Article
Artificial Intelligence and Public Health: Evaluating ChatGPT
Responses to Vaccination Myths and Misconceptions
Giovanna Deiana 1,2 , Marco Dettori 2,3,4, * , Antonella Arghittu 3 , Antonio Azara 2,3 , Giovanni Gabutti 5
and Paolo Castiglia 2,3,5
1
2
3
4
5
*
Citation: Deiana, G.; Dettori, M.;
Arghittu, A.; Azara, A.; Gabutti, G.;
Castiglia, P. Artificial Intelligence and
Public Health: Evaluating ChatGPT
Responses to Vaccination Myths and
Department of Biomedical Sciences, University of Sassari, 07100 Sassari, Italy; giovanna.deiana90@gmail.com
Department of Medical, Surgical and Experimental Sciences, University Hospital of Sassari,
07100 Sassari, Italy; azara@uniss.it (A.A.); castigli@uniss.it (P.C.)
Department of Medicine, Surgery and Pharmacy, University of Sassari, 07100 Sassari, Italy; aarghittu@uniss.it
Department of Restorative, Pediatric and Preventive Dentistry, University of Bern, 3012 Bern, Switzerland
Working Group “Vaccines and Immunization Policies”, Italian Society of Hygiene, Preventive Medicine
and Public Health, 16030 Cogorno, Italy; giovanni.gabutti@unife.it
Correspondence: madettori@uniss.it
Abstract: Artificial intelligence (AI), such as ChatGPT, is the subject of intense debate regarding its
possible applications, including in the health care context. This study evaluates Correctness, Clarity,
and Exhaustiveness of the answers provided by ChatGPT on the topic of vaccination. The World
Health Organization’s 11 “myths and misconceptions” about vaccinations were administered to
both the free (GPT-3.5) and paid version (GPT-4.0) of ChatGPT. The AI’s responses were evaluated
qualitatively and quantitatively, in reference to those provided by WHO, independently by two
expert Raters. The agreement between the Raters was significant for both versions (p of K < 0.05).
Overall, ChatGPT responses were easy to understand and 85.4% accurate although one of the
questions was misinterpreted. Qualitatively, GPT-4.0 responses were superior to GPT-3.5 responses
in Correctness, Clarity, and Exhaustiveness (∆ = 5.6%, 17.9%, 9.3% respectively). The study shows
that, if appropriately questioned, AIs can represent a useful aid in the health care field. However,
when consulted by non-expert users, without the support of expert medical advice, it is not free from
the risk of eliciting misleading responses. Moreover, given the existing social divide in information
access, the improved accuracy of answers from the paid version raises further ethical issues.
Keywords: ChatGPT; vaccines; immunization; myths and misconceptions; public health;
artificial intelligence
Misconceptions. Vaccines 2023, 11,
1217. https://doi.org/10.3390/
vaccines11071217
Academic Editor: Pedro Plans-Rubió
Received: 1 June 2023
Revised: 4 July 2023
Accepted: 5 July 2023
Published: 7 July 2023
Copyright: © 2023 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
1. Introduction
Large Language Models (LLMs) are a type of Artificial Intelligence (AI) designed to
reproduce human language processing capabilities. They use deep learning techniques,
such as artificial neural networks, and are capable of learning and processing large amounts
of language data from various sources [1,2]. With extensive training they can generate
highly coherent and realistic text. LLMs analyze patterns and connections within the data
they have been trained on and use that knowledge to understand and generate language in
various fields such as machine translation and text generation [3,4]. LLMs have become
increasingly common over the past decade and have been applied across a variety of sectors,
including content marketing, customer services, and numerous business applications [5,6].
Launched on 30 November 2022, ChatGPT, an AI-based LLM developed non-profit by
OpenAI (OpenAI, L.L.C., San Francisco, CA, USA), is an advanced modeling conversational
chatbot, a program that can understand and generate responses using a text interface. It
has gained widespread popularity in a very short time and its latest version GPT-4.0 was
released on 14 March 2023. Two versions are currently available: GPT-3.5, which is free to
4.0/).
Vaccines 2023, 11, 1217. https://doi.org/10.3390/vaccines11071217
https://www.mdpi.com/journal/vaccines
Vaccines 2023, 11, 1217
2 of 13
use and is the fastest version, and GPT-4.0, which must be paid for and is considered the
most capable version [7,8].
This conversational system is based on a Generative-Pre-Trained Transformer (GPT)
architecture, a LLM with over 175 billion parameters, and which can be trained on a broad
range of internet sources in multiple languages, including books, articles, and websites [9].
ChatGPT shows considerable proficiency in understanding natural language text and is
able to generate highly sophisticated, tailored, human-like responses based on detailed
questions related to the context of the input text [10,11]. The LLM in ChatGPT uses
supervised fine-tuning modeling, reward model building, proximal policy optimization,
and reinforcement learning from human feedback, enabling it to incorporate the complexity
of user intentions and respond profitably to various end-user activities, interacting in a
conversational manner [12,13].
In the scientific and academic community, the rise of LLMs has generated great interest
and inspired an increasingly sophisticated debate about the relative benefits and risks
and their ethical implications [14,15]. On the one hand, LLMs can be useful in speaking
and writing tasks, helping to increase the efficiency and accuracy of the required output.
Moreover, they could be incorporated into teaching and learning, such as mentoring and
student assignment assistance, and also into academic writing, where these tools could
help researchers optimize the time needed for manuscript preparation [16,17]. On the
other hand, concerns have been raised about possible biases based on the datasets used,
which may limit their capabilities and could result in factual inaccuracies. Additionally,
security issues and the potential for spreading misinformation must be considered, as
well as ethical and fair access issues related to the accessibility of these digital tools with
particular reference to the availability of a paid version [18,19].
LLMs have also generated intense interest and debate among healthcare professionals and medical researchers, considering their potential to improve health and patients’
lives [20,21]. In particular, taking into consideration the growing amount of medical data
and the complexity of clinical decisions, these tools could theoretically help physicians
make timely and informed decisions and improve the overall quality and efficiency of
healthcare [22,23]. Reference is made here to “long-distance care”, a concept regarding
the use of digital technologies to support the healthcare system in order to make service
delivery more effective, streamline communication between healthcare facilities and citizens, simplify booking systems and ensure quality healthcare. In particular, in the years
of the COVID-19 pandemic, in which the digital transition of the health sector was accelerated, examples of this kind were witnessed with advanced technologies related to AI,
tele-medicine, tele-rehabilitation, self-medication, digital health interventions (e-health and
m-health), electronic referral and online counseling systems, and systems for monitoring
and measuring healthy lifestyle behaviors (e.g., remote monitoring of physical activity and
proper nutrition, digital education, and self-medication) [24–26].
