By Aashi Lalit & Caldia Choi
When we think about how Artificial Intelligence (AI) could affect universities and colleges around the world, especially those teaching subjects as important as medicine, we don’t usually see the benefits, or the numerous possibilities that AI opens in places of education. However, when judging the impact, understanding both the benefits and the drawbacks are a must. On one hand, we tend to worry about students plagiarising assignments from chat-bots such as Chat-GPT or cheating on important exams, and therefore potentially not gaining the knowledge needed to appropriately and consistently treat patients. There’s also the issue of worrying about students over-relying on AI for their lectures, homework and revision which may result in students’ skills in organising and time management to decline .Moreover, it will cause universities a huge amount of expense in developing the suitable AI system for its students, as different medical schools have different methods of teaching styles. On the other hand, although a full ban of AI may mean these problems do not arise, they also prevent the use of AI in a productive way to in fact improve a student’s education. These benefits include curating curriculums to fit an individual, helping teachers spend less time on admin and more on engaging students in practical work, improving research and understanding of technology in medicine, and even using AI in virtual reality simulations.
The Benefits of Artificial Intelligence in Medical School
Aashi Lalit
Artificial Intelligence being used in medical training could lead to a more accessible, well-rounded, and overall enhanced education.
Understanding that AI is a massively beneficial tool is the first step to realising how effective it could prove to be in medical training. In medicine as a whole, AI could in the future do multiple different things, such as treatment recommendations to patient queries, removing the need for a doctor to be present, or curating treatments personalised to specific patients; medical record keeping, and other repetitive tasks, taking away the need for admin work to be done manually, decreasing medical costs; faster and accurate diagnosis, to help doctors make decisions quickly; and analysing medical images, from machines such as X-rays, MRIs or CT scans. All of these uses show just how instrumental AI could be in revolutionising the healthcare sector, however, this article is about the advantages of AI’s implementation in medical school. So, how does AI perform in a medical training context? One of the best ways to answer this question is to test AI’s abilities in medical exams, and there have been many experiments of this type. One big one was conducted in June/July 2023, and proved that Chat-GPT could successfully answer open-ended questions on the case-report portion of the United States Medical License Examination (USMLE), which is an exam that doctors have to achieve a passing mark in to practise. GPT-4 was used in this experiment, and compared to GPT-3.5, which was the AI program used in a similar previous experiment, there were fewer AI hallucinations, where an AI manufactures incorrect information. The AI scored, on average, more than four points higher than students on this part of the exam. This really shows the potential that AI has to be valuable in medical school, due to its ability to understand information and answer questions correctly, to an extent.
The biggest reason for Chat-GPT’s daily users shooting up to 100 million users in just 2 months, when popular social media applications such as Snapchat and Whatsapp took over 3.5 years, was that it was exciting, incredible, pioneering, the next big thing. It was also remarkably accessible for all: all that one needed was an email address. I, for one, created an account within days of finding out about it, and so did many others, diving headfirst into asking questions about recipes, advice about birthday presents, and so much more general and subject-specific knowledge, that Chat-GPT, this amazing invention, had immense knowledge about. This quickly brings me onto why the accessibility of AI tools like Chat-GPT is so important: any student of any subject, across the world, can use it. This, of course, includes medical students. Because of the simple access everyone with an email address has to AI chatbots, a student studying anatomy, or diseases in the human body, could get definitions and explanations tailored specifically to what they’re unsure about, which could be incredibly helpful, especially if they are unable to reach their professor or find concise information online. It saves time from searching multiple websites, and with AI chatbots increasingly linking websites where they’ve gotten information from, AI ‘hallucinations’, where the chatbots fabricate information, are less likely to take place. Overall, the accessibility of AI chatbots are one of the main reasons why they could be so useful in medical schools: the students, nor the school, will have to pay to be able to access their knowledge and assistance.
Furthermore, AI could be useful for medicine professors, in a variety of ways, such as curriculum mapping and inspiration, personalised learning, automated exams, grading and other repetitive tasks. Now that teachers increasingly can use AI-algorithms for a large part of administrative and organisational work, they can spend more time interacting with their students, answering detailed questions, and teaching practical things, such as lab work or surgery. A report showed, on average across countries, teachers spend half their working time on planning lessons, marking and monitoring performance, instead of teaching, and this could be cut down and spent more meaningfully with the implementation of AI in medical schools. Also, AI can help professors create immersive lesson plans and activities, by looking at student performance in specific areas of the curriculum There are also other possibilities such as creating exams curated for a specific student, by looking in areas that their answers in previous exams showed were weak, or even marking the exams, with new AI Image Recognition Software. Although unrelated to medicine, AQA, the British exam board, will be using data from 2024 GCSEs and A-Levels to understand how similar AI’s marking could be to a certified marker. An experiment like this was conducted before. With questions worth 3 marks, there was a match of 80%, however with 6 markers, the percentage lessened to 65%. Still, with research into this field, AI marking could become very feasible. Ultimately, this shows just how beneficial AI could be for professors, especially those who teach medicine, because it’s a subject that requires a lot of practical learning.
