assignment on test cross

References and Recommended Reading

Mendel, G. Versuche über Plflanzen-hybriden. Verhandlungen des naturforschenden Ver-eines in Brünn, Bd. IV für das Jahr 1865, Abhand-lungen, 3-47 (1866) (Bateson translation)

Pierce, B. Genetics: A Conceptual Approach , 2nd ed. (New York, W. H. Freeman, 2006)

Sadava, D., et al . Life: The Science of Biology , 8th ed. (New York, W. H. Freeman/Sinaeur Associates, 2008)

• Add Content to Group

Article History

Flag inappropriate.

Email your Friend

•  |  Lead Editor:  Laura Hoopes

Within this Subject (47)

• Consequences of Gene Regulation (9)
• From DNA to Protein (6)
• Gene Responses to Environment (9)
• Organization of Chromatin (5)
• Regulation of Transcription (7)
• RNA (4)
• Transcription Factors (7)

Other Topic Rooms

• Gene Inheritance and Transmission
• Gene Expression and Regulation
• Nucleic Acid Structure and Function
• Chromosomes and Cytogenetics
• Evolutionary Genetics
• Population and Quantitative Genetics
• Genes and Disease
• Genetics and Society
• Cell Origins and Metabolism
• Proteins and Gene Expression
• Subcellular Compartments
• Cell Communication
• Cell Cycle and Cell Division

© 2014 Nature Education

• Press Room |
• Terms of Use |
• Privacy Notice |

Visual Browse

Complimentary 1-hour tutoring consultation Schedule Now

Testcross (Backcross; concepts of parental, F1, and F2 generations)

A test cross is a way to determine whether an organism that expressed a dominant trait was homozygous or heterozygous ; backcross is the mating between parent and offspring to preserve the parental genotype ; P represents parent, F1 (filial 1) represents the children of the parent and F2 represents the children of the F1.

In genetics, dominant alleles are assigned capital letters (e.g., AA), and recessive alleles are assigned lowercase letters (aa). If both copies of the allele are the same, that individual is said to be homozygous. If they are different, the individual is heterozygous (Aa, aA). The parent or P generation refers to the individuals being crossed. The offspring are the filial or F generation; F₁  or the first filial represents the children of the parents while the F₂ represents children of the F1 or grandchildren of the parents.

Homozygous and Heterozygous Genotypes . Two dominant alleles (BB) or two recessive alleles (bb) are homozygous. One dominant allele and one recessive allele (Bb or bB) are heterozygous.

A  test cross  can be performed to determine whether an organism expressing a dominant trait is homozygous or heterozygous. Meaning, this is a way to tell what hidden genes are carried, but not shown. In a test cross, if you want to find out if the dominant trait is homozygous or heterozygous, it must be crossed with the homozygous recessive (bb). If the offspring is Bb, half the offspring will express the recessive trait. If it is BB, all offspring will express the dominant trait, and none will express the recessive trait. For instance, if the parent has blue eyes, and one of her children has brown eyes, then the parent is a heterozygous dominant. 

Because a test cross is used to determine the genotype of the parent based on the phenotypes of its offspring, test crosses are sometimes called backcrosses.  The backcross  is mating between the offspring and the parent to preserve the parental genotype.

Practice Questions

MCAT Official Prep (AAMC)

Biology Question Pack, Vol 2. Question 82

Sample Test B/B Section Passage 7 Question 38

Section Bank B/B Section Question 30

Practice Exam 4 B/B Section Question 58

• A test cross is a way to determine whether an organism that expressed a dominant trait was a heterozygote or a homozygote.

• In a test cross, the dominant trait must be crossed with the homozygous recessive to know whether it is homozygous or heterozygous it. If the offspring is recessive, half the offspring will express the recessive trait. If it is dominant, all offspring will express the dominant trait, and none will express the recessive trait.

• The parent or P generation refers to the individuals being crossed; the offspring are the filial or F generation.

• F₁ or the first filial represents the children of the parents; F₂ represents children of the F₁ or grandchildren of the parents.

• The backcross  is mating between the offspring and the parent to preserve the parental genotype.

Testcross : a cross between an organism showing a dominant trait and an organism showing a recessive trait to determine whether the former organism is homozygous or heterozygous for that trait

Filial : offspring, daughter or son

Phenotype : the set of observable characteristics of an individual resulting from the interaction of its genotype with the environment

Genotype : the part of the genetic makeup of a cell or an individual which determines one of its characteristics (phenotype)

Dominant : a relationship between alleles of a gene, in which one allele masks the expression (phenotype) of another allele at the same locus

Homozygous : of an organism in which both copies of a given gene have the same allele

Heterozygous : of an organism which has two different alleles of a given gene

Recessive : an allele that can be covered up by a dominant trait, represented by a lower case letter

Your Notifications Live Here

{{ notification.creator.name }} Spark

{{ announcement.creator.name }}

Trial Session Enrollment

Live trial session waiting list.

Next Trial Session:

{{ nextFTS.remaining.months }} {{ nextFTS.remaining.months > 1 ? 'months' : 'month' }} {{ nextFTS.remaining.days }} {{ nextFTS.remaining.days > 1 ? 'days' : 'day' }} {{ nextFTS.remaining.days === 0 ? 'Starts Today' : 'remaining' }} Starts Today

Recorded Trial Session

This is a recorded trial for students who missed the last live session.

Waiting List Details:

Due to high demand and limited spots there is a waiting list. You will be notified when your spot in the Trial Session is available.

• Learn Basic Strategy for CARS
• Full Jack Westin Experience
• Interactive Online Classroom
• Emphasis on Timing

Next Trial:

Free Trial Session Enrollment

Daily mcat cars practice.

New MCAT CARS passage every morning.

You are subscribed.

{{ nextFTS.remaining.months }} {{ nextFTS.remaining.months > 1 ? 'months' : 'month' }} {{ nextFTS.remaining.days }} {{ nextFTS.remaining.days > 1 ? 'days' : 'day' }} remaining Starts Today

Welcome Back!

Please sign in to continue..

Please sign up to continue.

{{ detailingplan.name }}.

• {{ feature }}

Want to create or adapt books like this? Learn more about how Pressbooks supports open publishing practices.

2.5 The Dihybrid Test Cross

While the cross of an F 1 x F 1 gives a ratio of 9:3:3:1, there is a better, easier cross to test for independent assortment: the dihybrid test cross. In a dihybrid test cross, independent assortment is seen as a ratio of 1:1:1:1, which is easier to score than the 9:3:3:1 ratio. This test cross will also be easier to use when testing for linkage.

Like in monohybrid crosses ( Chapter 1 ), you can do test crosses with dihybrids to determine the genotype of an individual with dominant phenotypes, to see if they are heterozygous or homozygous dominant. This type of cross is set up in the same fashion; an individual with an unknown genotype in two loci is crossed to an individual that is homozygous recessive for both loci.

Take a look at the video, Two-Gene Test Cross Explained , by Nicole Lantz (2020) on YouTube, for some worked examples.

Punnett squares should be done ahead of the crosses, so you know what to expect for any of the possible outcomes. Using the example from the rest of this chapter, you cross a double homozygous recessive pea plant (r/r ; y/y. green and wrinkled) to an unknown individual that has two dominant phenotypes (R/_ ; Y/_. yellow and round). There are four possible genotypes the unknown individual could be: R/R ; Y/Y or R/R ; Y/y or R/r ; Y/Y or R/r; Y/y. The Punnett squares for the first two are shown in Figure 2.5.1 . Notice on the left, you only get the dominant phenotype for both, so you know both genes in the unknown are homozygous dominant. On the right, you get only the dominant phenotype for round peas — but you get 50% yellow and 50% green peas, showing that the unknown is homozygous for round, but heterozygous for colour of the peas. Figure 2.5.2  is blank for you to fill in the other two gamete and genotype possibilities.

Media Attributions

• Figure 2.5.1 , Original by L.Canham (2017), CC BY-NC 3.0
• Figure 2.5.2 , Original by L.Canham (2017), CC BY-NC 3.0

Canham, L. (2017). Figures: 7. Punnett square for a test cross; 8. Blank Punnett squares to fill in the other two possibilities of the test cross [digital images]. In Locke, J., Harrington, M., Canham, L. and Min Ku Kang (Eds.),  Open Genetics Lectures, Fall 2017 (Chapter 17, p. 6-7). Dataverse/ BCcampus. http://solr.bccampus.ca:8001/bcc/file/7a7b00f9-fb56-4c49-81a9-cfa3ad80e6d8/1/OpenGeneticsLectures_Fall2017.pdf

Nicole Lantz. (2020, May 5). Two-gene test cross explained [Video file]. YouTube. https://youtu.be/GM0by2axiLM

Long Descriptions

• Figure 2.5.1 Two Punnett squares: The first is a testcross between a dihybrid homozygous dominant organism (RRYY) and the tester, which is a dihybrid homozygous recessive organism (rryy). The offspring and their genotypes in this test cross are shown, with all possessing the heterozygous condition for both traits (RrYy). The second is a testcross between a dihybrid heterozygous organism (RrYy) and the tester, which is a dihybrid homozygous recessive organism (rryy). The offspring and their genotypes in this test cross are shown, with half possessing the heterozygous condition for both traits (RrYy) and the other half expressing the genotype Rryy. [Back to Figure 2.5.1 ]
• Figure 2.5.2 A blank Punnett square can be used for practice, and to fill in the other two gamete and genotype possibilities for the question in the text directly preceding this figure. [Back to Figure 2.5.2 ]

Introduction to Genetics Copyright © 2023 by Natasha Ramroop Singh, Thompson Rivers University is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License , except where otherwise noted.