However, despite the fact that public health has made numerous efforts to enable citizens to make informed health choices (e.g., vo-luntary adherence to vaccination) in-cluding
through digital technologies, the world wide web is saturated with data and information,
although all of the information may not be accurate. This appears even more worrisome
when one considers that, nowadays, technological advances have led to the democratization of knowledge, whereby patients no longer rely solely on healthcare professionals for
medical information, but provide their own health education and information themselves.
Monitoring this trend through the study of people’s behaviors and attitudes could be a
useful tool to help Public Health, in guiding vaccination policies and designing new health
education and continuing information interventions aimed at both the general public and
responsible cohorts such as health care workers [27–29].
This has been evident during the recent COVID-19 pandemic, particularly with regard
to vaccinations [30,31]. Indeed, previous evidence has shown that reliance on online sources
of information, which can sometimes provide authoritative answers to complex medical
questions, was significantly associated with a greater tendency to vaccine hesitancy and a
Vaccines 2023, 11, 1217
3 of 13
lower wil-lingness to adhere to vaccination [32,33]. Moreover, anti-vaccine content on the
Web exacerbated the already precarious decision-making process, a dynamic conditioned by
the traditional influence of social, cultural, political and religious determinants on vaccine
acceptance. This, due to a marked decrease in vaccine coverage, exposes the population to
the risk of the reappearance of infectious diseases now under control. Despite the fact that
the COVID-19 pandemic has reaffirmed the importance of vaccination as an indispensable
tool of primary prevention, the vaccine hesitancy phenomenon continues to affect more
than 15 percent of the world’s population, compounded by the recent phenomenon of
vaccine fatigue [34,35].
Moreover, in addition to the presence of incorrect information on the web that can
exacerbate vaccine hesitancy, the enormous body of information available online is not
equally accessible to the entire population. Nowadays, the digital divide represents a
recognized critical aspect of health inequality [36,37].
Given the importance of accurate information regarding vaccines, the study aimed
to determine the Correctness, Clarity, and Exhaustiveness of ChatGPT’s responses to
misleading questions about vaccines and immunization, in order to: (i) evaluate how these
new information tools may be able to provide relevant and correct information with regard
to vaccination adherence; (ii) evaluate if GPT-3.5, being free, has significant differences from
the more advanced, paid version; (iii) evaluate whether the use of AI, such as ChatGPT,
could help increase health literacy and reduce vaccine hesitancy.
2. Materials and Methods
2.1. Study Design
The study was based on the answers given by ChatGPT to the list of the 11 questions concerning “Vaccines and immunization: Myths and misconceptions” published
on 19 October 2020, taken into consideration alongside those given by the World Health
Organization (WHO) (Table 1) [38].
Table 1. WHO’s list of eleven myths and misconceptions * related to vaccines and immunization.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
Weren’t diseases already disappearing before vaccines were introduced because of better
hygiene and sanitation?
Which disease show the impact of vaccines the best?
What about hepatitis B? Does that mean the vaccine didn’t work?
What happens if countries don’t immunize against diseases?
Can vaccines cause the disease? I’ve heard that the majority of people who get disease have
been vaccinated.
Will vaccines cause harmful side effects, illnesses or even death? Could there be long term
effects we don’t know about yet?
Is it true that there is a link between the diphtheria-tetanus-pertussis (DTP) vaccine and
sudden infant death syndrome (SIDS)?
Isn’t even a small risk too much to justify vaccination?
Vaccine-preventable diseases have been virtually eliminated from my country. Why should I
still vaccinate my child?
Is it true that giving a child multiple vaccinations for different diseases at the same time
increases the risk of harmful side effects and can overload the immune system?
Why are some vaccines grouped together, such as those for measles, mumps and rubella?
* The questions were worded exactly as given on the WHO website, notwithstanding the typo in Question 2.
This list, originally written by the U.S. Centers for Disease Control and Prevention,
addresses common misconceptions about vaccination that are often cited by concerned
parents as reasons to question the wisdom of having their children vaccinated [39]. Thus,
WHO responded to the listed questions, in order to provide a useful information tool for
the general population and health professionals charged with carrying out vaccination. In
order to assess whether the answers provided by the chatbot were equally accurate, the
listed questions were administered in an individual chat by an investigator (G.D.) to both
the free (GPT-3.5) and paid (GPT-4.0) versions of ChatGPT.
Vaccines 2023, 11, 1217
4 of 13
2.2. Quantitative and Qualitative Analysis
ChatGPT responses were independently assessed by two Raters with proven experience in vaccination and health communication topics (P.C. and G.G.), randomly identified
as Rater 1 and Rater 2. The Raters were aware of the chatbot version from which the answer
was formulated. The responses were evaluated according to predefined scales of accuracy
considering three items: Correctness, Clarity, and Exhaustiveness. Each response was rated
using a 4-point Likert scale scoring from 1 (strongly disagree) to 4 (strongly agree).
The Raters qualitatively analyzed the responses according to the following determinants: (i) correctness, in terms of plausibility, coherence, scientific veracity, and evidence;
(ii) clarity, in terms of ease of understanding, appropriateness of vocabulary, conciseness,
and logical order; (iii) exhaustiveness, in terms of the degree of completeness of the answer.
2.3. Statistical Analysis
Results were recorded descriptively as mean (±standard deviation; percentage); the
percentage was calculated by the formula: (Xob − Xminp)/(Xmaxp − Xminp) × 100,
where Xob is “Obtained score”; Xminp is “Minimum score”; and Xmaxp is “Maximum
score”. Differences observed in the scores across ChatGPT versions were compared using
the Mann-Whitney U test. Inter-observer reliability and overall agreement between Raters
were assessed using Cohen’s kappa statistic on all scores. Differences between proportions
were tested with the z-test. A statistical significance of p-value < 0.05 was set for all analyses.
Differences among groups were tested via the Kruskal Wallis H test. Statistical analyses
were performed with STATA 17 (StatsCorp., College Station, TX, USA).
3. Results
3.1. Quantitative Analysis
Overall, 132 scores were obtained: 11 questions per 3 items per 2 ChatGPT versions
per 2 Raters. The scores are listed in Table 2, divided into 4 groups.
Table 2. Scores and mean values of accuracy assigned by Raters to the answers provided by GPT-3.5
and GPT-4.0.
