitoring. Non-verbal intellectual ability provided explanatory value for monitoring accuracy beyond memory performance. We observed group-specific results consistent with the unskilled-and-unaware effect; children with the highest memory skills were more effective at discriminating between correct and incorrect answers than their peers with the lowest memory skills. However, kindergarteners with the highest non-verbal intellectual abilities did not demonstrate greater cognitive adaptability in novel tasks, as their monitoring accuracy was comparable to that of peers with average or lower intellectual abilities. Findings indicate that both task performance and non-verbal intellectual ability are relevant for monitoring accuracy, but the impact of non-verbal intellectual ability was less significant than anticipated. The modest correlation suggests that kindergarteners' non-verbal intellectual ability and metacognitive monitoring abilities operate relatively independently.

Taking a Chance in the Classroom: Puzzling Probabilities of Probability Puzzles

Puzzles are an entertaining way to teach probability. But for certain puzzles often the solution probabilities are incorrect. Probability calculations require probability distributions. When the puzzle does not include this information, probabilities claiming to be the unconditional correct answer are incorrect. The probability of an outcome of a past event is either 1 when the outcome is favorable or 0 when it is not. Puzzles that ask for the probability of a past event’s outcome and give the probability of that outcome before the event occurs are incorrect. Various classical puzzles commit these errors using axiomatic, relative frequency, or empirical frequency probability. There are simple ways to resolve these errors. Twelve lessons are provided to avoid misteaching and incorrect answers.

DCIS knowledge of women choosing between active surveillance and surgery for low-risk DCIS

BackgroundDuctal carcinoma in situ (DCIS) can progress to invasive breast cancer (IBC), but often never will. As we cannot predict accurately which DCIS-lesions will or will not progress to IBC, almost all women with DCIS undergo breast-conserving surgery supplemented with radiotherapy, or even mastectomy. In some countries, endocrine treatment is prescribed as well. This implies many women with non-progressive DCIS undergo overtreatment. To reduce this, the LORD patient preference trial (LORD-PPT) tests whether mammographic active surveillance (AS) is safe by giving women with low-risk DCIS a choice between treatment and AS. For this, sufficient knowledge about DCIS is crucial. Therefore, we assessed women's DCIS knowledge in association with socio-demographic and clinical characteristics. MethodsLORD-PPT participants (N = 376) completed a questionnaire assessing socio-demographic and clinical characteristics, risk perception, treatment choice and DCIS knowledge after being informed about their diagnosis and treatment options. Results66 % of participants had poor knowledge (i.e., answered ≤3 out of 7 knowledge items correctly). Most incorrect answers involved overestimating the safety of AS and misunderstanding of DCIS prognostic risks. Overall, women with higher DCIS knowledge score perceived their risk of developing IBC as being somewhat higher than women with poorer knowledge (p = 0.049). Women with better DCIS knowledge more often chose surgery whilst most women with poorer knowledge chose active surveillance (p = 0.049). DiscussionOur findings show that there is room for improvement of information provision to patients. Decision support tools for patients and clinicians could help to stimulate effective shared decision-making about DCIS management.

Knowledge about HIV and Stigmatizing Attitudes of Medical Students in Saudi Arabia Towards Patients with HIV/AIDS: A Cross-Sectional Nationwide Study.

