AI in Healthcare

Potential Errors & Lack of Accountability

A huge concern with the use of AI in healthcare is potential errors although, of course, human error also exists. Perhaps a unique issue with AI error is the lack of accountability. As in, you can’t fire an AI system or hold them accountable for mistakes like you could a human being.

AI tools are definitely known to be wrong — AKA hallucinating — and patient harm may result. Artificial intelligence could recommend the wrong prescription, fail to recognize a growth in imaging, or base guidance on racially-biased data collection.

In my medical opinion, a big problem I see is that AI could be trained on some outdated data. If AI has learned from incorrect information, it could easily produce incorrect results, diagnoses, and treatment recommendations.

If these AI tools become widespread, potential AI system errors could result in thousands of injuries — rather than a more limited number of injuries caused by a single healthcare provider’s error.

Who is accountable if an AI-enabled recommendation leads to harm — the clinician, the hospital, or the developer? Generally, the clinician or the hospital is still held accountable in cases of AI recommendations that lead to negative outcomes.

Transparency

Artificial intelligence never stops collecting information, but people are usually unaware that their data is being collected for machine learning purposes. There is very little legally-required transparency in AI, and hopefully that changes in the near future.

Yes, we sign away our rights to privacy with all those terms and conditions that we sign sight unseen. Data collection guidelines are often buried in user agreements. But that doesn’t mean patients have no ethical right to know when AI is being used.

Transparency in how AI-assisted models are trained would increase trust by hospitals, companies, and organizations disclosing when collected data has been used to create a tool, application, or message automation.

Individuals who use AI tools to compose a business email, a personal essay on a med school application, or a patient communication, seldom disclose that AI technology is being used. Ethically speaking, disclosing when AI is used benefits everyone, especially in healthcare.

Loss of Human Connection

AI may bring about a loss of individuality or a loss of human connection that we’ve already started to witness. When people rely on AI to answer their questions instead of human physicians or coworkers, the human element is slowly lost.

When communications — for example, between doctor and patient, or administrator and nurse — are automated and AI-generated, a loss of humanity can worsen the recipient’s comprehension or even their health outcome. AI could make provider communications feel less human and individualized. AI may weaken trust between payers and healthcare professionals.

Medical school applicants who use AI to generate essays or applications may lose a sense of individuality crucial to standing out. Med students who use AI to write content may be perceived as robotic or lacking passion.

Finally, AI could risk physicians listening to AI over the patient, which would impact trust and increase the risk of severe health outcomes. Many doctors already struggle to listen to their patients holistically, but AI could worsen this problem.

Bias & Equity

If AI is trained on biased data, it will almost certainly spit out biased recommendations.

Even though they don’t mean to be, many doctors have implicit racial biases that seep into patient care. Unfortunately — again, even if they aren’t biased on purpose — doctors may change how they communicate with or treat a patient based on race.

Algorithms may perpetuate such systemic bias if they’re trained on non-representative datasets. This could increase disparities instead of reducing them.

Wealthier communities have more access to healthcare resources. Though AI could, in theory, increase access to healthcare in lower-income communities, we must be wary not to perpetuate today’s disparities — since AI is trained on data from the past and present.

Read Next: Top Medical Schools in the US (Especially AI-Forward Schools)

Considerations for Med School Applicants and Students

It’s tempting to let ChatGPT fill out your whole med school or residency application, isn’t it? I must recommend you avoid ChatGPT for actual writing of any essay. However, you may utilize AI and chatbots for brainstorming and editing suggestions.

Let’s talk about some of the biggest considerations for medical school applicants and students when it comes to AI.

Gain a Strong Understanding of AI Fundamentals

Start with the basics. Here’s a list of topics you can research to gain a fundamental understanding of artificial intelligence:

• Artificial Intelligence (AI): Literally speaking, AI is computer systems that perform tasks that typically require human intelligence (reasoning, perception, learning, decision-making). Look up the difference between generative AI and other AI.
• Machine Learning (ML): This is a subset of AI where algorithms learn from data to make predictions or decisions without being explicitly programmed.
• Deep Learning: This is a subset of ML using neural networks (especially convolutional and transformer-based architectures) to process complex inputs like images, sound, or text.
• Large Language Models (LLMs): These are tools like ChatGPT and Med-PaLM trained on text data for summarization, reasoning, or patient communication.
• How AI is used in medicine: Reading this article is a great start! Above, we’ve talked about how AI is used in diagnostics, treatment, prevention, telemedicine, clinical trials, and so much more.
• Ethics of AI: Look up reputable articles on the ethics of using AI in the healthcare field. Again, this article is a great start! I would emphasize the importance of human oversight and accountability when it comes to applying AI to medical practices.

Learn to Think Critically About the Ethics

It’s essential for aspiring and current physicians to consider and weigh potential ethical scenarios regarding AI in healthcare. It will help with applying to medical school, going through med school and residency, and working as a doctor.

Familiarize yourself with the four pillars of medical bioethics: autonomy, non-maleficence, beneficence, justice. Think critically about how these bioethical principles apply to the use of AI models.

For instance, AI should not be able to override a patient’s wishes for treatment, as doing so would violate the ethical principle of autonomy — a patient’s right to make their own choices about their care. 

Another example: AI may recommend surgery because it produces the best long-term outcomes in the literature, even though surgery might be unnecessarily invasive for the patient’s condition. This could violate the principle of non-maleficence — AKA “do no harm.”

Prepare for Potential Interview Questions

There will probably be interview questions regarding AI in healthcare and medical education that you need to be prepared to answer in an interview for med school or residency.

Here are a few example interview questions about AI in healthcare:

• How do you see AI shaping patient care in the next decade?
• What ethical considerations should guide AI use in medicine?
• When do you think it’s acceptable to use AI in med school?

FAQs