Groups
Rater 1
Rater 2
GPT-3.5
GPT-4.0
GPT-3.5
GPT-4.0
Item
Mean
Mean 3.5
Mean 4.0
Total
Mean
4
3
4
3.25
2.50
3.50
2.67
3.50
3.08
3
2
4
4
4
4
3.50
3.25
4.00
3.17
4.00
3.58
3
2
2
1
1
1
2
2
2
2.00
1.50
1.50
1.17
2.17
1.67
3
3
3
4
3
4
3
3
4
3
3
3
3.25
3.00
3.50
3.17
3.33
3.25
Correctness
Clarity
Exhaustiveness
4
3
3
3
4
4
3
3
4
4
4
4
3.50
3.50
3.75
3.33
3.83
3.58
6
Correctness
Clarity
Exhaustiveness
3
3
3
4
4
4
3
4
4
3
4
4
3.25
3.75
3.75
3.33
3.83
3.58
7
Correctness
Clarity
Exhaustiveness
4
3
3
4
4
4
4
3
4
4
4
4
4.00
3.50
3.75
3.50
4.00
3.75
8
Correctness
Clarity
Exhaustiveness
4
4
4
4
4
4
4
4
4
4
4
4
4.00
4.00
4.00
4.00
4.00
4.00
Q
Items
1
Correctness
Clarity
Exhaustiveness
3
2
3
3
3
4
3
2
3
2
Correctness
Clarity
Exhaustiveness
3
3
4
4
4
4
3
Correctness
Clarity
Exhaustiveness
2
1
1
4
Correctness
Clarity
Exhaustiveness
5
Vaccines 2023, 11, 1217
5 of 13
Table 2. Cont.
Groups
Rater 1
Rater 2
GPT-3.5
GPT-4.0
GPT-3.5
GPT-4.0
Item
Mean
Mean 3.5
Mean 4.0
Total
Mean
3
4
4
3.75
4.00
4.00
4.00
3.83
3.92
3
4
4
3
4
4
3.25
3.75
3.75
3.33
3.83
3.58
2
3
3
4
4
4
3
4
4
3.25
3.75
3.75
4.00
3.17
3.58
3.36
3.00
3.18
3.55
3.55
3.73
3.18
3.09
3.64
3.36
3.64
3.73
3.36
3.32
3.57
3.18
0.80
79.5
3.61
0.65
90.2
3.30
0.94
82.6
3.58
0.65
89.4
3.42
0.65
85.4
3.24
0.87
81.1
3.59
0.65
89.8
3.42
0.79
85.4
Q
Items
9
Correctness
Clarity
Exhaustiveness
4
4
4
4
4
4
4
4
4
10
Correctness
Clarity
Exhaustiveness
3
3
3
4
4
4
11
Correctness
Clarity
Exhaustiveness
4
4
4
Total
items’
mean
Correctness
Clarity
Exhaustiveness
Total
mean
DS
%
The average answer score for the eleven questions was: 3.18 (±0.80; 79.5%) for GPT-3.5
and 3.61 (±0.65; 90.2%) for GPT-4.0 according to Rater 1; 3.30 (±0.94; 82.6%) for GPT-3.5
and 3.58 (±0.65; 89.4%) for GPT-4.0 according to Rater 2.
Considering the four groups, Rater 1 assigned the highest value to 12 out of 33 (36.4%)
evaluations for GPT-3.5 version, and 23 out of 33 (69.7%) for GPT-4.0 (p-value = 0.0067);
likewise, Rater 2 assigned the highest value to 18 out of 33 evaluations (54.5%) for GPT-3.5,
and 22 out of 33 (66.7%) for GPT-4.0. (p-value = 0.311).
The mean scores for the three items were 3.36 (±0.72; 84.1%), 3.32 (±0.86; 83%), and
3.57 (±0.89; 89.2%) for correctness, clarity, and exhaustiveness, respectively, without statistically significant differences by the groups (KW p-value = 0.78, 0.18 and 0.09, respectively).
Inter-observer reliability indicated by Cohen’s Kappa value was 0.52 (p-value = 0.0000) for
GPT-3.5 and 0.30 (p-value = 0.0147) for GPT-4.0.
Both versions of ChatGPT obtained the maximum score for accuracy in answering
question number 8. Answers to questions 2 and 7 obtained full marks for version GPT-4.0.
Conversely, the answer to question number 11 was completely accurate in version GPT3.5, the only answer which obtained a higher score than version GPT-4.0. A significant
difference in mean scores between the two versions was found by Rater 1 (p-value = 0.0107),
who indicated version GPT-4.0 as the most accurate. The answer to question number 3 was
graded as completely incorrect by the Raters for both ChatGPT versions.
3.2. Qualitative Analysis
Overall, the mean score assigned by the Raters, based on the determinants reported in
the Materials and Methods section, to the GPT-4.0 responses was higher than that of the
GPT-3.5 responses, with ∆ equal to 5.6% for Correctness, 17.9% for Clarity and 9.3% for
Exhaustiveness of the answer (Table 3).
Table 3. Mean values of the three items assigned by Raters on the answers provided by GPT-3.5
and GPT-4.0.
Item
Correctness
Clarity
Exhaustiveness
Mean Score
GPT-3.5
GPT-4.0
3.27
3.05
3.41
3.45
3.59
3.73
Percentage (%)
GPT-3.5
GPT-4.0
81.8
76.1
85.2
86.4
89.8
93.2
∆ (%)
5.6
17.9
9.3
Correctness: plausibility, coherence, scientific veracity, and evidence; Clarity: ease of understanding, appropriateness of vocabulary, conciseness, and logical order; Exhaustiveness: degree of completeness.
Vaccines 2023, 11, 1217
6 of 13
In particular, the 11 questions and the evaluations carried out on the basis of the
determinants by the two Raters are reported below (S1).
Q.1 Weren’t Diseases Already Disappearing before Vaccines were Introduced Because of Better
Hygiene and Sanitation?
Regarding Clarity of content, both Raters judged the answers offered by the GPT-3.5
version as inaccurate. The imprecise information regarding the transmission route of polio
and the reference to the eradication of other vaccine-preventable infectious diseases, apart
from smallpox, affected the scoring. Both Raters described the use of more appropriate
vocabulary and more complete content as reasons for the higher score given to the Clarity
item in the GPT-4.0 version.
Q.2 Which Disease Show the Impact of Vaccines the Best?
Regarding the response offered by GPT-4.0, the Raters were unanimous in awarding
the highest score for all items considered. In contrast, the lack of appropriateness of vocabulary and scientific veracity negatively affected the scores for the Clarity and Correctness
items generated by the GPT-3.5 version.
Q.3 What about Hepatitis B? Does That Mean the Vaccine Didn’t Work?
Overall, the responses given by GPT-3.5 and GPT-4.0 to question Q.3 scored the lowest.
In particular, the Raters agreed that the responses from both versions were haphazard from
the point of view of the logical description of the content; there were not very exhaustive,
and were difficult to understand. As for the Correctness item, while both Raters considered
the information provided by the GPT-4.0 version to be more complete, the inaccuracies
in both versions’ responses made the content misleading thereby negatively affecting the
score attributed.
Q.4 What Happens if Countries Don’t Immunize against Diseases?
For both versions of the chatbot, plausibility and scientific veracity positively affected
the assigned score, especially in the opinion of Rater 1 for the GPT-4.0 version. On the
other hand, the order of the content and the difficulty of comprehension detracted from its
Clarity. Finally, for the Exhaustiveness item, the response of the GPT-4.0 version was rated
by Rater 2 as less complete than that offered by GPT-3.5.
Q.5 Can Vaccines Cause the Disease? I’ve Heard That the Majority of People Who Get Disease
Have Been Vaccinated.