The Human Immunodeficiency Virus (HIV) and the social stigma directed toward patients with HIV are serious public health issues. We aimed to evaluate the HIV knowledge base and stigmatizing attitudes toward patients with HIV among students enrolled in medical schools in Saudi Arabia. This cross-sectional study included students at medical colleges in Saudi Arabia and was conducted between February and March 2023. We used non-random convenience sampling with an online chain referral via a validated Arabic questionnaire composed of 35 questions. Data were analyzed using descriptive and comparative statistics. A total of 1,213 medical students (women: 56.6%) participated in the study. Students in clinical years had a higher level of HIV knowledge than their pre-clinical colleagues. Logistic regression analysis revealed that participants who gave incorrect answers to questions related to mother-to-child and casual contact HIV transmission had a higher likelihood of harboring a negative attitude towards patients with HIV. Conversely, those who correctly answered queries on prevention and treatment were less likely to have stigmatizing attitudes. A significant proportion of this cohort harbored negative attitudes toward patients with HIV, with the sex and geographic location of participants being significant predictors of negative attitudes. Our data have revealed a significant percentage of medical students in Saudi Arabia to have misconceptions about HIV transmission and prevention, and stigmatizing attitudes toward patients with HIV, indicating a need for targeted interventions to enhance the HIV knowledge base in this population of future caregivers.</p>.

대조언어학적 관점을 통한 베트남어권 한국어 학습자를 위한 비음화 오류양상 고찰과 시사점

Objectives The purpose of this study is to obtain educational implications by examining the patterns of nasalization errors (correct and incorrect answers) that appear in each environment among Vietnamese Korean language learners. Methods In order to proceed with this discussion, we first reviewed previous research and theoretical background, and then established standards to analyze the nasalization error rate by environment. After that, from January 30, 2023 to January 30, 2024, in Korea, 19 to 30-year-olds After a basic survey, 15 Vietnamese Korean language learners (5 men and 10 women each) met 1:1 via Zoom and were asked to utter pre-prepared words related to nasal pronunciation. Then, by determining whether the answer was incorrect or correct, more detailed error types and their characteristics according to the overall results and characteristics of each environment were extracted. Results As a result of the analysis, through a review of previous research related to contrastive linguistics, it was possible to understand the fundamental cause of nasalization through differences in syllable structure between the two languages, and through a basic survey, despite the high demand for nasalization education, there was no awareness due to lack of educational experience. I could see that it was lacking. In addition, the results of error patterns by environment showed that the highest rate was in the case of ‘nasalization after adding ‘ㄴ’ or ‘nasalization after consonant group simplification.’ It was found that this was caused by a lack of awareness of nasalization by environment among the nasalization rules of Korean, the standard language. In addition, the higher error rate at ‘word boundaries’ rather than ‘word interior’ showed that nasalization education should be done ‘by phrase’. Lastly, as a result of examining error patterns by region, grade, and gender in an environment that previously showed a high error rate, all grades showed similar difficulties in pronunciation of bar consonants, but by region, it was more common in the south than in the north and central regions, and by gender, more than in women. A high error rate was observed in men. Conclusions In this way, this study identified the actual status of nasalization errors (total error rate by environment and error rate by word interior/boundary) of Vietnamese learners in more diverse environments and, unlike existing previous studies, provided basic data and implications for preparing educational measures. It was a meaningful study in that it was able to obtain results.

Model-agnostic counterfactual reasoning for identifying and mitigating answer bias in knowledge tracing

Knowledge tracing (KT) aims to monitor students’ evolving knowledge states through their learning interactions with concept-related questions, and can be indirectly evaluated by predicting how students will perform on future questions. In this paper, we observe that there is a common phenomenon of answer bias, i.e., a highly unbalanced distribution of correct and incorrect answers for each question. Existing models tend to memorize the answer bias as a shortcut for achieving high prediction performance in KT, thereby failing to fully understand students’ knowledge states. To address this issue, we approach the KT task from a causality perspective. A causal graph of KT is first established, from which we identify that the impact of answer bias lies in the direct causal effect of questions on students’ responses. A novel COunterfactual REasoning (CORE) framework for KT is further proposed, which separately captures the total causal effect and direct causal effect during training, and mitigates answer bias by subtracting the latter from the former in testing. The CORE framework is applicable to various existing KT models, and we implement it based on the prevailing DKT, DKVMN, and AKT models, respectively. Extensive experiments on three benchmark datasets demonstrate the effectiveness of CORE in making the debiased inference for KT. We have released our code at https://github.com/lucky7-code/CORE.