Share This Book

Biology (Year 12) - Genetics

Ben Whitten

A test cross is a technique used by geneticists in which an individual whose genotype is unknown for a dominant phenotype (it could be homozygous dominant or heterozygous dominant ) is crossed with an individual that is homozygous recessive at the locus in question. The ratio of phenotypes in the offspring reveals the unknown genotype. Drosophila melanogaster is the scientific name for the fruit fly and is an excellent example of where test crosses are used as they have a large number of variants. The long-winged condition is dominant (V) to the vestigial, recessive (non-useful) wing (v). If a pure-breeding long-winged fly (VV) is mated with a vestigial-winged fly (vv), then the F1 generation (first filial) will all be heterozygous (Vv) and have long wings. However, how can it be decided whether a given F2 long-winged fly will be homozygous dominant (VV) or heterozygous (Vv)? This is determined by crossing the unknown genotype with a vestigial-winged fly. As a vestigial-winged fly must be homozygous recessive (vv), if the long-winged fly whose genotype is unknown is VV, crossing with the homozygous recessive fly will produce all long-winged flies. However, if the unknown fly is heterozygous (Vv), there will be a mix of long and vestigial-winged flies in equal numbers (approximately).

One bed, two blankets: We put the Scandinavian sleep method to the test

If you sleep with a cover hog, this might be for you.

There are plenty of Scandinavian exports worth embracing, including saunas , Lego , Dansk tableware , the Billy bookcase , Kransekake , Marimekko . So when the Scandinavian sleep method started popping up everywhere on social media , I took note.

The idea: Two people share a bed but use separate covers. The promise: a better night’s sleep. The bedding arrangement is popular in other European countries as well; on a trip to Paris a couple of years ago, our hotel room bed was outfitted with two sets of blankets and top sheets.

But does the method work? And how do you make a bed with two blankets? We put it to the test. We gave five couples two twin-size down alternative comforters each and asked them to try the method for a week. Here’s what they had to say.

Typical sleeping arrangement: Queen-size bed with a flat sheet, blanket and comforter.

Pain points: Tester 1’s husband accuses her of being a chronic sheet stealer, but she thinks the main issue is that they operate at different temperatures: “I am a nuclear furnace when I sleep, while my husband is a block of ice,” she says. Regardless, it seems like someone is always getting shortchanged on covers, or someone is tossing and turning.

How they carried out the test: They used the separate comforters, no flat sheet or blankets.

The experience: She likes to throw her foot or leg out from under the covers when she gets too hot, and the separate covers allowed her to do this on both sides, instead of just one. Another bonus: Lounging in bed on weekends was better with the extra covers. “It was absolutely luxuriant to take over the whole bed and both comforters and do the crossword puzzle,” she says. Her husband liked that he could roll over and move around in bed without affecting his wife.

The bed-making problem: They don’t typically make their bed, so this wasn’t an issue. “No shade on anyone who is put together enough in the morning to arrange their bedding, but we usually leave it in a chaotic state, and tend to reassemble the tangle of sheets into something more coherent as we’re going to bed, not when we wake up,” she says.

The verdict: The arrangement worked well for them, and they may implement it in the future. “The only hesitation is that we have so much bedding, buying more almost becomes a storage problem at this point,” she says. “But I think we slept better and more comfortably, so if I see twin comforters on sale anytime soon, there’s a good chance we will make the switch.”

Typical sleeping arrangement: Queen-size bed with a top sheet and a duvet.

Pain points: Tester 2 and her husband often go to bed at different times, and many nights one of them has to get up to tend to their toddler, which can be disruptive. There are also issues with sharing covers, she says: “One of us (a.k.a. me) is allegedly a cover hog, which means the other of us sleeps in a defensive crouch with respect to the shared covers as a way to maintain some blanket.”

How they carried out the test: They started with the two twin-size down alternative duvets provided, but her husband didn’t like the texture, so he swapped his out for a twin-size quilt they had. And she realized she preferred a bigger blanket, so she swapped in a queen-size quilt. They kept their queen-size flat sheet.

The experience: The test went well for this couple. Even the difference of opinions on the blankets ended up working in their favor. “The idea that we could each seek out a blanket with our preferred texture and warmth level was an improvement on sharing one quilt,” she says. “It was great to each have our own quilt and made for more restful sleeping.”

The bed-making problem: They don’t make the bed daily but like to do it when time allows. Settling on one twin quilt for him and a queen-size quilt for her helped here. During the day, they made the bed with the queen-size cover and folded the twin cover across the bottom of the bed.

The verdict: They said the arrangement was a big improvement on their previous blanket-sharing situation and plan to continue to use it.

Typical sleeping arrangement: Queen-size bed with a top sheet and comforter.

Pain points: Tester 3 says her husband is an occasional cover-stealer, and she’s a light sleeper who wakes easily if he is tossing and turning.

How they carried out the test: They used the twin comforters and skipped the flat sheet.

The experience: Her husband tossed and turned a good bit one night, and got up to use the bathroom several times on another night, but it didn’t disturb her. She says, though, that she’s so used to his restlessness that she might not have been disturbed even while sharing covers. She liked being able to stick her leg out from under the covers on either side. But there were also drawbacks, including getting overheated. “I think during the night, the excess parts of our blankets would overlap on one person or the other,” she says.

The bed-making problem: They folded the twin comforters lengthwise and placed them side by side to give the bed a neat appearance.

The verdict: They plan to keep their original bedding configuration. “I don’t think we saw enough of a positive impact to switch to two blankets,” she says. She also likes having a neatly made bed, which was another strike against the Scandinavian method.

Typical sleeping arrangement: King-size bed with a flat sheet, blanket and duvet.

Pain points: Tester 4 likes to have covers tucked snugly around her, while her husband struggles to keep his side tucked. Sometimes that leads to uneven distribution of blankets. She maintains that because his covers are untucked, they bunch up at the bottom, creating the impression that she’s stolen them when, in fact, she has not. He admits this possibility: “I think I kick covers off and then overcompensate when trying to recover (ha),” he says. “Then the groggy tug-of-war begins.”

How they carried out the test: They used the two twin duvets provided and dragged out old twin-size top sheets from when their kids were little, giving them completely separate covers.

The experience: The test took cover-stealing out of the equation, and she reports that she was less aware of when her husband got in and out of bed (he goes to bed later — and gets up earlier — than she does). “Something has to go pretty wrong for you to end up with the other person’s covers,” her husband says. “I don’t really see a downside to it. It’s the happy medium between the freedom of separate beds and the traditional way that just doesn’t work for many people.” It didn’t solve his issues with keeping his covers in order; sometimes he woke up with just the duvet and couldn’t find the top sheet. She liked the fact that he couldn’t blame her for this problem anymore. Overall, it improved their quality of sleep.

The bed-making problem: They typically don’t make their bed, so this wasn’t an issue.

The verdict: They have returned to their usual sleeping arrangements for now, but may switch at some point in the future if they find bedding they like.

Typical sleeping arrangement: A queen bed with a flat sheet, blanket and comforter.

Pain points: Tester 5 says she and her partner’s covers come untucked at times, but overall, they don’t have issues with sharing. She sleeps “warm,” so prefers a cool room with a fan pointing at her, and often wakes up under just the sheet.

How they carried out the test: They started with just the separate comforters but missed having a top sheet, so after a week, they brought back a shared queen-size sheet.

The experience: While they slept well during the test, they didn’t notice much of a difference between sharing a comforter or having separate ones. “We already have a comforter we like enough, and both of us prefer to have a flat sheet between the comforter and ourselves,” she says. After a few nights, they added a sheet and blanket for her partner, who sleeps cooler. And after a week, they brought back the queen-size top sheet. Overall, they didn’t notice much of a difference sleeping with two duvets, possibly because they don’t struggle with sharing covers.

The bed-making problem: It’s easier to make the bed with fewer covers, she says, but it definitely looked less tidy. When her parents came to visit, they tossed their regular comforter over the bed for a neater appearance.

The verdict: They felt pretty neutral about the test and have gone back to their regular sleeping arrangement. “I think this is fine when we travel … but isn’t a huge benefit at home,” she says.

What Is Test Cross?

Test cross: a test cross is a cross between an individual with an unknown genotype with a homozygous recessive genotype. the test cross research was initially used by gregor johann mendel. it is used to know whether the trait which is dominant is heterozygous or homozygous..

Explain what is test cross?

What is test cross and its significance?