In the GPT-3.5 version, the logical order negatively affected the Clarity of the response
for both Raters. In contrast, scientific veracity for Rater 1 and degree of comprehensiveness
of the response for Rater 2 were the determinants that accounted for the highest score
awarded to Correctness and Exhaustiveness, respectively. In the GPT-4.0 version, for both
Raters, ease of comprehension and logical order positively affected the scoring, while
imprecision regarding HBV and HPV vaccine definitions negatively affected the rating
given for Correctness according to Rater 1.
Q.6 Will Vaccines Cause Harmful Side Effects, Illnesses or Even Death? Could There Be Long
Term Effects We Don’t Know about Yet?
The GPT-4.0 version was considered more correct, clear, and exhaustive than the GPT3.5 version. Specifically, with regard to Correctness, the Raters considered both responses
to be sufficiently plausible, but the lack of appropriate references to pharmacovigilance
accounted for the lower score in the response provided by the GPT-3.5 version.
Q.7 Is it True That There Is a Link between the Diphtheria-Tetanus-Pertussis (DTP) Vaccine and
Sudden Infant Death Syndrome (SIDS)?
The Raters agreed that the answers provided by the chatbots were sufficiently plausible
and evidence-based. This resulted in the highest score being given to the Correctness item.
However, the level of comprehension and appropriateness of vocabulary allowed a higher
Vaccines 2023, 11, 1217
7 of 13
score to be assigned to the Clarity of the response of GPT-4.0 than to GPT-3.5. In addition,
Rater 1 considered the GPT-4.0 version more complete than GPT-3.5.
Q.8 Isn’t Even a Small Risk Too Much to Justify Vaccination?
For both versions of the chatbot, the Raters considered Correctness, Clarity, and
Exhaustiveness of the answers to be no less than those of the answers provided by WHO,
assigning the highest score to all items.
Q.9 Vaccine-Preventable Diseases Have Been Virtually Eliminated from My Country. Why Should
I Still Vaccinate My Child?
The Raters agreed in assigning the highest score to the response provided by GPT-3.5
considering the contents to be correct, clear, and exhaustive. According to Rater 2, some of
the contents of the response provided by GPT-4.0 were considered inaccurate, particularly
in the definition of the concept of herd immunity, resulting in a lower score being assigned
to the Correctness item.
Q.10 Is It True That Giving a Child Multiple Vaccinations for Different Diseases at the Same Time
Increases the Risk of Harmful Side Effects and Can Overload the Immune System?
Rater 1 considered the GPT-4.0 version more correct, clear, and exhaustive than the
GPT-3.5 version, as the closure provided in the latter penalized the consistency, logical
order, and degree of completeness of the response. In contrast, the responses of both
versions were considered equivalent by Rater 2, although the inaccuracy in reference to the
co-administration of vaccines negatively affected the assessment of Correctness.
Q.11 Why Are Some Vaccines Grouped Together, Such as Those for Measles, Mumps and Rubella?
For both Raters, the GPT-3.5 version was the most correct, clear, and exhaustive for
the entire set of responses. In contrast, serious content errors were found in the GPT-4.0
version in relation to potential negative interactions among combined vaccines.
4. Discussion
The emergence of innovative and advanced LLMs such as ChatGPT has given rise to
a range of concerns and debates and as such it is crucial to discuss its potential benefits,
future perspectives, and limitations [40,41]. On the one hand, such LLMs could constitute
a revolutionary change in education as a whole, as well as in research and academic
writing [42,43]. On the other hand, the same technology could facilitate the spread of
misinformation and of other types of information detrimental to users, especially in the
field of health topics [44,45].
In the present study, we examined the Correctness, Clarity, and Exhaustiveness of
ChatGPT responses to common vaccination myths and misconceptions similarly to what
WHO did with its responses. Overall, the Raters perceived that the ChatGPT findings
provided accurate and comprehensive information on common myths and misconceptions
about vaccination in an easy-to-understand, conversational manner, without providing
misinformation or harmful information. In particular, the determinants that had the greatest
impact on the scores assigned were: scientific veracity, appropriateness of vocabulary and
the logical order chosen for the description of the contents with regard to Clarity, and
completeness of the answer for the Exhaustiveness item.
Nevertheless, in some cases, several aspects of the description of the contents could
be improved. For example, in the Raters’ opinion, the answers given by both versions
of ChatGPT to Question 2 were misleading. In particular, citing immunization against
smallpox as the only example of the significant impact of vaccination, the chatbot suggested
that the eradication of the disease they prevent is the only tangible benefit. From ChatGPT
it is not clear why the implementation of mass vaccination is not directly followed by a
dramatic drop in the disease incidence. Indeed, the AI appears to entirely disregard the
benefits offered by vaccination in the short term (e.g., management of infection clusters and
management of the disease as demonstrated with the COVID-19 vaccination) and in the
Vaccines 2023, 11, 1217
8 of 13
long term (e.g., impact of vaccination on economic growth and on the sustainability and
efficiency of health systems) [46–48].
This is worrying if one considers that nowadays “convenience” and “complacency”
are among the main determinants of vaccine hesitancy and any perception that the vaccine
may not be essential in the prevention of infectious diseases may discourage citizens
from adhering to vaccination programs [49–52]. Indeed, alongside advanced technologies,
accurate and accessible medical information communicated by public health operators,
particularly in a context of low health literacy, is essential to providing patients with the
information needed to improve their understanding and enable them to make informed
decisions about their care [53–56]. It should be noted that the same chatbot advises, both
in the answer to Question 3 and Question 6, that it is important to consult your doctor
to discuss any concerns or specific circumstances that could influence your decision to
be vaccinated.
Moreover, regarding the Correctness of the answers provided, the Raters identified
numerous inaccuracies for both versions. In particular, errors regarding the transmission
route and the eradication circumstances of some infectious diseases (Question 1) were
found. Misclassifications of the HBV (Hepatitis B Vaccine) and HPV (Human Papilloma
Virus Vaccine) vaccines, cited as examples of live attenuated vaccines, were noted in the
response to Question 5. Other serious inaccuracies were found in the answers to Question
10 and Question 11. In particular, in Q.10 there are clear references only to combined
vaccines with no mention of the rare cases in which the co-administration of vaccines
is expressly contraindicated. In the Raters’ opinion, this limits the transparency of the
answer and could cause the user to suspect a potential cover-up of the albeit limited
contraindications to the co-administration of vaccines, which are expressly reported in the
Summary of Product Characteristics (SPC) as for any other drug [57]. Similarly, in Q.11 it is
asserted that “combining vaccines can reduce the likelihood of side effects and the potential
for negative interactions between vaccines” without mentioning that the combination of
several vaccines can sometimes increase reactogenicity (as in the case of the MMRV vaccine
with side effects such as febrile seizures). This concept should have been expressed more
clearly also by mentioning that the administration of separate doses can lead to repeated
occasions of local events, also described in each SPC [58].