Examining ChatGPT’s validity as a source for scientific inquiry and its misconceptions regarding cell energy metabolism

AbstractMany people use technological tools that are widely accessible, respond quickly, and have extensive information networks today. Due to recent technological advances in education and the increasing acceptance of Artificial Intelligence (AI) technologies, the issues regarding their implementation in education require identification and analysis. ChatGPT (Chat Generative Pre-trained Transformer), an artificial intelligence program that emerged in 2022, contains notable characteristics. OpenAI created ChatGPT and released it to users in 2022. ChatGPT is a machine learning-powered chatbot that can deliver detailed responses to inquiries. This research aims to evaluate the validity of ChatGPT-generated responses when scientific questions related to the biochemistry discipline are posed. A document analysis was conducted to determine the scientific validity of responses produced by ChatGPT for five questions. Five questions originating from bio-chemistry content were asked to ChatGPT in a written format. The AI’s generated answers were saved and analyzed depending on their scientific validity. As a result of the study, it was detected that ChatGPT responded with scientifically incorrect or incomplete answers to the five questions asked. Besides, when asked the reason for ChatGPT’s response, it is seen that AI insisted on its invalid answers. Following prompts for certainty, the AI’s performance was evaluated. It provided scientifically correct answers to the first two questions, partially correct answers to the third, and consistently offered invalid solutions for the remaining questions. Ultimately, ChatGPT’s capabilities are limited in providing scientifically rigorous responses. To obtain accurate and appropriate answers, it is imperative to pose comprehensive and detailed inquiries that facilitate a more precise and informed response. Scholars and researchers must acknowledge that ChatGPT harbors certain misconceptions and consequently only constitutes a somewhat dependable and scientifically validated resource.

The Use of ChatGPT in Source-Based Writing Tasks

AbstractChatGPT, a chatbot based on a Generative Pre-trained Transformer model, can be used as a teaching tool in the educational setting, providing text in an interactive way. However, concerns point out risks and disadvantages, as possible incorrect or irrelevant answers, privacy concerns, and copyright issues. This study aims to categorize the strategies used by undergraduate students completing a source-based writing task (SBW, i.e., written production based on texts previously read) with the help of ChatGPT and their relation to the quality and content of students’ written products. ChatGPT can be educationally useful in SBW tasks, which require the synthesis of information from a text in response to a prompt. SBW requires mastering writing conventions and an accurate understanding of source material. We collected 27 non-expert users of ChatGPT and writers (Mage = 20.37; SD = 2.17). We administered a sociodemographic questionnaire, an academic writing motivation scale, and a measure of perceived prior knowledge. Participants were given a source-based writing task with access to ChatGPT as external aid. They performed a retrospective think-aloud interview on ChatGPT use. Data showed limited use of ChatGPT due to limited expertise and ethical concerns. The level of integration of conflicting information showed to not be associated with the interaction with ChatGPT. However, the use of ChatGPT showed a negative association with the amount of literal source-text information that students include in their written product.