• Open access
• Published: 26 June 2024

Comparative accuracy of ChatGPT-4, Microsoft Copilot and Google Gemini in the Italian entrance test for healthcare sciences degrees: a cross-sectional study

• Giacomo Rossettini   ORCID: orcid.org/0000-0002-1623-7681 1 , 2 ,
• Lia Rodeghiero 3 ,
• Federica Corradi 4 ,
• Chad Cook   ORCID: orcid.org/0000-0001-8622-8361 5 , 6 , 7 ,
• Paolo Pillastrini   ORCID: orcid.org/0000-0002-8396-2250 8 , 9 ,
• Andrea Turolla   ORCID: orcid.org/0000-0002-1609-8060 8 , 9 ,
• Greta Castellini   ORCID: orcid.org/0000-0002-3345-8187 10 ,
• Stefania Chiappinotto   ORCID: orcid.org/0000-0003-4829-1831 11 ,
• Silvia Gianola   ORCID: orcid.org/0000-0003-3770-0011 10   na1 &
• Alvisa Palese   ORCID: orcid.org/0000-0002-3508-844X 11   na1  

BMC Medical Education volume  24 , Article number:  694 ( 2024 ) Cite this article

355 Accesses

3 Altmetric

Metrics details

Artificial intelligence (AI) chatbots are emerging educational tools for students in healthcare science. However, assessing their accuracy is essential prior to adoption in educational settings. This study aimed to assess the accuracy of predicting the correct answers from three AI chatbots (ChatGPT-4, Microsoft Copilot and Google Gemini) in the Italian entrance standardized examination test of healthcare science degrees (CINECA test). Secondarily, we assessed the narrative coherence of the AI chatbots’ responses (i.e., text output) based on three qualitative metrics: the logical rationale behind the chosen answer, the presence of information internal to the question, and presence of information external to the question.

An observational cross-sectional design was performed in September of 2023. Accuracy of the three chatbots was evaluated for the CINECA test, where questions were formatted using a multiple-choice structure with a single best answer. The outcome is binary (correct or incorrect). Chi-squared test and a post hoc analysis with Bonferroni correction assessed differences among chatbots performance in accuracy. A p -value of < 0.05 was considered statistically significant. A sensitivity analysis was performed, excluding answers that were not applicable (e.g., images). Narrative coherence was analyzed by absolute and relative frequencies of correct answers and errors.

Overall, of the 820 CINECA multiple-choice questions inputted into all chatbots, 20 questions were not imported in ChatGPT-4 ( n  = 808) and Google Gemini ( n  = 808) due to technical limitations. We found statistically significant differences in the ChatGPT-4 vs Google Gemini and Microsoft Copilot vs Google Gemini comparisons ( p -value < 0.001). The narrative coherence of AI chatbots revealed “Logical reasoning” as the prevalent correct answer ( n  = 622, 81.5%) and “Logical error” as the prevalent incorrect answer ( n  = 40, 88.9%).

Conclusions

Our main findings reveal that: (A) AI chatbots performed well; (B) ChatGPT-4 and Microsoft Copilot performed better than Google Gemini; and (C) their narrative coherence is primarily logical. Although AI chatbots showed promising accuracy in predicting the correct answer in the Italian entrance university standardized examination test, we encourage candidates to cautiously incorporate this new technology to supplement their learning rather than a primary resource.

Trial registration

Not required.

Peer Review reports

Being enrolled in a healthcare science degree in Italy requires a university examination, which is a highly competitive and selective process that demands intensive preparation worldwide [ 1 ]. Conventional preparation methods involve attending classes, studying textbooks, and completing practical exercises [ 2 ]. However, with the emergence of artificial intelligence (AI), digital tools like AI chatbots to assist in exam preparation are becoming more prevalent, presenting novel opportunities for candidates [ 2 ].

AI chatbots such as ChatGPT, Microsoft Bing, and Google Bard are advanced language models that can produce responses similar to humans through a user-friendly interface [ 3 ]. These chatbots are trained using vast amounts of data and deep learning algorithms, which enable them to generate coherent responses and predict text by identifying the relationships between words [ 3 ]. Since their introduction, AI chatbots have gained considerable attention and sparked discussions in medical and health science education and clinical practice [ 4 , 5 , 6 , 7 ]. AI chatbots can provide simulations with digital patients, personalized feedback, and help eliminate language barriers; they also present biases, ethical and legal concerns, and content quality issues [ 8 , 9 ]. As such, the scientific community recommends evaluating the AI chatbot’s accuracy of predicting the correct answer (e.g., passing examination tests) to inform students and academics of their value [ 10 , 11 ].

Several studies have assessed the accuracy of AI chatbots to pass medical education tests and exams. A recent meta-analysis found that ChatGPT-3.5 correctly answered most multiple-choice questions across various medical educational fields [ 12 ]. Further research has shown that newer versions of AI chatbots, such as ChatGPT-4, have surpassed their predecessors in passing Specialty Certificate Examinations in dermatology [ 13 , 14 ], neurology [ 15 ], ophthalmology [ 16 ], rheumatology [ 17 ], general medicine [ 18 , 19 , 20 , 21 ], and nursing [ 22 ]. Others have reported mixed results when comparing the accuracy of multiple AI chatbots (e.g., ChatGPT-4 vs Microsoft Bing, ChatGPT-4 vs Google Bard) in several medical examinations tests [ 23 , 24 , 25 , 26 , 27 , 28 , 29 ]. Recently, two studies observed the superiority of ChatGPT-3.5 over Microsoft Copilot and Google Bard in hematology [ 30 ] and physiology [ 31 ] case solving. Recent work has also observed that ChatGPT-4 outperformed other AI Chatbots in clinical dentistry-related questions [ 32 ], whereas another revealed that ChatGPT-4 and Microsoft Bing outperformed Google Bard and Claude in the Peruvian National Medical Licensing Examination [ 33 ].

These findings suggest a potential hierarchy in accuracy of AI chatbots, although continued study in medical education is certainly warranted [ 3 ]. Further, current studies are limited by predominantly investigating: (A) a single AI chatbot rather than multiple ones; (B) examination tests for students and professionals already in training rather than newcomers to the university; and (C) examination tests for medical specialities rather than for healthcare science (e.g., rehabilitation and nursing). Only two studies [ 34 , 35 ] have attempted to address these limitations, identifying ChatGPT-3.5 as a promising, supplementary tool to pass several standardised admission tests in universities in the UK [ 34 ] and in France [ 35 ]. To our knowledge, no study has been performed on admission tests for admissions to a healthcare science degree program. Healthcare Science is a profession that includes over 40 areas of applied science that support the diagnosis, rehabilitation and treatment of several clinical conditions [ 36 ]. Moreover, the only studies conducted in Italy concerned ChatGPT's accuracy in passing the Italian Residency Admission National Exam for medical graduates [ 37 , 38 ] offering opportunities for further research setting.

Accordingly, to overcome existing knowledge gaps, this study aimed to assess the comparative accuracy of predicting the correct answer of three updated AI chatbots (ChatGPT-4, Microsoft Copilot and Google Gemini) in the Italian entrance university standardized examination test of healthcare science. The secondary aim was to assess the narrative coherence of the text responses offered by the AI chatbots. Narrative coherence was defined as the internally consistency and sensibility of the internal or external explanation provided by the chatbot.

Study design and ethics

We conducted an observational cross-sectional study following the Strengthening of Reporting of Observational Studies in Epidemiology (STROBE) high-quality reporting standards [ 39 ]. Because no human subjects were included, ethical approval was not required [ 40 ].

This study was developed by an Italian multidisciplinary group of healthcare science educators. The group included professors, lecturers, and educators actively involved in university education in different healthcare disciplines (e.g., rehabilitation, physiotherapy, speech therapy, nursing).

In Italy, the university’s process of accessing the healthcare professions is regulated by the laws according to short- and long-term workforce needs [ 41 ]. Consequently, the placements available for each degree are established in advance; to be enrolled in an academic year, candidates should take a standardized examination test occurring on the same day for all universities. This process, in most Italian universities, is annually managed by the CINECA (Consorzio Interuniversitario per il Calcolo Automatico dell'Italia Nord Orientale), a governmental organization composed of 70 Italian universities, 45 national public research centers, the Italian Ministry of University and Research, and the Italian Ministry of Education [ 42 ]. CINECA prepares the standardized test common to all healthcare disciplines (e.g., nursing and midwifery, rehabilitation, diagnostics and technical, and prevention) for entrance to University [ 43 ]. The test assesses basic knowledge useful as a prerequisite for their future education [ 44 ], in line with the expected knowledge possessed by candidates that encompass students at the end of secondary school, including those from high schools, technical, and professional institutes [ 45 ].

For this study, we adopted the official CINECA Tests from the past 13 years (2011–2023) obtained from freely available public repositories [ 46 , 47 ]. The CINECA Test provided 60–80 range of independent questions per year for a total of 820 multiple-choice questions considered for the analysis. Every question presents five multiple-choice options, with only one being the correct answer and the remaining four being incorrect [ 44 ]. According to the law, over the years, the CINECA test consisted of multiple-choice questions covering four areas: (1) logical reasoning and general culture, (2) biology, (3) chemistry, and (4) physics and mathematics. The accuracy of each AI chatbot was evaluated as the sum of the proportion of correct answers provided among all possible responses for each area and for the total test. In Additional file 1, we reported all the standardized examination tests used in the Italian language and an example of the question stem that was exactly replicated.