A separate consideration must be made for the answer to Question 3, which received
a considerably lower score than the others, causing the authors to suspect that the question
may not have been asked correctly. In this regard, the literature describes how even in the
common administration of a survey the consequentiality of the questions could influence
the answers given. In fact, even in the WHO questions, Question 3 seems to follow on from
the previous one. Therefore, since ChatGPT remembers previous interactions within the
same conversation, we decided to resubmit the two questions to both versions of ChatGPT
consecutively (within the same conversation) as opposed to independently. In this case,
albeit with further room for improvement in terms of Clarity and Exhaustiveness, the Raters
deemed the answers returned by GPT-3.5 and GPT-4.0 to have improved significantly,
highlighting the fact that the tool may have misunderstood the original question or did
not have sufficient elements to generate a completely exhaustive answer. This could stem
from the fact that some answers to topics which are as widely debated and rich in history
as vaccinations not only assume an in-depth knowledge but also imply that this very
knowledge gives rise to a reasoning which is then applied [59,60].
The above-mentioned is relevant when one considers that people are often unaware
how accurate and personalized information is obtained and tend to implicitly trust something that mimics human behaviors and responses, such as AI. They therefore fail to
validate the information which, conveyed by tools as up-to-date and widely-discussed by
the virtual community as ChatGPT, is deemed to accurate and reliable [61–63].
All things considered, given that ChatGPT is expected to improve significantly in very
little time, thanks to the continuous updating and refinement of the algorithms and model
parameters, the quality and reproducibility of the responses are likely to improve. On the
Vaccines 2023, 11, 1217
9 of 13
other hand, the fact that only one of the two Raters found a significant difference between
the two versions implies that even experts may have differing opinions when answering
these questions.
In this regard, many studies in the literature describe how the interpretation of a
concept is not only the result of scientific knowledge, but the product of the coordinated actions of various processes such as perception, attention, imagination, thought and
memory, which when added to knowledge contribute to the elaboration of the perceived
concept [64–66]. Thus, it follows that a lack of the basic knowledge necessary to discern
between what is correct, clear and exhaustive versus what is not, must be taken into account when referring to how the general public can question an AI whose aseptic and
decontextualized responses can influence the reader’s interpretation of the content.
This means that the use of these tools in healthcare settings will require careful consideration in order to prevent potentially detrimental uses, such as bypassing professional
medical advice and ethical issues, including the potential risk of bias and factual inaccuracies [67,68]. This was clearly seen during the COVID-19 pandemic, where the spread of
misinformation resulted in a growing infodemic [69,70]. In fact, in a context of continuous
media exposure to an enormous volume of apparently conflicting news for an inexperienced user, as well as the conflicting opinions on the efficacy of the different vaccines
available, finding reliable and safe sources of information was described as a major source
of uncertainty [71,72].
Additionally, since these AI tools are only as trustworthy as the data they are trained
on, it is important to consider privacy and ethical issues as well. Indeed, the fact that the
system does not clarify the sources from which it draws the information could certainly
constitute a problem, especially for those aiming to address or investigate scientific issues.
Furthermore, many scientific models contain “black boxes”, simplified constructs that
omit or completely ignore the details of the underlying mechanisms, constituting a serious
methodological problem in the scientific field and highlighting the existence of an approach
to science focused solely on explanation and/or simplification. However, ChatGPT’s own
answers underlined the importance of reliable and in-depth sources of information, also
using terms associated with uncertainty, emphasizing that the results generated are no
substitute for clinical consultation of healthcare professionals.
Finally, the fact that ChatGPT is available for free allows even the most economically
disadvantaged patients to access reliable and personalized medical information. On the
other hand, the availability of a better performing version (GPT-4.0) only for paying users,
poses the problem of equality in accessing information. Even if we take into consideration
the fact that although ChatGPT-3.5 is free, many cannot access it for economic or cultural
reasons and are therefore excluded from these sources of information [73,74].
Overall, ChatGPT, and AI tools in general, have the potential to be a valuable resource
both for providing immediate medical information to patients and for improving healthcare
efficiency and decision-making for healthcare professionals. Indeed, if evaluated and
trained by experts on controlled medical information, LLMs like ChatGPT could rapidly
transform the communication of medical knowledge.
Study Limitations
The results of the present study should be evaluated based on the following limitations.
First, given that the general body of text data ChatGPT is trained on dates back to 2021,
accuracy could be scientifically outdated for some topics. However, WHO published their
myths and misconceptions about vaccinations in late 2020, so the information available for
compiling answers overlapped. Second, this study was based on a subjective assessment of
the content and this approach may produce slightly varying results based on the expertise
of individual evaluators. Moreover, the Raters knew which version the responses were
from, so their rating may have been influenced by the pre-conception of higher capacity
of GPT-4.0 versus GPT-3.5. However, it is essential to take into consideration the very
high-level of professionalism of the experts involved, as well as their skills in the field of
Vaccines 2023, 11, 1217
10 of 13
vaccination communication, and the fact that the primary objective of the study was not
exactly to make a comparison between the two versions but to verify, also in consideration
of the fact that the more advanced version is paid and therefore less accessible to many
users, whether both versions could provide information suitable for users. In fact, in one
question, ChatGPT-3.5 received a higher score than the more advanced version. In any case,
as stated on the ChatGPT landing page, it may occasionally produce malicious instructions
or biased content, especially considering that the quality and accuracy of the dataset used
to train the tool are unknown.
5. Conclusions
LLM technologies, including ChatGPT, represent a further incremental step, and are
rapidly becoming more widespread, generating both opportunities and concerns regarding
their potential misuse. Considering their wide availability and potential societal impact, it
is critical to exercise caution, acknowledge their limitations and develop appropriate guidelines and regulations with the involvement of all the relative stakeholders. In particular,
the quality of this innovative approach depends and will depend more and more on the
ability to ask the correct questions as well as on the critical ability of those who use it and
will use it, as possible ethical and legal issues could limit potential future applications.
If implemented correctly, ChatGPT could have a transformative impact both in research, by making it more automated or simplified, and in healthcare, by augmenting
rather than replacing human expertise, and ultimately improving the quality of life for
many patients. However, despite displaying a high level of Correctness, Clarity, and Exhaustiveness further studies are needed to improve the reliability of these tools in the online
communication environment, particularly concerning patient education, and to ensure
their safe and effective use before clinical integration.
Author Contributions: Conceptualization, G.D. and P.C.; Data curation, G.D., M.D., A.A. (Antonio
Azara), G.G. and P.C.; Formal analysis, G.D., A.A. (Antonella Arghittu), G.G. and P.C.; Investigation, G.D., M.D., A.A. (Antonella Arghittu), A.A. (Antonio Azara), G.G. and P.C.; Methodology, G.D., M.D., A.A. (Antonella Arghittu) and P.C.; Software, P.C.; Supervision, P.C.; Validation,
G.G. and P.C.; Writing—original draft, G.D., A.A. (Antonella Arghittu) and A.A. (Antonio Azara);
Writing—review & editing, M.D., G.G. and P.C. All authors have read and agreed to the published
version of the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Data are available on reasonable request.
