ChatGPT Can ‘Reduce Vaccine Hesitancy’ — by Parroting Public Health Narratives

(The Defender)—ChatGPT, an artificial intelligence (AI) chatbot, can be used to reduce “vaccine hesitancy” among the general public and also provide advice on sexually transmitted infections (STIs), according to a new study.

The research, which will be presented at the end of April at the Conference of the European Society of Clinical Microbiology and Infectious Diseases, found that ChatGPT’s responses to questions about vaccination were similar to those given by professional medical organizations and official guidelines.

According to a press release about the research, “vaccine hesitancy, directly linked to misinformation — false, inaccurate information promoted as factual — is on the rise.”

People who mistrust public health institutions may turn to ChatGPT for answers, as the popularity of the AI tool grows, it said. And if they do, the research shows they will get the responses the public health agencies would provide anyway.

Given those findings, the researchers hope ChatGPT can be “an effective tool to help reduce vaccine hesitancy among the general public.”

They also found similar results for advice about STIs. For the study, researchers from the National University Health System in Singapore tested how ChatGPT would respond to common questions about vaccination.

The researchers, who run infectious disease clinics, asked ChatGPT “15 commonly asked questions on vaccine hesitancy” including questions about efficacy, adverse effects and “cultural concerns.”

They also asked 17 questions about risk factors, access to care and prophylaxis for STIs. Two independent infectious disease experts compared ChatGPT’s answers about vaccines to recommendations from the Centers for Disease Control and Prevention (CDC) Advisory Committee on Immunization Practices (ACIP).

They assessed ChatGPT’s answers on STIs against the 2021 CDC STI Treatment Guidelines. According to the press release, ChatGPT provided “factual responses and reassurance to vaccine hesitancy concerns” — using the public health agencies’ statements as the basis for determining fact.

For example, the chatbot highlighted recommendations for measles vaccination in low-income settings and discussed the benefits of HPV vaccination, but failed to discuss age limits for vaccination. ChatGPT also “addressed misconceptions around mRNA vaccination and permanent alterations in DNA with high accuracy.”

“Overall, ChatGPT’s responses to vaccine hesitancy were accurate and may help individuals who have vaccine-related misconceptions,” said lead author Dr. Matthew Koh, an infectious diseases physician at National University Health System Singapore.

“Our results demonstrate the potential power of AI models to assist in public health campaigns and aid health professionals in reducing vaccine hesitancy.”

Biases and errors in AI

ChatGPT, a type of generative AI software that creates text and content, is increasingly gaining traction in science and healthcare. Research has found that in some cases it is beginning to displace typical search engines or Wikipedia as a go-to source for information.

As AI begins to be used in medicine, many researchers and AI designers have celebrated its potential as an administrative tool for writing research and generating patient information. Politico said 2024 may be the year “artificial intelligence transforms medicine.” Forbes and BMC Medical Education both claimed AI is “revolutionizing healthcare.”

However, generative AI tools have proven extremely unreliable in diagnosing health conditions. A study published in January in JAMA Pediatrics found ChatGPT misdiagnosed 83% of children’s health conditions.

The researchers in that study explained that chatbots “are typically non-specifically trained on a massive amount of internet data, which can often be inaccurate.” They “do not discriminate between reliable and unreliable information but simply regurgitate text from the training data to generate a response.”

This is significant, according to The New York Times, because people tend to adopt ideas they are repeatedly exposed to. The Times reported last month on a new preprint study that found AI language models like ChatGPT had clear political biases based on its “moral judgments, the way they frame their answers, which information they choose to share or omit and which questions they will or won’t answer.”

The Times also noted that the responses given by language AI models like ChatGPT are determined by the inputs that train them — the data they draw from and especially the “fine-tuning” done by system designers. The “misinformation” that the researchers in the study from the National University Health System Singapore are concerned about is the same misinformation the U.S. Food and Drug Administration (FDA) deems problematic.

Since FDA Commissioner Robert Califf began his second tenure as the agency’s head in February 2022, he made combating “misinformation” one of his top priorities, arguing it is “a leading cause of preventable death in America now” — though “this cannot be proved,” he said.

The FDA uses a tactic known as “prebunking,” by which the agency defines something as “misinformation” before readers encounter it elsewhere as possibly true. The agency can do this because Google “prioritizes credible websites” like the FDA’s in its searches.

Earlier studies tout ChatGPT’s ability to fight ‘COVID-19 conspiracy beliefs’

At least two previous studies also tested ChatGPT’s ability to “combat misinformation,” although both focused only on the COVID-19 vaccine. Research published last year in the journal Cureus reported ChatGPT could successfully be used to combat “COVID-19 conspiracy beliefs” and that it responded with “clear, concise and non-biased content.”

ChatGPT’s success, it said, could be seen in responses where the software was “dismissive” about “conspiratorial ideas” regarding COVID-19 origins, labeling it as non-credible and lacking scientific evidence — meaning the software dismissed the “lab-leak theory” and instead repeated the World Health Organization’s talking point that the virus had a “natural animal origin.”

No evidence has been found for the zoonotic origins of COVID-19 with a bat as the article claims. Substantive evidence from internal government documents released through Freedom of Information Act requests has suggested a lab leak is the more likely of the COVID-19 origin theories.

The software responses in that study also confirmed that the vaccines had undergone rigorous safety testing, there was no evidence that pharmaceutical companies obscured any data about adverse effects, and “the benefits of vaccination far outweigh the risks.”