Variable and measurements

We assessed the accuracy of three AI chatbots in providing accurate responses for the Italian entrance university standardized examination test for healthcare disciplines. We utilized the latest versions of ChatGPT-4 (OpenAI Incorporated, Mission District, San Francisco, United States) [ 48 ], Microsoft Copilot (Microsoft Corporation, WA, US) [ 49 ] and Google Gemini (Alphabet Inc., CA, US) [ 50 ] that were updated in September 2023. We considered the following variables: (A) the accuracy of predicting the correct answer of the three AI chatbots in the CINECA Test and (B) the narrative coherence and errors of the three AI chatbots responses.

The accuracy of three AI chatbots was assessed by comparing their responses to the correct answers from the CINECA Test. AI Chatbots’ answers were entered into an Excel sheet and categorized as correct or incorrect. Ambiguous or multiple responses were marked as incorrect [ 51 ]. Since none of the three chatbots has integrated multimodal input at this point, questions containing imaging data were evaluated based solely on the text portion of the question stem. However, technical limitations can be present, and a sensitivity analysis was performed, excluding answers that were not applicable (e.g., images).

The narrative coherence and errors [ 52 ] of AI chatbot answers for each question were assessed using a standardized system for categorization [ 53 ]. Correct answers were classified as [ 53 ]: (A) “Logical reasoning”, if they clearly demonstrated the logic presented in the response; (B) “Internal information”, if they included information from the question itself; and (C) “External information”, if they referenced information external to the question.

On the other side, incorrect answers were categorized as [ 53 ]: (A) “Logical error”, when they correctly identify the relevant information but fail to convert it into an appropriate answer; (B) “Information error”, if AI chatbots fail to recognize a key piece of information, whether present in the question stem or through external information; and (C) “Statistical error”, for arithmetic mistakes. An example of categorisation is displayed in Additional file 2. Two authors (L.R., F.C.) independently analyzed the narrative coherence, with a third (G.R.) resolving uncertainties. Inter-rater agreement was measured using Cohen’s Kappa, according to the scale offered by Landis and Koch: < 0.00 “poor”, 0–0.20 “slight”; 0.21–0.40 “fair”, 0.41–0.60 “moderate”, 0.61–0.80 “substantial”, 0.81–1.00 “almost perfect” [ 54 ].

We used each multiple-choice question of the CINECA Test, formatted for proper structure and readability. Because prompt engineering significantly affects generative output, we standardized the input formats of the questions following the Prompt-Engineering-Guide [ 55 , 56 ]. First, we manually entered each question in a Word file, left one line of space and then inserted the five answer options one below the other on different lines. If the questions presented text-based answers, they were directly inputted into the 3 AI chatbots. If the questions were presented as images containing tables or mathematical formulae, they were faithfully rewritten for AI chatbot processing [ 57 ]. If the answers had images with graphs or drawings, they were imported only into Microsoft Copilot because ChatGPT-4 and Google Gemini only accept textual input in their current form and could not process and interpret the meaning of complex images, as present in the CINECA Test, at the time of our study [ 58 ].

On 26th of September 2023, the research group copied and pasted each question onto each of the 3 AI chatbots in the same order in which it was presented in the CINECA Test [ 59 ] and without translating it from the original Italian language to English because the AIs are language-enabled [ 60 ]. To avoid learning bias and that the AI chatbots could learn or be influenced by conversations that existed before the start of the study, we: (A) created and used a new account [ 2 , 51 ], (B) always asked each question only once [ 61 , 62 ], (C) did not provide positive or negative feedback on the answer given [ 60 ], and (D) deleted conversations with the AI chatbots before entering each new question into a new chat (with no previous conversations). We presented an example of a question and answer in Additional file 3.

Statistical analyses

Categorical variables are presented as the absolute frequency with percent and continuous variables as mean with confidence interval (CI, 95%) or median with interquartile range (IQR). The answers were collected as binomial outcomes for each AI chatbot respect to the reference (CINECA Tests). A chi-square test was used to ascertain whether the CINECA Test percentage of correct answers differed among the three AI chatbots according to different taxonomic subcategories (logical reasoning and general culture, biology, chemistry, and physics and mathematics). A sensitivity analysis was performed, excluding answers that were not applicable (e.g., if the answers had images with graphs or drawings). A p -value of < 0.05 was considered significant. Since we are comparing three groups/chatbots, Bonferroni adjustment, Familywise adjustment for multiple measures, for multiple comparisons was applied. Regarding narrative coherence and errors, we calculated the overall correct answers as the relative proportion of correct answers provided among the overall test answers of each AI chatbot accuracy. A descriptive analysis of reasons for logical argumentation of correct answers and categorization of type error was reported by percentage in tables. Statistical analyses were performed with STATA/MP 16.1 software.

AI chatbots’ multiple-choice questions

From our original sample, we inputted all the multiple-choice questions in Microsoft Copilot ( n  = 820). Twelve multiple-choice questions were not imported in ChatGPT-4 ( n  = 808) and Google Gemini ( n  = 808) since they were images with graphs or drawings. The flowchart of the study is shown in Fig.  1 .

The study flow chart

AI chatbots’ accuracy

Overall, we found a statistically significant difference in accuracy between the answers of the three chatbots ( p  < 0.001). The results of the Bonferroni adjustment, as a Familywise adjustment for multiple measures and tests between couples, are presented in Table  1 . We found a statistically significant difference in the ChatGPT-4 vs Google Gemini ( p  < 0.001) and Microsoft Copilot vs Google Gemini ( p  < 0.001) comparisons, which indicate a better ChatGPT-4 and Microsoft Copilot accuracy than Google Gemini (Table  1 ). A sensitivity analysis excluding answers that were not applicable (e.g., if the answers had images with graphs or drawings) showed similar results reported in Additional file 4.

AI chatbots’ narrative coherence: correct answers and errors

The Inter-rater agreement regarding AI chatbots’ narrative coherence was “almost perfect” ranging from 0.84–0.88 kappa for internal and logical answers (Additional file 5). The narrative coherence of AI chatbots is reported in Tables 2 and 3 . We excluded from these analyses all not applicable answers (ChatGPT-4: n  = 12, Microsoft Copilot: n  = 0, Google Gemini: n  = 12).

About the category of correct answer (Table  2 ), in ChatGPT-4 (tot = 763), the most frequent feature was “Logical reasoning” ( n  = 622, 81.5%) followed by “Internal information” ( n  = 141, 18.5%). In Microsoft Copilot (tot = 737), the main frequent feature was “Logical reasoning” ( n  = 405, 55%), followed by “External information” ( n  = 195, 26.4%) and “Internal information” ( n  = 137, 18.6%). In Google Gemini (tot = 574), the most frequent feature was “Logical reasoning” ( n  = 567, 98.8%), followed by a few cases of “Internal information” ( n  = 7, 1.2%).

With respect to category of errors (Table  3 ), in ChatGPT-4 (tot = 45), the main frequent reason was “Logical error” ( n  = 40, 88.9%), followed by a few cases of “Information error” ( n  = 4, 8.9%) and statistic ( n  = 1, 2.2%) errors. In Microsoft Copilot (tot = 83), the main frequent reason was “Logical error” ( n  = 66, 79.1%), followed by a few cases of “Information error” ( n  = 9, 11.1%) and “Statistical error” ( n  = 8, 9.8%) errors. In Google Gemini (tot = 234), the main frequent reason was “Logical error” ( n  = 233, 99.6%), followed by a few cases of “Information error” ( n  = 1, 0.4%).

Main findings

The main findings reveal that: (A) AI chatbots reported an overall high accuracy in predicting the correct answer; (B) ChatGPT-4 and Microsoft Copilot performed better than Google Gemini; and (C) considering the narrative coherence of AI chatbots, the most prevalent modality to present correct and incorrect answers were “Logical” (“Logical reasoning” and “Logical error”, respectively).

Comparing our study with existing literature poses a challenge due to the limited number of research that have examined the accuracy of multiple AI chatbots [ 30 , 31 , 32 , 33 ]. Our research shows that AI chatbots can accurately answer questions from the CINECA Test, regardless of the topics (logical reasoning and general culture, biology, chemistry, physics and mathematics). This differs from the fluctuating accuracy found in other studies [ 34 , 35 ]. Our findings support Torres-Zegarra et al.'s observations that the previous version of ChatGPT-4 and Microsoft Bing were superior to Google Bard [ 33 ], while other research groups did not confirm it [ 30 , 31 , 32 ]. This discrepancy may be due to differences in the tests used (e.g., medical specialties vs university entrance), the types of questions targeted at different stakeholders (e.g. professionals vs students), and the version of AI chatbots used (e.g., ChatGPT-3.5 vs 4).

The accuracy ranking of AI chatbots in our study might be due to differences in their neural network architecture. ChatGPT-4 and Microsoft Copilot AI use the GPT (Generative Pre-trained Transformer) architecture, while Google Gemini adopts LaMDA (Language Model for Dialogue Application) and later PaLM 2 (Pathways Language Model) in combination with web search [ 32 ]. The differences in the quality, variety, and quantity of data used for training, the optimization strategies adopted (e.g., fine-tuning), and the techniques applied to create the model could also account for the accuracy differences between AI chatbots [ 63 ]. Therefore, the variations mentioned above could lead to different responses to the same questions, affecting their overall accuracy.