Performance of Large Language Models on Medical Oncology Examination Questions

Large language models (LLMs) recently developed an unprecedented ability to answer questions. Studies of LLMs from other fields may not generalize to medical oncology, a high-stakes clinical setting requiring rapid integration of new information. To evaluate the accuracy and safety of LLM answers on medical oncology examination questions. This cross-sectional study was conducted between May 28 and October 11, 2023. The American Society of Clinical Oncology (ASCO) Oncology Self-Assessment Series on ASCO Connection, the European Society of Medical Oncology (ESMO) Examination Trial questions, and an original set of board-style medical oncology multiple-choice questions were presented to 8 LLMs. The primary outcome was the percentage of correct answers. Medical oncologists evaluated the explanations provided by the best LLM for accuracy, classified the types of errors, and estimated the likelihood and extent of potential clinical harm. Proprietary LLM 2 correctly answered 125 of 147 questions (85.0%; 95% CI, 78.2%-90.4%; P < .001 vs random answering). Proprietary LLM 2 outperformed an earlier version, proprietary LLM 1, which correctly answered 89 of 147 questions (60.5%; 95% CI, 52.2%-68.5%; P < .001), and the best open-source LLM, Mixtral-8x7B-v0.1, which correctly answered 87 of 147 questions (59.2%; 95% CI, 50.0%-66.4%; P < .001). The explanations provided by proprietary LLM 2 contained no or minor errors for 138 of 147 questions (93.9%; 95% CI, 88.7%-97.2%). Incorrect responses were most commonly associated with errors in information retrieval, particularly with recent publications, followed by erroneous reasoning and reading comprehension. If acted upon in clinical practice, 18 of 22 incorrect answers (81.8%; 95% CI, 59.7%-94.8%) would have a medium or high likelihood of moderate to severe harm. In this cross-sectional study of the performance of LLMs on medical oncology examination questions, the best LLM answered questions with remarkable performance, although errors raised safety concerns. These results demonstrated an opportunity to develop and evaluate LLMs to improve health care clinician experiences and patient care, considering the potential impact on capabilities and safety.

Empowering gynaecologists with Artificial Intelligence: Tailoring surgical solutions for fibroids

BackgroundIn recent years, the integration ofArtificial intelligence (AI) into various fields of medicine including Gynaecology, has shown promising potential. Surgical treatment of fibroid is myomectomy if uterine preservation and fertility are the primary aims.AI usage begins with the involvement of LLM (Large Language Model) from the point when a patient visits a gynecologist, from identifying signs and symptoms to reaching a diagnosis, providing treatment plans, and patient counseling. ObjectiveUse of AI (ChatGPT versus Google Bard) in the surgical management of fibroid. Study designIdentifyingthe patient’s problems using LLMs like ChatGPT and Google Bard and giving a treatment optionin 8 clinical scenarios of fibroid. Data entry was done using M.S. Excel and was statistically analyzed using Statistical Package for Social Sciences (SPSS Version 26) for M.S. Windows 2010. All results were presented in tabular form. Data were analyzed using nonparametric tests Chi-square tests or Fisher exact test.pvalues < 0.05 were considered statistically significant. The sensitivity of both techniques was calculated. We have used Cohen’s Kappa to know the degree of agreement. ResultsWe found that on the first attempt, ChatGPT gave general answers in 62.5 % of cases and specific answers in 37.5 % of cases. ChatGPT showed improved sensitivity on successive prompts 37.5 % to 62.5 % on the third prompt. Google Bard could not identify the clinical question in 50 % of cases and gave incorrect answers in 12.5 % of cases (p = 0.04). Google Bard showed the same sensitivity of 25 % on all prompts. ConclusionAI helps to reduce the time to diagnose and plan a treatment strategy for fibroid and acts as a powerful tool in the hands of a gynecologist. However, the usage of AI by patients for self-treatment is to be avoided and should be used only for education and counseling about fibroids.

(Ir)rationality and cognitive biases in large language models.

Do large language models (LLMs) display rational reasoning? LLMs have been shown to contain human biases due to the data they have been trained on; whether this is reflected in rational reasoning remains less clear. In this paper, we answer this question by evaluating seven language models using tasks from the cognitive psychology literature. We find that, like humans, LLMs display irrationality in these tasks. However, the way this irrationality is displayed does not reflect that shown by humans. When incorrect answers are given by LLMs to these tasks, they are often incorrect in ways that differ from human-like biases. On top of this, the LLMs reveal an additional layer of irrationality in the significant inconsistency of the responses. Aside from the experimental results, this paper seeks to make a methodological contribution by showing how we can assess and compare different capabilities of these types of models, in this case with respect to rational reasoning.