However, those responses contradict what is known about how the vaccines were rushed through the regulatory process and how vaccine makers overstated the efficacy and understated risks of the vaccines.

A second study published in September 2023 in Human Vaccines & Immunotherapeutics made similar findings. The researchers asked the software the 50 questions most posed to World Health Organization collaborators about the COVID-19 vaccines and had public health professionals evaluate the responses.

It concluded that ChatGPT offers information consistent with existing scientific evidence, effectively debunking many myths circulating on social media, which could potentially promote a higher vaccine uptake. After those studies, mainstream scientific publications celebrated ChatGPT’s ability to debunk vaccine myths and increase vaccine uptake.

The Defender chats with ChatGPT

The Defender also “chatted” with ChatGPT to test its answers on some of the claims that public health officials often identify as vaccine “misinformation.”

The Defender: What are the origins of COVID-19?

Chat GPT: The origins of COVID-19, the disease caused by the novel coronavirus SARS-CoV-2, are believed to be zoonotic, meaning the virus likely originated in animals before spreading to humans. The exact origins are still under investigation, but it’s widely thought that the virus originated in bats and might have been transmitted to humans through an intermediate animal host, possibly in a wet market in Wuhan, China, where live animals were sold for food.

The Defender: Have pharmaceutical companies misrepresented the safety and efficacy of the COVID-19 vaccines?

ChatGPT: There’s no evidence to suggest that pharmaceutical companies have intentionally misrepresented the safety and efficacy of COVID-19 vaccines. The development and testing of vaccines are highly regulated processes, overseen by regulatory agencies such as the U.S. Food and Drug Administration (FDA), the European Medicines Agency (EMA), and others worldwide.

… extensive clinical trials were conducted to evaluate the safety and efficacy of COVID-19 vaccines …

However, it’s essential to recognize that no vaccine is 100% effective or without risks. Like all medical interventions, COVID-19 vaccines have potential side effects, but these are typically mild and temporary. 

The Defender: Do masks work to reduce the spread of COVID-19?

ChatGPT: Yes, masks are an essential tool in reducing the spread of COVID-19 …

The Defender: According to Cochrane systematic review on masks, they are not effective at reducing the spread of respiratory viruses.

ChatGPT: … In July 2020, Cochrane published a systematic review on the effectiveness of physical interventions, including masks, in reducing the transmission of respiratory viruses such as COVID-19. This review analyzed data from studies conducted before the COVID-19 pandemic and found limited evidence to support the effectiveness of wearing masks in preventing the transmission of respiratory viruses in the community.

However, it’s important to note that the findings of this review were based on evidence primarily from studies conducted prior to the COVID-19 pandemic and did not specifically focus on the effectiveness of masks in preventing the spread of COVID-19. Since the review’s publication, numerous studies have provided evidence supporting the effectiveness of masks, especially in reducing the transmission of COVID-19.

The Defender: How many people have suffered serious adverse events from the COVID-19 shots?

ChatGPT: As of my last update, COVID-19 vaccines have been administered to billions of people worldwide, and adverse events following vaccination have been reported … Serious adverse events following COVID-19 vaccination are rare. Most adverse events reported are mild and temporary, such as soreness at the injection site, fatigue, headache, or mild fever. Serious adverse events, while rare, can include severe allergic reactions (anaphylaxis) and other uncommon side effects …

Overall, the benefits of COVID-19 vaccination in preventing severe illness, hospitalization, and death from COVID-19 far outweigh the risks of potential adverse events. Vaccination remains a crucial tool in controlling the spread of the virus and protecting public health, particularly in the face of ongoing variants and the persistence of the pandemic …

The Defender: Is the Human Papillomavirus (HPV) vaccine linked to autoimmune disorders?

ChatGPT: The HPV vaccine has been extensively studied for its safety and efficacy, and there is currently no conclusive evidence linking the HPV vaccine to autoimmune disorders. Numerous large-scale studies and systematic reviews have found no significant association between HPV vaccination and autoimmune disorders …

While isolated reports of adverse events following HPV vaccination, including autoimmune disorders, may occur, extensive monitoring systems continue to evaluate vaccine safety. These systems, such as the Vaccine Adverse Event Reporting System (VAERS) in the United States and similar systems in other countries, investigate and monitor reports of adverse events to identify potential safety concerns.

Overall, the benefits of HPV vaccination in preventing HPV-related cancers, including cervical cancer, far outweigh the potential risks of adverse events.

ChatGPT didn’t mention that extensive research linking the Gardasil HPV vaccine to serious side effects, including postural orthostatic tachycardia syndrome (POTS) and premature ovarian failure and other autoimmune issues.

There are nearly 170 lawsuits in federal court from people who suffered a range of serious injuries after taking Gardasil, including autoimmune disorders, premature ovarian failure and cancer. There are also over 200 Gardasil injury claims pending in the ”vaccine court.”

The Defender also experimented with several questions with premises that clearly diverged from the regulatory agencies’ official positions on vaccination, such as “Can increased cancer rates be linked to the COVID-19 vaccines? Is there a link between vaccines and some autism? Has there been DNA contamination caused by mRNA vaccines?”

To those questions, ChatGPT’s response always included this statement, or something similar, at the end:

“Misinformation and conspiracy theories about vaccines can circulate, but it’s crucial to rely on credible sources of information, such as public health agencies and reputable scientific organizations, when evaluating the safety and efficacy of COVID-19 vaccines.”