In our study, the narrative coherence shows that AI chatbots mainly offer a broader perspective on the discussed topic using logical processes rather than just providing a simple answer [ 53 ]. This can be explained by the computational abilities of AI chatbots and their capacity to understand and analyze text by recognizing word connections and predicting future words in a sentence [ 63 ]. However, it is important to note that our findings are preliminary, and more research is needed to investigate how narrative coherence changes with advancements in AI chatbot technology and updates.

Implications and future perspective

Our study identifies two contrasting implications of using AI chatbots in education. The positive implication regards AI chatbots as a valuable resource, while the negative implication perceives them as a potential threat. First, our study sheds light on the potential role of AI chatbots as supportive tools to assist candidates in preparation for the Italian entrance university standardized examination test of healthcare science. They can complement the traditional learning methods such as textbooks or in-person courses [ 10 ]. AI chatbots can facilitate self-directed learning, provide explanations and insights on the topics studied, select and filter materials and can be personalized to meet the needs of individual students [ 10 ]. In addition to the knowledge components, these instruments contribute to developing competencies, as defined by the World Health Organization [ 64 ]. Virtual simulation scenarios could facilitate the development of targeted skills and attitudes where students have a virtual interlocutor with a dynamic and human-like approach driven by AI. However, we should highlight that they cannot replace the value of reflection and discussion with peers and teachers, which are crucial for developing meta-competencies of today's students and tomorrow's healthcare professionals [ 10 ]. Conversely, candidates must be protected from simply attempting to use these tools to answer questions while administering exams. Encouraging honesty by avoiding placing and using devices (e.g., mobile phones, tablets) in classrooms is important. Candidates must be encouraged to respond with their preparation and knowledge, given that they are mostly applying for professions where honesty and ethical principles are imperative.

Strengths and limitations

As a strength, we evaluated the comparative accuracy of three AI chatbots in the Italian health sciences university admissions test over the past 13 years on a large sample of questions, considering the narrative consistency of their responses. This enriches the international debate on this topic and provides valuable insights into the strengths and limitations of AI chatbots in the context of university education [ 2 , 3 , 8 , 9 , 11 ].

However, limitations exist and offer opportunities for future study. Firstly, we only used the CINECA Test, while other universities in Italy adopted different tests (e.g., CASPUR and SELECTA). Secondly, we studied three AI Chatbots without considering others presented in the market (e.g., Cloude, Perplexity) [ 31 ]. Thirdly, we adopted both paid (ChatGPT-4) and free (Microsoft Copilot and Google Gemini) versions of AI Chatbots. Although this choice may be a limitation, we aimed to use the most up-to-date and recent versions of the AI Chatbots available when the study was performed. Fourthly, although we inputted all queries into AI chatbots, we processed only some of them as only Microsoft Copilot was able to analyse complex images, as reported in the CINECA Tests, at the time of our study [ 65 , 66 , 67 ]. Fifthly, we inputted the test questions only once to simulate the test execution conditions in real educational contexts [ 32 ], although previous studies have prompted the test questions multiple times in AI chatbots to obtain better results [ 68 ]. However, an AI language model operates differently from regular, deterministic software. These models are probabilistic in nature, forming responses by estimating the probability of the next word according to statistical patterns in their training data [ 69 ]. Consequently, posing the same question twice may not always yield identical answers. Sixthly, we did not calculate the response time of the AI chatbots since this variable is affected by the speed of the internet connection and data traffic [ 51 ]. Seventhly, we assessed the accuracy of AI chatbots in a single country by prompting questions in Italian, which may limit the generalizability of our findings to other contexts and languages [ 70 , 71 ]. Finally, we did not compare the responses of AI chatbots with those of human students since there is no national ranking for admission in Italy, and each university draws up its ranking on its own.

AI chatbots have shown promising accuracy in quickly predicting correct answers, producing writing that is grammatically correct and coherent in a conversation for the Italian entrance university standardized examination test of healthcare science degrees. However, the study provides data regarding the overall performances of different AI Chatbots with regard to the standardized examinations provided in the last 13 years to all candidates willing to enter a healthcare science degree in Italy. Therefore, findings should be placed in the context of a research exercise and may support the current debate regarding the use of AI chatbots in the academic context. Further research is needed to explore the potential of AI chatbots in other educational contexts and to address their limitations as an innovative tool for education and test preparation.

Availability of data and materials

The datasets generated and/or analysed during the current study are available in the Open Science Framework (OSF) repository, https://osf.io/ue5wf/ .

Abbreviations

• Artificial intelligence

Confidence interval

Consorzio Interuniversitario per il Calcolo Automatico dell'Italia Nord Orientale

Generative pre-trained transformer

Interquartile range

Language model for dialogue application

Pathways language model

Strengthening of Reporting of Observational Studies in Epidemiology

Redazione. Test d’ammissione professioni sanitarie, il 14 settembre 2023. Sanità Informazione. 2023. https://www.sanitainformazione.it/professioni-sanitarie/1settembre-test-dammissione-alle-professioni-sanitarie-fissato-per-il-14-settembre-2023-alle-ore-13-in-tutta-italia/ . Accessed 6 May 2024.

Kung TH, Cheatham M, Medenilla A, Sillos C, De Leon L, Elepaño C, et al. Performance of ChatGPT on USMLE: Potential for AI-assisted medical education using large language models. PLOS Digit Health. 2023;2:e0000198.

Article   Google Scholar  

Rossettini G, Cook C, Palese A, Pillastrini P, Turolla A. Pros and cons of using artificial intelligence Chatbots for musculoskeletal rehabilitation management. J Orthop Sports Phys Ther. 2023;53:1–17.

Fütterer T, Fischer C, Alekseeva A, Chen X, Tate T, Warschauer M, et al. ChatGPT in education: global reactions to AI innovations. Sci Rep. 2023;13:15310.

Mohammadi S, SeyedAlinaghi S, Heydari M, Pashaei Z, Mirzapour P, Karimi A, et al. Artificial intelligence in COVID-19 Management: a systematic review. J Comput Sci. 2023;19:554–68.

Mehraeen E, Mehrtak M, SeyedAlinaghi S, Nazeri Z, Afsahi AM, Behnezhad F, et al. Technology in the Era of COVID-19: a systematic review of current evidence. Infect Disord Drug Targets. 2022;22:e240322202551.

SeyedAlinaghi S, Abbaspour F, Mehraeen E. The Challenges of ChatGPT in Healthcare Scientific Writing. Shiraz E-Med J. 2024;25(2):e141861. https://doi.org/10.5812/semj-141861 .

Karabacak M, Ozkara BB, Margetis K, Wintermark M, Bisdas S. The advent of generative language models in medical education. JMIR Med Educ. 2023;9:e48163.

Mohammad B, Supti T, Alzubaidi M, Shah H, Alam T, Shah Z, et al. The pros and cons of using ChatGPT in medical education: a scoping review. Stud Health Technol Inform. 2023;305:644–7.

Google Scholar  

Abd-Alrazaq A, AlSaad R, Alhuwail D, Ahmed A, Healy PM, Latifi S, et al. Large language models in medical education: opportunities, challenges, and future directions. JMIR Med Educ. 2023;9:e48291.

Azer SA, Guerrero APS. The challenges imposed by artificial intelligence: are we ready in medical education? BMC Med Educ. 2023;23:680.

Levin G, Horesh N, Brezinov Y, Meyer R. Performance of ChatGPT in medical examinations: a systematic review and a meta-analysis. BJOG Int J Obstet Gynaecol. 2023. https://doi.org/10.1111/1471-0528.17641 .

Passby L, Jenko N, Wernham A. Performance of ChatGPT on Specialty Certificate Examination in Dermatology multiple-choice questions. Clin Exp Dermatol. 2023:llad197. https://doi.org/10.1093/ced/llad197 .

Lewandowski M, Łukowicz P, Świetlik D, Barańska-Rybak W. ChatGPT-3.5 and ChatGPT-4 dermatological knowledge level based on the Specialty Certificate Examination in Dermatology. Clin Exp Dermatol. 2023:llad255. https://doi.org/10.1093/ced/llad255 .

Giannos P. Evaluating the limits of AI in medical specialisation: ChatGPT’s performance on the UK Neurology Specialty Certificate Examination. BMJ Neurol Open. 2023;5:e000451.

Teebagy S, Colwell L, Wood E, Yaghy A, Faustina M. Improved performance of ChatGPT-4 on the OKAP examination: a comparative study with ChatGPT-3.5. J Acad Ophthalmol. 2023;15:e184–7.

Madrid-García A, Rosales-Rosado Z, Freites-Nuñez D, Pérez-Sancristóbal I, Pato-Cour E, Plasencia-Rodríguez C, et al. Harnessing ChatGPT and GPT-4 for evaluating the rheumatology questions of the Spanish access exam to specialized medical training. Sci Rep. 2023;13:22129.