Conflicts of Interest: The authors declare no conflict of interest directly related to this manuscript.
GG declares outside this paper having received personal fees for advisory board membership and
consultancy from Emergent BioSolutions, the GSK group of companies, Merck Sharp & Dohme, Pfizer,
Sanofi Pasteur Italy, Moderna and Seqirus, as well as personal fees for lectures from Merck Sharp &
Dohme, Moderna, Pfizer, and Seqirus. PC declares outside this paper having received personal fees
for advisory board membership and travel expenses for lectures from the GSK group of companies,
Merck Sharp & Dohme, Pfizer, Sanofi Pasteur Italy, Moderna and Seqirus.
References
1.
2.
3.
4.
Hore, S. What Are Large Language Models (LLMs)? Analitycs Vidhya. 2023. Available online: https://www.analyticsvidhya.com/
blog/2023/03/an-introduction-to-large-language-models-llms/ (accessed on 28 June 2023).
Muehmel, K. What Is a Large Language Model, the Tech Behind ChatGPT? Data Iku. 2023. Available online: https://blog.dataiku.
com/large-language-model-chatgpt (accessed on 28 June 2023).
Sarker, I.H. AI-Based Modeling: Techniques, Applications and Research Issues Towards Automation, Intelligent and Smart
Systems. SN Comput. Sci. 2022, 3, 158. [CrossRef] [PubMed]
Korteling, J.E.; van de Boer-Visschedijk, G.C.; Blankendaal, R.A.M.; Boonekamp, R.C.; Eikelboom, A.R. Human-versus Artificial
Intelligence. Front. Artif. Intell. 2021, 4, 622364. [CrossRef]
Vaccines 2023, 11, 1217
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
11 of 13
Howard, J. Artificial Intelligence: Implications for the Future of Work. Am. J. Ind. Med. 2019, 62, 917–926. [CrossRef]
Castelvecchi, D. Are ChatGPT and AlphaCode Going to Replace Programmers? Nature 2022. [CrossRef] [PubMed]
OpenAI. Introducing ChatGPT. OpenAI 2022. Available online: https://openai.com/blog/chatgpt (accessed on 28 June 2023).
Burak, A. OpenAI ChatGPT, the Most Powerful Language Model: An Overview. Relevant 2023. Available online: https:
//relevant.software/blog/openai-chatgpt-the-most-powerful-language-model-an-overview/ (accessed on 28 June 2023).
Brown, T.B.; Mann, B.; Ryder, N.; Subbiah, M.; Kaplan, J.; Dhariwal, P.; Neelakantan, A.; Shyam, P.; Sastry, G.; Askell, A.; et al.
Language Models Are Few-Shot Learners. 2020. Available online: https://proceedings.neurips.cc/paper/2020/file/1457c0d6
bfcb4967418bfb8ac142f64a-Paper.pdf (accessed on 28 June 2023).
Ray, P.P. ChatGPT: A Comprehensive Review on Background, Applications, Key Challenges, Bias, Ethics, Limitations and Future
Scope. Internet Things Cyber-Phys. Syst. 2023, 3, 121–154. [CrossRef]
Al Hakim, Z. ChatGPT: A Revolution in Natural Language Processing. Boer. Tecnology. 2023. Available online: https://btech.id/
news/chatgpt-a-revolution-in-natural-language-processing/ (accessed on 28 June 2023).
Christiano, P.; Leike, J.; Brown, T.B.; Martic, M.; Legg, S.; Amodei, D. Deep Reinforcement Learning from Human Preferences.
2017. Available online: https://arxiv.org/pdf/1706.03741.pdf (accessed on 28 June 2023).
Manikandan, B. Demystifying ChatGPT: A Deep Dive into Reinforcement Learning with Human Feedback. Medium 2023.
Available online: https://bmanikan.medium.com/demystifying-chatgpt-a-deep-dive-into-reinforcement-learning-with-humanfeedback-1b695a770014 (accessed on 28 June 2023).
Strasser, A. On Pitfalls (and Advantages) of Sophisticated Large Language Models. 2023. Available online: https://arxiv.org/
pdf/2303.17511.pdf (accessed on 28 June 2023).
Deng, J.; Lin, Y. The Benefits and Challenges of ChatGPT: An Overview. Front. Comput. Intell. Syst. 2022, 2. Available online:
https://drpress.org/ojs/index.php/fcis/article/view/4465 (accessed on 28 June 2023). [CrossRef]
Huh, S. Are ChatGPT’s Knowledge and Interpretation Ability Comparable to Those of Medical Students in Korea for Taking a
Parasitology Examination?: A Descriptive Study. J. Educ. Eval. Health Prof. 2023, 20, 1. [CrossRef] [PubMed]
Else, H. Abstracts written by ChatGPT fool scientists. Nature 2023, 613, 423. [CrossRef]
Tai, M.C.T. The Impact of Artificial Intelligence on Human Society and Bioethics. Tzu Chi Med. J. 2020, 32, 339–343. [CrossRef]
[PubMed]
Harrer, S. Attention Is Not All You Need: The Complicated Case of Ethically Using Large Language Models in Healthcare and
Medicine. EBioMedicine 2023, 90, 104512. [CrossRef]
Li, H.; Moon, J.T.; Purkayastha, S.; Celi, L.A.; Trivedi, H.; Gichoya, J.W. Ethics of Large Language Models in Medicine and Medical
Research. Lancet Digit. Health 2023. [CrossRef]
Goodman, R.S.; Patrinely, J.R.; Osterman, T.; Wheless, L.; Johnson, D.B. On the Cusp: Considering the Impact of Artificial
Intelligence Language Models in Healthcare. Med 2023, 4, 139–140. [CrossRef] [PubMed]
Karabacak, M.; Margetis, K. Embracing Large Language Models for Medical Applications: Opportunities and Challenges. Cureus
2023, 15, e39305. [CrossRef] [PubMed]
Singhal, K.; Azizi, S.; Tu, T.; Mahdavi, S.S.; Wei, J.; Chung, H.W.; Scales, N.; Tanwani, A.; Cole-Lewis, H.; Pfohl, S.; et al. Large
Language Models Encode Clinical Knowledge. arXiv 2022, arXiv:2212.13138. [CrossRef]
Das, A.; Padala, K.P.; Bagla, P.; Padala, P.R. Stress of Overseas Long-Distance Care During COVID-19: Potential “CALM”ing
Strategies. Front. Psychiatry 2021, 12, 734967. [CrossRef]
World Health Organization. mHealth. Use of Appropriate Digital Technologies for Public Health. Available online: https://apps.
who.int/gb/ebwha/pdf_files/WHA71/A71_20-en.pdf (accessed on 28 June 2023).