A few years ago, a survey was conducted of trainee doctors’ estimated impact of AI on clinical training, at the UK NHS postgraduate centres in London in the final 3 months of 2020. 210 doctors responded, and 58% of them believed that AI would have a positive impact. As the figure on the right shows, a large proportion of the surveyees believed that AI could improve their research, audit and quality improvement skills’, which was largely backed by ideas given by the doctors about AI’s analysing, gathering data protocols, and generating suggestions abilities. One also talked about its ‘ability to create simulations as close to reality as possible’, and another mentioned how it would ‘free up time to do practical procedures’. These predictions, although now a few years old, have proved to be mostly correct, with the plentiful research and numerous advances in AI-powered software, and will only improve in the future.
An important reason as to why AI-based training should be implemented in medical schools, is that it allows for exposure to newly-pioneered treatments with the aid of technology. There are multiple examples of this: the University of Florida has a program where Radiology residents work with a company to create a computer-aided software to detect mammography, the Stanford University Centre for AI in Medicine and Imaging teaches students to solve healthcare problems with the aid of machine learning and the University of Pittsburgh School of Medicine has a course where they teach the application of AI and Machine Learning in Pediatric Medicine. All of these prove the great scope that AI has when combined with human knowledge to more efficiently and accurately think of solutions: learning about it in school gives doctors a wider range of knowledge for when they are finding better cures and treatments for diseases.
The most riveting way, in my opinion, that AI could be used in medical school, is in an AI-powered virtual reality. This involves a VR headset, which when put on allows a user to be completely immersed in a virtual world, which is often completely interactive. This allows medical students to learn from experience, as they perform surgeries or take medical histories. There’s even possibilities to have whole scenarios played out in a virtual reality, where the student interacts with a virtual patient, their family, nurses, and there is also virtual stress created with made-up emergencies. This all can be improved with AI, as researchers at the Stanford Computational Imaging Lab showed, with a creation called the neural holographic display, training a neural network to imitate the real-world physics of what was happening in real life, and this was joint to a ‘camera in the loop’ calibration system. With this algorithm and calibration strategy, which ran alongside the image seen, vivid and realistic visuals were created. The researchers said that this technique was the ‘big future’, and in that period of time, medical students will likely be able to use VR/AR for training, and for visualising medical data from tests, such as CT scans and X-rays, onto the patients themselves, which could have a massive impact on their understanding of the human body and the diseases which affect it. Another benefit is that they’re repeatable: the same scenarios can be run again and again, for the student to be fully confident.
The Drawbacks of Artificial Intelligence in Medical School
Caldia Choi
All cases of misused and misconduct happened during exams or coursework has always been a huge headache to teachers and education professionals.
The trend of malpractice has been increasing, according to ofqual, the number of penalties issued to students increased since the last time exam took place in the United Kingdom : 4,335 penalties were issued to students in 2022, up from 3,040 in 2019, and representing 0.03% of entries. Undoubtedly, with the introduction of AI into the educational sector, the cases of malpractice will increase; especially with the examinations now being done electronically. With the increase of malpractice cases, the exam results could be unfair compared to those who work hard for the test, leading to a false ranking of the students’ learning abilities. Moreover, it could pose a danger to patients if the medical student cheated on their exams throughout their studies in medical school, without actually understanding the material. This may lead to future issues of poor and unsafe practice in healthcare.
With the advancement of AI technology, there are progressively increasing AI tools or programs on the Web that are able to aid students to write their essays ; or even write their essays for them. For instance, ‘Editpad’, ‘Eduwriterai’, and ‘The Perfect Essay Writer’ are all examples of them. These applications claim that they generate free narrative and persuasive essays. If students use one of these AI chatbots to generate essays for their coursework, this should be considered as a form of plagiarism; even if there is no original work being copied. But at the same time, the student did not actually write and research for this piece of work ,hence it is not recognised as a piece of ‘ original work’ . There is still controversy of whether using AI tools should be counted as cheating due to the ethics behind using AI in educational settings. The ethics behind it are complicated, as different medical schools have their own understanding of it. It is important for the Medical School Council to publish guidelines addressing this issue.
In addition, there is a growing concern of students over relying on AI tools for their essays, for example, one may over rely on grammar correcting AI tools such as Grammarly to check their grammar. While some may argue that they increase the productivity and efficiency of writing scholar essays, on the contrary, it concerns some that this may cause a decline in students’ abilities to proofread their work, affecting their communications skills, as the AI tool will do that for them. The overreliance may affect how medical students perform in a healthcare environment once they graduate. Due to their over-relying on grammar corrections tools, there is a chance of miscommunication happening between different healthcare sectors, or between a doctor and their patient. This could significantly decline the quality of healthcare provided to patients.