Haze T, Kawano R, Takase H, Suzuki S, Hirawa N, Tamura K. Influence on the accuracy in ChatGPT: Differences in the amount of information per medical field. Int J Med Inf. 2023;180:105283.

Yanagita Y, Yokokawa D, Uchida S, Tawara J, Ikusaka M. Accuracy of ChatGPT on medical questions in the national medical licensing examination in Japan: evaluation study. JMIR Form Res. 2023;7:e48023.

Rosoł M, Gąsior JS, Łaba J, Korzeniewski K, Młyńczak M. Evaluation of the performance of GPT-3.5 and GPT-4 on the polish medical final examination. Sci Rep. 2023;13:20512.

Brin D, Sorin V, Vaid A, Soroush A, Glicksberg BS, Charney AW, et al. Comparing ChatGPT and GPT-4 performance in USMLE soft skill assessments. Sci Rep. 2023;13:16492.

Kaneda Y, Takahashi R, Kaneda U, Akashima S, Okita H, Misaki S, et al. Assessing the performance of GPT-3.5 and GPT-4 on the 2023 Japanese nursing examination. Cureus. 2023;15:e42924.

Kleinig O, Gao C, Bacchi S. This too shall pass: the performance of ChatGPT-3.5, ChatGPT-4 and new bing in an Australian medical licensing examination. Med J Aust. 2023;219:237.

Roos J, Kasapovic A, Jansen T, Kaczmarczyk R. Artificial intelligence in medical education: comparative analysis of ChatGPT, Bing, and medical students in Germany. JMIR Med Educ. 2023;9:e46482.

Ali R, Tang OY, Connolly ID, Fridley JS, Shin JH, Zadnik Sullivan PL, et al. Performance of ChatGPT, GPT-4, and google bard on a neurosurgery oral boards preparation question bank. Neurosurgery. 2023. https://doi.org/10.1227/neu.0000000000002551 .

Patil NS, Huang RS, van der Pol CB, Larocque N. Comparative Performance of ChatGPT and Bard in a Text-Based Radiology Knowledge Assessment. Can Assoc Radiol J. 2024;75(2):344–50. https://doi.org/10.1177/08465371231193716 .

Toyama Y, Harigai A, Abe M, Nagano M, Kawabata M, Seki Y, et al. Performance evaluation of ChatGPT, GPT-4, and bard on the official board examination of the Japan radiology society. Jpn J Radiol. 2023. https://doi.org/10.1007/s11604-023-01491-2 .

Fowler T, Pullen S, Birkett L. Performance of ChatGPT and Bard on the official part 1 FRCOphth practice questions. Br J Ophthalmol. 2023:bjo-2023-324091. https://doi.org/10.1136/bjo-2023-324091 . Online ahead of print.

Meo SA, Al-Khlaiwi T, AbuKhalaf AA, Meo AS, Klonoff DC. The Scientific Knowledge of Bard and ChatGPT in Endocrinology, Diabetes, and Diabetes Technology: Multiple-Choice Questions Examination-Based Performance. J Diabetes Sci Technol. 2023:19322968231203987. https://doi.org/10.1177/19322968231203987 . Online ahead of print.

Kumari A, Kumari A, Singh A, Singh SK, Juhi A, Dhanvijay AKD, et al. Large language models in hematology case solving: a comparative study of ChatGPT-3.5, google bard, and microsoft bing. Cureus. 2023;15:e43861.

Dhanvijay AKD, Pinjar MJ, Dhokane N, Sorte SR, Kumari A, Mondal H. Performance of large language models (ChatGPT, Bing Search, and Google Bard) in solving case vignettes in physiology. Cureus. 2023;15:e42972.

Giannakopoulos K, Kavadella A, Aaqel Salim A, Stamatopoulos V, Kaklamanos EG. Evaluation of generative artificial intelligence large language models ChatGPT, google bard, and microsoft bing chat in supporting evidence-based dentistry: a comparative mixed-methods study. J Med Internet Res. 2023. https://doi.org/10.2196/51580 .

Torres-Zegarra BC, Rios-Garcia W, Ñaña-Cordova AM, Arteaga-Cisneros KF, Chalco XCB, Ordoñez MAB, et al. Performance of ChatGPT, Bard, Claude, and Bing on the Peruvian National licensing medical examination: a cross-sectional study. J Educ Eval Health Prof. 2023;20.

Giannos P, Delardas O. Performance of ChatGPT on UK standardized admission tests: insights from the BMAT, TMUA, LNAT, and TSA examinations. JMIR Med Educ. 2023;9:e47737.

Guigue P-A, Meyer R, Thivolle-Lioux G, Brezinov Y, Levin G. Performance of ChatGPT in French language Parcours d’Accès Spécifique Santé test and in OBGYN. Int J Gynaecol Obstet Off Organ Int Fed Gynaecol Obstet. 2023. https://doi.org/10.1002/ijgo.15083 .

Healthcare Science. NSHCS. https://nshcs.hee.nhs.uk/healthcare-science/ . Accessed 6 May 2024.

Alessandri Bonetti M, Giorgino R, Gallo Afflitto G, De Lorenzi F, Egro FM. How does ChatGPT perform on the Italian residency admission national exam compared to 15,869 medical graduates? Ann Biomed Eng. 2023. https://doi.org/10.1007/s10439-023-03318-7 .

Scaioli G, Moro GL, Conrado F, Rosset L, Bert F, Siliquini R. Exploring the potential of ChatGPT for clinical reasoning and decision-making: a cross-sectional study on the Italian medical residency exam. Ann DellIstituto Super Sanità. 2023;59:267–70.

von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP, et al. The strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. PLoS Med. 2007;4:e296.

Nowell J. Guide to ethical approval. BMJ. 2009;338:b450.

Accesso programmato a livello nazionale. Mi - Ministero dell’istruzione. https://www.miur.gov.it/accesso-programmato-a-livello-nazionale . Accessed 6 May 2024.

Il Consorzio. Cineca. http://www.cineca.it/chi-siamo/il-consorzio . Accessed 6 May 2024.

Salute M della. Professioni sanitarie. https://www.salute.gov.it/portale/professioniSanitarie/dettaglioContenutiProfessioniSanitarie.jsp?lingua=italiano&id=808&area=professioni-sanitarie&menu=vuoto&tab=1 . Accessed 6 May 2024.

Test d’ingresso ai corsi ad accesso programmato e alle scuole di specializzazione. Cineca. http://www.cineca.it/sistemi-informativi-miur/studenti-carriere-offerta-formativa-e-altri-servizi/test-dingresso-ai . Accessed 6 May 2024.

Scuola secondaria di secondo grado. Mi - Ministero dell’istruzione. https://www.miur.gov.it/scuola-secondaria-di-secondo-grado . Accessed 6 May 2024.

Test ammissione professioni sanitarie anni precedenti. TaxiTest. https://taxitest.it/test-ingresso-professioni-sanitarie-anni-passati/ . Accessed 6 May 2024.

Soluzioni dei Test d’Ingresso per Professioni Sanitarie 2023. https://www.studentville.it/app/uploads/2023/09/soluzioni-test-cineca-professioni-sanitarie-2023.pdf . Accessed 6 May 2024.

ChatGPT. https://chat.openai.com . Accessed 6 May 2024.

Microsoft Copilot: il tuo AI Companion quotidiano. Microsoft Copilot: il tuo AI Companion quotidiano. https://ceto.westus2.binguxlivesite.net/ . Accessed 6 May 2024.

Gemini: chatta per espandere le tue idee. Gemini. https://gemini.google.com . Accessed 6 May 2024.

Mihalache A, Popovic MM, Muni RH. Performance of an artificial intelligence chatbot in ophthalmic knowledge assessment. JAMA Ophthalmol. 2023;141:589–97.

Trabasso T. The Development of Coherence in Narratives by Understanding Intentional Action. In: Stelmach GE, Vroon PA, editors. Advances in Psychology. Vol. 79. North-Holland; 1991. p. 297–314. ISSN 0166-4115, ISBN 9780444884848. https://doi.org/10.1016/S0166-4115(08)61559-9 .

Gilson A, Safranek CW, Huang T, Socrates V, Chi L, Taylor RA, et al. How does ChatGPT perform on the United States medical licensing examination? the implications of large language models for medical education and knowledge assessment. JMIR Med Educ. 2023;9:e45312.

Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33:159–74.

Saravia E. Prompt Engineering Guide. https://github.com/dair-ai/Prompt-Engineering-Guide . 2022. Accessed 6 May 2024.

Giray L. Prompt engineering with ChatGPT: a guide for academic writers. Ann Biomed Eng. 2023. https://doi.org/10.1007/s10439-023-03272-4 .

Massey PA, Montgomery C, Zhang AS. Comparison of ChatGPT–3.5, ChatGPT-4, and orthopaedic resident performance on orthopaedic assessment examinations. JAAOS -. J Am Acad Orthop Surg. 2023;31:1173.

Guerra GA, Hofmann H, Sobhani S, Hofmann G, Gomez D, Soroudi D, et al. GPT-4 artificial intelligence model outperforms ChatGPT, medical students, and neurosurgery residents on neurosurgery written board-like questions. World Neurosurg. 2023;S1878–8750(23):01144.