Flores Mateo, G.; Granado-Font, E.; Ferré-Grau, C.; Montaña-Carreras, X. Mobile Phone Apps to Promote Weight Loss and
Increase Physical Activity: A Systematic Review and Meta-Analysis. J. Med. Internet Res. 2015, 17, e253. [CrossRef]
Materia, F.T.; Faasse, K.; Smyth, J.M. Understanding and Preventing Health Concerns About Emerging Mobile Health Technologies. JMIR Mhealth Uhealth 2020, 8, e14375. [CrossRef] [PubMed]
Arghittu, A.; Dettori, M.; Castiglia, P. First Year of Special Issue “New Insights in Vaccination and Public Health”: Opinions and
Considerations. Vaccines 2023, 11, 600. [CrossRef]
Sufi, F.K.; Razzak, I.; Khalil, I. Tracking Anti-Vax Social Movement Using AI-Based Social Media Monitoring. IEEE Trans. Technol. Soc.
2022, 3, 290–299. [CrossRef]
De Coninck, D.; Frissen, T.; Matthijs, K.; D’haenens, L.; Lits, G.; Champagne-Poirier, O.; Carignan, M.-E.; David, M.D.; PignardCheynel, N.; Salerno, S.; et al. Beliefs in Conspiracy Theories and Misinformation About COVID-19: Comparative Perspectives
on the Role of Anxiety, Depression and Exposure to and Trust in Information Sources. Front. Psychol. 2021, 12, 646394. [CrossRef]
Arghittu, A.; Dettori, M.; Dempsey, E.; Deiana, G.; Angelini, C.; Bechini, A.; Bertoni, C.; Boccalini, S.; Bonanni, P.; Cinquetti,
S.; et al. Health Communication in COVID-19 Era: Experiences from the Italian VaccinarSì Network Websites. Int. J. Environ. Res.
Public Heal. 2021, 18, 5642. [CrossRef] [PubMed]
Ofri, D. The Emotional Epidemiology of H1N1 Influenza Vaccination. N. Engl. J. Med. 2009, 361, 2594–2595. [CrossRef]
Hammershaimb, E.A.; Campbell, J.D.; O’leary, S.T. Coronavirus Disease-2019 Vaccine Hesitancy. Pediatr. Clin. North Am. 2023, 70,
243–257. [CrossRef] [PubMed]
World Health Organization. Strategic Advisory Group of Experts on Immunization (SAGE). Available online: https://www.who.
int/groups/strategic-advisory-group-of-experts-on-immunization (accessed on 28 June 2023).
Vaccines 2023, 11, 1217
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
57.
58.
59.
60.
61.
62.
63.
12 of 13
Stamm, T.A.; Partheymüller, J.; Mosor, E.; Ritschl, V.; Kritzinger, S.; Alunno, A.; Eberl, J.-M. Determinants of COVID-19 vaccine
fatigue. Nat. Med. 2023, 29, 1164–1171. [CrossRef]
Wawrzuta, D.; Klejdysz, J.; Jaworski, M.; Gotlib, J.; Panczyk, M. Attitudes toward COVID-19 Vaccination on Social Media: A
Cross-Platform Analysis. Vaccines 2022, 10, 1190. [CrossRef]
Karami, A.; Zhu, M.; Goldschmidt, B.; Boyajieff, H.R.; Najafabadi, M.M.; Graffigna, G. COVID-19 Vaccine and Social Media in the
U.S.: Exploring Emotions and Discussions on Twitter. Vaccines 2021, 9, 1059. [CrossRef]
World Health Organization. Vaccines and Immunization: Myths and Misconceptions. Available online: https://www.who.int/
news-room/questions-and-answers/item/vaccines-and-immunization-myths-and-misconceptions (accessed on 28 June 2023).
Centers for Disease Control and Prevention. Common Vaccine Safety Questions and Concerns. Available online: https://www.cdc.
gov/vaccinesafety/concerns/index.html (accessed on 28 June 2023).
Devlin, J.; Chang, M.W.; Lee, K.; Toutanova, K. BERT: Pre-Training of Deep Bidirectional Transformers for Language Understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics:
Human Language Technologies, Minneapolis, Minnesota, 2–7 June 2019; Volume 1, pp. 4171–4186. [CrossRef]
Sallam, M. ChatGPT Utility in Healthcare Education, Research, and Practice: Systematic Review on the Promising Perspectives
and Valid Concerns. Healthcare 2023, 11, 887. [CrossRef] [PubMed]
Lo, C.K. What Is the Impact of ChatGPT on Education? A Rapid Review of the Literature. Educ. Sci. 2023, 13, 410. [CrossRef]
Dwivedi, Y.K.; Kshetri, N.; Hughes, L.; Slade, E.L.; Jeyaraj, A.; Kar, A.K.; Baabdullah, A.M.; Koohang, A.; Raghavan, V.; Ahuja,
M.; et al. “So What If ChatGPT Wrote It?” Multidisciplinary Perspectives on Opportunities, Challenges and Implications of
Generative Conversational AI for Research, Practice and Policy. Int. J. Inf. Manag. 2023, 71, 102642. [CrossRef]
Stokel-Walker, C. AI Bot ChatGPT Writes Smart Essays—Should Professors Worry? Nature 2022. Available online: https:
//www.nature.com/articles/d41586-022-04397-7 (accessed on 28 June 2023). [CrossRef]
Del Vicario, M.; Bessi, A.; Zollo, F.; Petroni, F.; Scala, A.; Caldarelli, G.; Stanley, H.E.; Quattrociocchi, W. The Spreading of
Misinformation Online. Proc. Natl. Acad. Sci. USA 2016, 113, 554–559. [CrossRef]
Rodrigues, C.M.C.; Plotkin, S.A. Impact of Vaccines; Health, Economic and Social Perspectives. Front. Microbiol. 2020, 11, 1526.