Without a doubt, developing suitable AI systems to aid students will cost great funds. Different medical schools have different teaching styles tailored to their students. The Medical School of Manchester uses a problem-solving style, whereas the Medical School of Bristol uses a case-study teaching style. While some believe that different medical schools should tailor their own AI teaching system to aid students, some argue those funds could go to a better place. For instance, building more student labs or renewing and rebuilding teaching funding, which is a more urgent issue than developing AI tools for students.
Artificial intelligence (AI) relies on large amounts of data. Machine learning algorithms learn to find inter-dependencies and patterns amongst data sets and apply those learnings to any new data they are presented with. It tends to follow a rule of the higher the volume and quality of the data (being uniform, diverse, comprehensive and relevant), the more accurate the algorithms. However, there’s a concern that there will be inaccurate data due to incomplete datasets, which can lead to biased AI predictions, especially in medicine, where there are always new findings. Medical students who use an inaccurate AI dataset may be given false, biased results, which may significantly affect their grade, or teach them inaccurate knowledge. Hence, universities must make sure that AI datasets are correct and unbiased before letting students use them.
Overall, the different applications of Artificial Intelligence within medical schools bring a range of benefits to both teachers and students, enhancing their education and ability to treat patients, however, they may hold students back from developing essential skills, as well as having many ethics issues that universities have to consider. Ultimately, there are multiple ways that AI can benefit medical training, including its accessibility, capacity to aid teachers, vast knowledge, ability to widen the range of treatment possibilities, and potential to improve VR simulations. However, the ethics issues that AI brings need an immediate solution. The Department of Education from the government should set out strict rules and regulations alongside different universities to monitor students’ usage of ai on their essays. It is recommended by different professionals in the education sector that a mature AI checking system should be developed before students are allowed to use AI for aiding their studies. Therefore, universities must balance the pros and cons of AI being implemented in medical schools before spending funds on developing specialised AI systems. Let’s see what the future holds.
References
Benefits – Aashi Lalit
- Rao, P. (2023). How Long it Took for Popular Apps to Reach 100 Million Users. [online] Visual Capitalist. Available at: https://www.visualcapitalist.com/threads-100-million-users/
- Banerjee, M., Chiew, D., Patel, K.T., Johns, I., Chappell, D., Linton, N., Cole, G.D., Francis, D.P., Szram, J., Ross, J. and Zaman, S. (2021). The impact of artificial intelligence on clinical education: perceptions of postgraduate trainee doctors in London (UK) and recommendations for trainers. BMC Medical Education, 21(1). doi:https://doi.org/10.1186/s12909-021-02870-x
- Paranjape, K., Schinkel, M., Nannan Panday, R., Car, J. and Nanayakkara, P. (2019). Introducing Artificial Intelligence Training in Medical Education (Preprint). JMIR Medical Education, 5(2). doi:https://doi.org/10.2196/16048
- news.stanford.edu. (n.d.). Using AI to create better virtual reality experiences. [online] Available at: https://news.stanford.edu/stories/2021/11/using-ai-create-better-virtual-reality-experiences
- Stanford HAI. (n.d.). ChatGPT Out-scores Medical Students on Complex Clinical Care Exam Questions. [online] Available at: https://hai.stanford.edu/news/chatgpt-out-scores-medical-students-complex-clinical-care-exam-questions
- Pottle, J. (2019). Virtual reality and the transformation of medical education. Future Healthcare Journal, [online] 6(3), pp.181–185. doi:https://doi.org/10.7861/fhj.2019-0036
- OECD (2015). How Much Time Do Teachers Spend on Teaching and Non-teaching Activities? [online] OECD iLibrary. Available at: https://www.oecd-ilibrary.org/education/how-much-time-do-teachers-spend-on-teaching-and-non-teaching-activities_5js64kndz1f3-en#:~:text=On%20average%20across%20countries%2C%20teachers%20spend%20half%20of
- www.tes.com. (n.d.). AQA to test how AI can ‘mark the marker’ in GCSE exams | Tes. [online] Available at: https://www.tes.com/magazine/news/general/gcses-2024-exam-board-trial-ai-summer-exams
Drawbacks – Caldia Choi
- LG (n.d.). What data is used to train an AI, where does it come from, and who owns it? [online] Potter Clarkson. Available at: https://www.potterclarkson.com/insights/what-data-is-used-to-train-an-ai-where-does-it-come-from-and-who-owns-it/#:~:text=Artificial%20intelligence%20(AI)%20relies%20on
- Lee, C. (2023). What is the potential of AI writing? Is cheating its greatest purpose? [online] www.turnitin.com. Available at: https://www.turnitin.com/blog/what-is-the-potential-of-ai-writing-is-cheating-its-greatest-purpose
GOV.UK. (n.d.). Malpractice in GCSE, AS and A level: summer 2022 exam series. [online] Available at: https://www.gov.uk/government/statistics/malpractice-in-gcse-as-and-a-level-summer-2022-exam-series/malpractice-in-gcse-as-and-a-level-summer-2022-exam-series