Cuthbert R, Simpson AI. Artificial intelligence in orthopaedics: can chat generative pre-trained transformer (ChatGPT) pass Section 1. Postgrad Med J. 2023;99:1110–4.

Friederichs H, Friederichs WJ, März M. ChatGPT in medical school: how successful is AI in progress testing? Med Educ Online. 2023;28.

Weng T-L, Wang Y-M, Chang S, Chen T-J, Hwang S-J. ChatGPT failed Taiwan’s family medicine board exam. J Chin Med Assoc JCMA. 2023;86:762–6.

Bhayana R, Krishna S, Bleakney RR. Performance of ChatGPT on a radiology board-style examination: insights into current strengths and limitations. Radiology. 2023;307:e230582.

Thirunavukarasu AJ, Ting DSJ, Elangovan K, Gutierrez L, Tan TF, Ting DSW. Large language models in medicine. Nat Med. 2023;29:1930–40.

Global competency framework for universal health coverage. https://www.who.int/publications-detail-redirect/9789240034686 . Accessed 6 May 2024.

ChatGPT — Release Notes | OpenAI Help Center. https://help.openai.com/en/articles/6825453-chatgpt-release-notes . Accessed 6 May 2024.

Microsoft. Visual Search API | Microsoft Bing. Bingapis. https://www.microsoft.com/en-us/bing/apis/bing-visual-search-api . Accessed 6 May 2024.

What’s ahead for Bard: More global, more visual, more integrated. Google. 2023. https://blog.google/technology/ai/google-bard-updates-io-2023/ . Accessed 6 May 2024.

Zhu L, Mou W, Yang T, Chen R. ChatGPT can pass the AHA exams: Open-ended questions outperform multiple-choice format. Resuscitation. 2023;188:109783.

Probabilistic machine learning and artificial intelligence | Nature. https://www.nature.com/articles/nature14541 . Accessed 6 May 2024.

Ebrahimian M, Behnam B, Ghayebi N, Sobhrakhshankhah E. ChatGPT in Iranian medical licensing examination: evaluating the diagnostic accuracy and decision-making capabilities of an AI-based model. BMJ Health Care Inform. 2023;30:e100815.

Seghier ML. ChatGPT: not all languages are equal. Nature. 2023;615:216.

Download references

Acknowledgements

The authors thanks Sanitätsbetrieb der Autonomen Provinz Bozen/Azienda Sanitaria della Provincia Autonoma di Bolzano for covering the open access publication costs.

The authors declare that they receive fundings from the Department of Innovation, Research, University and Museums of the Autonomous Province of Bozen/Bolzano for covering the open access publication costs of this study.

Author information

Silvia Gianola and Alvisa Palese both authors have contributed equally.

Authors and Affiliations

School of Physiotherapy, University of Verona, Verona, Italy

Giacomo Rossettini

Department of Physiotherapy, Faculty of Sport Sciences, Universidad Europea de Madrid, Villaviciosa de Odón, 28670, Spain

Department of Rehabilitation, Hospital of Merano (SABES-ASDAA), Teaching Hospital of Paracelsus Medical University (PMU), Merano-Meran, Italy

Lia Rodeghiero

School of Speech Therapy, University of Verona, Verona, Italy

Federica Corradi

Department of Orthopaedics, Duke University, Durham, NC, USA

Duke Clinical Research Institute, Duke University, Durham, NC, USA

Department of Population Health Sciences, Duke University, Durham, NC, USA

Department of Biomedical and Neuromotor Sciences (DIBINEM), Alma Mater University of Bologna, Bologna, Italy

Paolo Pillastrini & Andrea Turolla

Unit of Occupational Medicine, IRCCS Azienda Ospedaliero-Universitaria Di Bologna, Bologna, Italy

Unit of Clinical Epidemiology, IRCCS Istituto Ortopedico Galeazzi, Milan, Italy

Greta Castellini & Silvia Gianola

Department of Medical Sciences, University of Udine, Udine, Italy

Stefania Chiappinotto & Alvisa Palese

You can also search for this author in PubMed   Google Scholar

Contributions

GR, SG, AP conceived and designed the research and wrote the first draft. LR, FC, managed the acquisition of data. SG, GC, SC, CC, PP, AT managed the analysis and interpretation of data. GR, SG, AP wrote the first draft. All authors read, revised, wrote and approved the final version of manuscript.

Authors' information

A multidisciplinary group of healthcare science educators promoted and developed this study in Italy. The group consisted of professors, lecturers, and tutors actively involved in university education in different healthcare science disciplines (e.g., rehabilitation, physiotherapy, speech therapy, nursing).

Corresponding authors

Correspondence to Giacomo Rossettini , Lia Rodeghiero , Stefania Chiappinotto , Silvia Gianola or Alvisa Palese .

Ethics declarations

Ethics approval and consent to participate.

Not applicable, no humans and patients have been involved in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Supplementary material 1., supplementary material 2., supplementary material 3., supplementary material 4., supplementary material 5., rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Rossettini, G., Rodeghiero, L., Corradi, F. et al. Comparative accuracy of ChatGPT-4, Microsoft Copilot and Google Gemini in the Italian entrance test for healthcare sciences degrees: a cross-sectional study. BMC Med Educ 24 , 694 (2024). https://doi.org/10.1186/s12909-024-05630-9

Download citation

Received : 24 January 2024

Accepted : 04 June 2024

Published : 26 June 2024

DOI : https://doi.org/10.1186/s12909-024-05630-9

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Health occupations
• Physical therapy modalities
• Speech therapy

BMC Medical Education

ISSN: 1472-6920

• Submission enquiries: [email protected]
• General enquiries: [email protected]

Information

• Author Services

Initiatives

You are accessing a machine-readable page. In order to be human-readable, please install an RSS reader.

All articles published by MDPI are made immediately available worldwide under an open access license. No special permission is required to reuse all or part of the article published by MDPI, including figures and tables. For articles published under an open access Creative Common CC BY license, any part of the article may be reused without permission provided that the original article is clearly cited. For more information, please refer to https://www.mdpi.com/openaccess .

Feature papers represent the most advanced research with significant potential for high impact in the field. A Feature Paper should be a substantial original Article that involves several techniques or approaches, provides an outlook for future research directions and describes possible research applications.

Feature papers are submitted upon individual invitation or recommendation by the scientific editors and must receive positive feedback from the reviewers.

Editor’s Choice articles are based on recommendations by the scientific editors of MDPI journals from around the world. Editors select a small number of articles recently published in the journal that they believe will be particularly interesting to readers, or important in the respective research area. The aim is to provide a snapshot of some of the most exciting work published in the various research areas of the journal.

Original Submission Date Received: .

• Active Journals
• Find a Journal
• Proceedings Series
• For Authors
• For Reviewers
• For Editors
• For Librarians
• For Publishers
• For Societies
• For Conference Organizers
• Open Access Policy
• Institutional Open Access Program
• Special Issues Guidelines
• Editorial Process
• Research and Publication Ethics
• Article Processing Charges
• Testimonials
• Preprints.org
• SciProfiles
• Encyclopedia

Article Menu

• Subscribe SciFeed
• Recommended Articles
• Author Biographies
• Google Scholar
• on Google Scholar
• Table of Contents

Find support for a specific problem in the support section of our website.

Please let us know what you think of our products and services.

Visit our dedicated information section to learn more about MDPI.

JSmol Viewer

Design and research of a strain elastic element with a double-layer cross floating beam for strain gauge wireless rotating dynamometers.

1. Introduction

2. design and research of strain elastic element with double-layer cross floating beam, 2.1. structural design of strain gauge wireless rotating dynamometer, 2.2. structural design of the strain elastic element, 2.3. strain and deformation segmented rigid body model of double-layer cross floating beam, 2.3.1. strain and deformation analysis of double-layer cross floating beam under the action of f x, 2.3.2. strain and deformation analysis of double-layer cross floating beam under the action of f z, 2.4. comparison of fea solution and model solution for strain and deformation of double-layer cross floating beam, 2.5. size optimization of double-layer cross floating beam, 2.6. overall analysis of the strain elastic element, 3. strain gauge arrangement and testing, 3.1. strain gauge arrangement, 3.2. static calibration testing, 3.3. free modal testing, 3.4. cutting test, 4. conclusions.

• This paper designed a compact strain elastic element with a double-layer cross floating beam. The maximum outer diameter is only 75 mm and the maximum thickness is 25 mm. The design of the double-layer cross floating beam allows the strain elastic element to improve the sensitivity and reduce the cross-sensitivity error while ensuring the overall stiffness;
• Based on the proposed strain elastic element, a strain gauge wireless rotating dynamometer with compact size is designed. The overall diameter of the dynamometer is 170 mm and the axial size is 286 mm. In order to facilitate integration, a ring-shaped data acquisition and wireless transmission module PCB is designed;
• The static model of the double-layer cross floating beam on the strain elastic element was established by the segmented rigid body method. The rationality of the model was verified by comparison with the finite element results. According to the obtained static model, the structural parameters of the double-layer cross floating beam were optimized using the sequential quadratic programming algorithm to maximize the sensitivity of the floating beam;
• The strain elastic element was analyzed using finite element software, and the strain of the structure under simulation conditions was obtained, which provided a reference for subsequent calibration tests and circuit design;
• Through static calibration tests, the sensitivities of the strain elastic element in the four directions of F X , F Y , F Z , and M Z are determined to be 3.3 mV/N, 2.7 mV/N, 1.6 mV/N, and 104.1 mV/Nm, respectively, and the maximum cross-sensitivity error does not exceed 1%. Through modal tests in the free state, the natural frequency of the strain elastic element is determined to be 2954 Hz. The results of cutting tests show that the strain elastic element can obtain the change in cutting force and the tool-passing frequency. The results show that the designed strain elastic element can be applied to the strain gauge wireless rotating dynamometer to measure four-component cutting forces under medium- and low-speed conditions.