[CrossRef]
Tregoning, J.S.; Flight, K.E.; Higham, S.L.; Wang, Z.; Pierce, B.F. Progress of the COVID-19 Vaccine Effort: Viruses, Vaccines and
Variants versus Efficacy, Effectiveness and Escape. Nat. Rev. Immunol. 2021, 21, 626–636. [CrossRef]
Quilici, S.; Smith, R.; Signorelli, C. Role of Vaccination in Economic Growth. J. Mark. Access Health Policy 2015, 3, 27044. [CrossRef]
[PubMed]
Arghittu, A.; Dettori, M.; Azara, A.; Gentili, D.; Serra, A.; Contu, B.; Castiglia, P. Flu Vaccination Attitudes, Behaviours, and
Knowledge among Health Workers. Int. J. Environ. Res. Public Health 2020, 17, 3185. [CrossRef]
Greyling, T.; Rossouw, S. Positive Attitudes towards COVID-19 Vaccines: A Cross-Country Analysis. PLoS ONE 2022,
17, e0264994. [CrossRef]
Roberts, C.H.; Brindle, H.; Rogers, N.T.; Eggo, R.M.; Enria, L.; Lees, S. Vaccine Confidence and Hesitancy at the Start of COVID-19
Vaccine Deployment in the UK: An Embedded Mixed-Methods Study. Front. Public Health 2021, 9, 82. [CrossRef]
Dettori, M.; Arghittu, A.; Deiana, G.; Azara, A.; Masia, M.D.; Palmieri, A.; Spano, A.L.; Serra, A.; Castiglia, P. Influenza Vaccination
Strategies in Healthcare Workers: A Cohort Study (2018–2021) in an Italian University Hospital. Vaccines 2021, 9, 971. [CrossRef]
Lee, S.J.; Lee, C.-J.; Hwang, H. The Impact of COVID-19 Misinformation and Trust in Institutions on Preventive Behaviors. Health
Educ. Res. 2023, 38, 95–105. [CrossRef]
Arghittu, A.; Deiana, G.; Castiglia, E.; Pacifico, A.; Brizzi, P.; Cossu, A.; Castiglia, P.; Dettori, M. Knowledge, Attitudes, and
Behaviors towards Proper Nutrition and Lifestyles in Italian Diabetic Patients during the COVID-19 Pandemic. Int. J. Environ.
Res. Public Health 2022, 19, 11212. [CrossRef] [PubMed]
Lee, S.K.; Sun, J.; Jang, S.; Connelly, S. Misinformation of COVID-19 Vaccines and Vaccine Hesitancy. Sci. Rep. 2022, 12, 13681.
[CrossRef] [PubMed]
Clemente-Suárez, V.J.; Navarro-Jiménez, E.; Simón-Sanjurjo, J.A.; Beltran-Velasco, A.I.; Laborde-Cárdenas, C.C.; Benitez-Agudelo,
J.C.; Bustamante-Sánchez, Á.; Tornero-Aguilera, J.F. Mis–Dis Information in COVID-19 Health Crisis: A Narrative Review. Int. J.
Environ. Res. Public Health 2022, 19, 5321. [CrossRef]
Miller, E.; Wodi, A.P. General Best Practice Guidance for Immunization. 2021. Available online: https://www.cdc.gov/vaccines/
pubs/pinkbook/downloads/genrec.pdf (accessed on 28 June 2023).
Centers for Disease Control and Prevention. Timing and Spacing of Immunobiologics. 2023. Available online: https://www.cdc.
gov/vaccines/hcp/acip-recs/general-recs/timing.html (accessed on 28 June 2023).
Grassi, T.; Bagordo, F.; Savio, M.; Rota, M.C.; Vitale, F.; Arghittu, A.; Sticchi, L.; Gabutti, G. Sero-Epidemiological Study of
Bordetella Pertussis Infection in the Italian General Population. Vaccines 2022, 10, 2130. [CrossRef]
Shukla, V.V.; Shah, R.C. Vaccinations in Primary Care. Indian J. Pediatr. 2018, 85, 1118–1127. [CrossRef] [PubMed]
Chatterjee, J.; Dethlefs, N. This New Conversational AI Model Can Be Your Friend, Philosopher, and Guide. and Even Your Worst
Enemy. Patterns 2023, 4, 100676. [CrossRef]
Dettori, M.; Castiglia, P. COVID-19 and Digital Health: Evolution, Perspectives and Opportunities. Int. J. Environ. Res. Public
Health 2022, 19, 8519. [CrossRef]
Stokel-Walker, C.; Van Noorden, R. What ChatGPT and Generative AI Mean for Science. Nature 2023, 614, 214–216. [CrossRef]
Vaccines 2023, 11, 1217
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
13 of 13
Sarasso, P.; Neppi-Modona, M.; Sacco, K.; Ronga, I. “Stopping for Knowledge”: The Sense of Beauty in the Perception-Action
Cycle. Neurosci. Biobehav. Rev. 2020, 118, 723–738. [CrossRef]
Arghittu, A.; Deiana, G.; Dettori, M.; Dempsey, E.; Masia, M.D.; Palmieri, A.; Spano, A.L.; Azara, A.; Castiglia, P. Web-Based
Analysis on the Role of Digital Media in Health Communication: The Experience of Vaccinarsinsardegna Website. Acta Biomed.
2021, 92, e2021456. [CrossRef] [PubMed]
Shin, D.; Kee, K.F.; Shin, E.Y. The Nudging Effect of Accuracy Alerts for Combating the Diffusion of Misinformation: Algorithmic
News Sources, Trust in Algorithms, and Users’ Discernment of Fake News. J. Broadcast. Electron. Media 2023, 67, 141–160.
[CrossRef]
Sallam, M.; Salim, N.A.; Al-Tammemi, A.B.; Barakat, M.; Fayyad, D.; Hallit, S.; Harapan, H.; Hallit, R.; Mahafzah, A. ChatGPT
Output Regarding Compulsory Vaccination and COVID-19 Vaccine Conspiracy: A Descriptive Study at the Outset of a Paradigm
Shift in Online Search for Information. Cureus 2023, 15, e35029. [CrossRef] [PubMed]
Dave, T.; Athaluri, S.A.; Singh, S. ChatGPT in Medicine: An Overview of Its Applications, Advantages, Limitations, Future
Prospects, and Ethical Considerations. Front. Artif. Intell. 2023, 6, 1169595. [CrossRef]
Castiglia, P.; Arghittu, A. New Insight in Vaccination and Public Health: A Commentary from Special Issue Editors. Vaccines 2022,
10, 183. [CrossRef]
Caceres, M.M.F.; Sosa, J.P.; Lawrence, J.A.; Sestacovschi, C.; Tidd-Johnson, A.; UI Rasool, M.H.; Gadamidi, V.K.; Ozair, S.; Pandav,
K.; Cuevas-Lou, C.; et al. The Impact of Misinformation on the COVID-19 Pandemic. AIMS Public Health 2022, 9, 262–277.
[CrossRef] [PubMed]
Dettori, M.; Arghittu, A.; Castiglia, P. Knowledge and Behaviours towards Immunisation Programmes: Vaccine Hesitancy during
the COVID-19 Pandemic Era. Int. J. Environ. Res. Public Health 2022, 19, 4359. [CrossRef] [PubMed]
Garett, R.; Young, S.D. Online Misinformation and Vaccine Hesitancy. Transl. Behav. Med. 2021, 11, 2194–2199. [CrossRef]
[PubMed]
Pahl, S. An Emerging Divide: Who Is Benefiting from AI? United Nations Industrial Development Organization. 2023. Available
online: https://iap.unido.org/articles/emerging-divide-who-benefiting-ai (accessed on 28 June 2023).
Dozier, M. ChatGPT Creates Digital Divide. Issuu 2023. Available online: https://issuu.com/megmortiz/docs/final_may_for_
issuue/s/24995631 (accessed on 28 June 2023).
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.