Author Contributions

Data availability statement, conflicts of interest.

• Imad, M.; Hopkins, C.; Hosseini, A.; Yussefian, N.Z.; Kishawy, H.A. Intelligent machining: A review of trends, achievements and current progress. Int. J. Comput. Integr. Manuf. 2022 , 35 , 359–387. [ Google Scholar ] [ CrossRef ]
• Sousa, V.F.C.; Silva, F.J.G.; Fecheira, J.S.; Lopes, H.M.; Martinho, R.P.; Casais, R.B.; Ferreira, L.P. Cutting Forces Assessment in CNC Machining Processes: A Critical Review. Sensors 2020 , 20 , 4536. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Duan, Z.J.; Li, C.H.; Ding, W.F.; Zhang, Y.B.; Yang, M.; Gao, T.; Cao, H.J.; Xu, X.F.; Wang, D.Z.; Mao, C.; et al. Milling Force Model for Aviation Aluminum Alloy: Academic Insight and Perspective Analysis. Chin. J. Mech. Eng. 2021 , 34 , 35. [ Google Scholar ] [ CrossRef ]
• Wang, B.; Liu, Z.Q.; Cai, Y.K.; Luo, X.C.; Ma, H.F.; Song, Q.H.; Xiong, Z.H. Advancements in material removal mechanism and surface integrity of high speed metal cutting: A review. Int. J. Mach. Tools Manuf. 2021 , 166 , 32. [ Google Scholar ] [ CrossRef ]
• Zhu, L.D.; Liu, C.F. Recent progress of chatter prediction, detection and suppression in milling. Mech. Syst. Signal Process. 2020 , 143 , 37. [ Google Scholar ] [ CrossRef ]
• Pimenov, D.Y.; Bustillo, A.; Wojciechowski, S.; Sharma, V.S.; Gupta, M.K.; Kuntoglu, M. Artificial intelligence systems for tool condition monitoring in machining: Analysis and critical review. J. Intell. Manuf. 2023 , 34 , 2079–2121. [ Google Scholar ] [ CrossRef ]
• Lu, C. Study on prediction of surface quality in machining process. J. Mater. Process. Technol. 2008 , 205 , 439–450. [ Google Scholar ] [ CrossRef ]
• Xie, Z.Y.; Lu, Y.; Chen, X.L. A multi-sensor integrated smart tool holder for cutting process monitoring. Int. J. Adv. Manuf. Technol. 2020 , 110 , 853–864. [ Google Scholar ] [ CrossRef ]
• Totis, G.; Wirtz, G.; Sortino, M.; Veselovac, D.; Kuljanic, E.; Klocke, F. Development of a dynamometer for measuring individual cutting edge forces in face milling. Mech. Syst. Signal Process. 2010 , 24 , 1844–1857. [ Google Scholar ] [ CrossRef ]
• Luo, M.; Luo, H.; Axinte, D.; Liu, D.S.; Mei, J.W.; Liao, Z.R. A wireless instrumented milling cutter system with embedded PVDF sensors. Mech. Syst. Signal Process. 2018 , 110 , 556–568. [ Google Scholar ] [ CrossRef ]
• Ma, L.; Melkote, S.N.; Castle, J.B. PVDF sensor-based monitoring of milling torque. Int. J. Adv. Manuf. Technol. 2014 , 70 , 1603–1614. [ Google Scholar ] [ CrossRef ]
• Wu, F.H.; Li, Y.X.; Guo, B.S.; Zhang, P.F. The Design of Force Measuring Tool Holder System Based on Wireless Transmission. IEEE Access 2018 , 6 , 38556–38566. [ Google Scholar ] [ CrossRef ]
• Qin, Y.F.; Zhao, Y.L.; Li, Y.X.; Zhao, Y.; Wang, P. A novel dynamometer for monitoring milling process. Int. J. Adv. Manuf. Technol. 2017 , 92 , 2535–2543. [ Google Scholar ] [ CrossRef ]
• Xie, Z.Y.; Lu, Y.; Li, J.G. Development and testing of an integrated smart tool holder for four-component cutting force measurement. Mech. Syst. Signal Process. 2017 , 93 , 225–240. [ Google Scholar ] [ CrossRef ]
• Zhang, P.F.; Gao, D.; Lu, Y.; Xie, Z.Y.; Wang, Z.Q. Development and testing of a wireless multi-axis toolholder dynamometer for milling and drilling process. Proc. Inst. Mech. Eng. Part B-J. Eng. Manuf. 2023 , 237 , 181–191. [ Google Scholar ] [ CrossRef ]
• Liu, M.Y.; Bing, J.J.; Xiao, L.; Yun, K.; Wan, L. Development and Testing of an Integrated Rotating Dynamometer Based on Fiber Bragg Grating for Four-Component Cutting Force Measurement. Sensors 2018 , 18 , 1254. [ Google Scholar ] [ CrossRef ] [ PubMed ]
• Rizal, M.; Ghani, J.A.; Nuawi, M.Z.; Haron, C.H.C. An embedded multi-sensor system on the rotating dynamometer for real-time condition monitoring in milling. Int. J. Adv. Manuf. Technol. 2018 , 95 , 811–823. [ Google Scholar ] [ CrossRef ]
• Rizal, M.; Ghani, J.A.; Nuawi, M.Z.; Haron, C.H.C. Development and testing of an integrated rotating dynamometer on tool holder for milling process. Mech. Syst. Signal Process. 2015 , 52–53 , 559–576. [ Google Scholar ] [ CrossRef ]
• Zhang, P.F.; Gao, D.; Lu, Y.; Wang, F.L.; Liao, Z.R. A novel smart toolholder with embedded force sensors for milling operations. Mech. Syst. Signal Process. 2022 , 175 , 25. [ Google Scholar ] [ CrossRef ]
• Kang, M.K.; Lee, S.; Kim, J.H. Shape optimization of a mechanically decoupled six-axis force/torque sensor. Sens. Actuator A-Phys. 2014 , 209 , 41–51. [ Google Scholar ] [ CrossRef ]
• Akbari, H.; Kazerooni, A. Improving the coupling errors of a Maltese cross-beams type six-axis force/moment sensor using numerical shape-optimization technique. Measurement 2018 , 126 , 342–355. [ Google Scholar ] [ CrossRef ]
• Kim, G.S. Design of a six-axis wrist force/moment sensor using FEM and its fabrication for an intelligent robot. Sens. Actuator A-Phys. 2007 , 133 , 27–34. [ Google Scholar ] [ CrossRef ]
• Kim, H.M.; Hong, T.K.; Kim, G.S. Design of a wrist rotation rehabilitation robot. In Proceedings of the 4th Annual IEEE International Conference on Cyber Technology in Automation, Control and Intelligent, Hong Kong, China, 4–7 June 2014; pp. 240–245. [ Google Scholar ]
• Rizal, M.; Ghani, J.A.; Usman, H.; Dirhamsyah, M.; Mubarak, A.Z. Development and testing of a stationary dynamometer using cross-beam-type force-sensing elements for three-axis cutting force measurement in milling operations. Proc. Inst. Mech. Eng. Part B-J. Eng. Manuf. 2024 , 238 , 950–961. [ Google Scholar ] [ CrossRef ]

Click here to enlarge figure

Share and Cite

Wang, Q.; Wu, W.; Zhao, Y.; Cheng, Y.; Liu, L.; Yan, K. Design and Research of a Strain Elastic Element with a Double-Layer Cross Floating Beam for Strain Gauge Wireless Rotating Dynamometers. Micromachines 2024 , 15 , 857. https://doi.org/10.3390/mi15070857

Wang Q, Wu W, Zhao Y, Cheng Y, Liu L, Yan K. Design and Research of a Strain Elastic Element with a Double-Layer Cross Floating Beam for Strain Gauge Wireless Rotating Dynamometers. Micromachines . 2024; 15(7):857. https://doi.org/10.3390/mi15070857

Wang, Qinan, Wenge Wu, Yongjuan Zhao, Yunping Cheng, Lijuan Liu, and Kaiqiang Yan. 2024. "Design and Research of a Strain Elastic Element with a Double-Layer Cross Floating Beam for Strain Gauge Wireless Rotating Dynamometers" Micromachines 15, no. 7: 857. https://doi.org/10.3390/mi15070857

Article Metrics

Article access statistics, further information, mdpi initiatives, follow mdpi.

Subscribe to receive issue release notifications and newsletters from MDPI journals