Artificial intelligence and clinical anatomical education: Promises and perils

May combine two claims.

Existing AI cannot match the full capabilities of human clinicians or educators, but can carry out specific tasks at high speed, at large scale, and with remarkable consistency.

“As such, existing AI cannot match the full capabilities of human clinicians or educators, but can carry out specific tasks at high speed, at large scale, and with remarkable consistency.”

May combine two claims.

Historical approaches to anatomy education were disrupted during the Covid-19 pandemic, with remote teaching nearly universally shared by the anatomy discipline at some point between 2020 and 2022.

“Historical approaches to anatomy education were disrupted during the Covid-19 pandemic, with the onset of remote teaching which was nearly universally shared by the anatomy discipline at some point between 2020 and 2022 (Pather et al., 2020;Sun et al., 2020;Harmon et al., 2021).”

• cites Pather et al., 2020Not checked yet

  Not checked in the free run.

• cites Sun et al., 2020Not checked yet

  Not checked in the free run.

• cites Harmon et al., 2021Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim. May combine two claims.

The collective pandemic experience has led to some calls to further integrate AI into anatomy education as an ongoing standard teaching tool, with select groups of students and educators valuing these technological advances over traditional person-led teaching approaches like dissection.

“This collective experience has led to some calls to further integrate AI into anatomy education as an ongoing standard teaching tool, with select groups of students and educators valuing these technological advances over traditional person-led teaching approaches like dissection (Papa et al., 2022).”

• cites Papa et al., 2022Not checked yet

  Not checked in the free run.

May combine two claims.

AIEd offers opportunities for enhanced consistency and precision across anatomy education delivery and assessments, and capacity to detect issues and patterns that may otherwise go undetected by humans.

“This AIEd also offers opportunities for enhanced consistency and precision across anatomy education delivery and assessments, as well as the capacity to detect issues and patterns that may otherwise go undetected by humans.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

The ability for AI to be rapidly adapted and adopted into anatomy classrooms could make it more influential than other, less technological, anatomy teaching tools.

“Furthermore, the ability for AI to be rapidly adapted and adopted into anatomy classrooms could make it more influential than these other, less technological, anatomy teaching tools.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

Anatomy education is brimming with uncertainty.

“Many in the anatomy community acknowledge that anatomy education is brimming with uncertainty (Evans et al., 2018;Stephens et al., 2021).”

• cites Evans et al., 2018Not checked yet

  Not checked in the free run.

• cites Stephens et al., 2021Not checked yet

  Not checked in the free run.

May combine two claims.

Clinical anatomy educators are embracing uncertainties relating to individual variation and diversity of the human form, ambiguities of dissection, and the breadth of content required for effective healthcare practice.

“Clinical anatomy educators are embracing uncertainties relating to issues such as individual variation and diversity of the human form (Willan & Humpherson, 1999;Wheble & Channon, 2021;Cullinane & Barry, 2022), the ambiguities of dissection (e.g., differences between textbook anatomy and that found through dissection) (Stephens et al., 2021), and the sheer breadth of content required for effective”

• cites Willan & Humpherson, 1999Not checked yet

  Not checked in the free run.

• cites Wheble & Channon, 2021Not checked yet

  Not checked in the free run.

• cites Cullinane & Barry, 2022Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable). May combine two claims.

Incorporating AI into anatomy education may provide unique benefits but may also have unintended consequences.

“Incorporating AI into anatomy education may therefore provide unique benefits, but there may also be unintended consequences.”

May combine two claims.

AI excels in areas such as pattern recognition, object matching, and image identification.

“Artificial intelligence excels in areas such as pattern recognition, object matching, and image identification (Bundy, 2017).”

• cites Bundy, 2017Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Integrating opportunities for learners to practice and develop uncertainty tolerance in anatomical education is one way of preparing students for this healthcare competency.

“Integrating opportunities for learners to practice and develop their uncertainty tolerance in anatomical education is one way of preparing students for this healthcare competency (Domen, 2016;Papanagnou et al., 2021).”

• cites Domen, 2016Not checked yet

  Not checked in the free run.

• cites Papanagnou et al., 2021Not checked yet

  Not checked in the free run.

May combine two claims.

AI developers require precision and definitive categorization, and operate under the pragmatic expectation that AI technologies function in social settings in ways that are broadly quantifiable, calculable, and capable of operating effectively given correct inputs.

“Artificial intelligence developers require precision and definitive categorization, and operate under the pragmatic expectation that AI technologies function in social settings (such as the anatomy classroom) in ways that are broadly quantifiable, calculable, and capable of operating effectively given the correct inputs (Selwyn & Gallo Cordoba, 2021).”

• cites Selwyn & Gallo Cordoba, 2021Not checked yet

  Not checked in the free run.

May combine two claims.

The accuracy of AI tools depends on high-volume and high-quality data, therefore biases may be newly created or reinforced depending on the representativeness of the data these tools are trained upon.

“The accuracy of AI tools depends on high-volume and high-quality data, therefore, biases may be newly created or reinforced depending on the representativeness of the data these tools are trained upon.”

• AssumptionRests on an assumption: The accuracy of AI tools depends on high-volume and high-quality data.

May combine two claims.

Anatomical education has progressed over the past 20 years to better reflect the clinical relevance of human anatomical variation, and is now calling into question historical and contemporary anatomical representations that fail to consider differences based on race, skin tone, sex, gender, body type, and age.

“Anatomical education has progressed over the past 20 years to better reflect the clinical relevance of human anatomical variation (Willan & Humpherson, 1999;Wheble & Channon, 2021;Cullinane & Barry, 2022), and is now calling into question historical and contemporary anatomical representations that fail to consider differences based on race, skin tone, sex, gender, body type, and age (Liscum &”

• cites Willan & Humpherson, 1999Not checked yet

  Not checked in the free run.

• cites Wheble & Channon, 2021Not checked yet

  Not checked in the free run.

• cites Cullinane & Barry, 2022Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

There is increasing awareness of and attention to the potential for AI to reinforce, hide, or accelerate racism and other forms of discrimination across multiple spheres of life.

“There is increasing awareness of, and attention to, the potential for AI to reinforce, hide, or accelerate racism and other forms of discrimination across multiple spheres of life (Benjamin, 2019).”

• cites Benjamin, 2019Not checked yet

  Not checked in the free run.

May combine two claims.

Chatbots trained on a corpus of online interactions tend to be unreliable conveyors of information, which can result in undesirable or unintended lines of communication.

“Here, too, chatbots trained on a corpus of online interactions (e.g., from Reddit) tend to be unreliable conveyors of information, which can result in undesirable or unintended lines of communication.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

Integrating AI into anatomy education to support and evaluate learners is tempting given the large and sustained teaching load of anatomy educators globally.

“Integrating AI into anatomy education to support and evaluate learners is tempting given the large and sustained teaching load of anatomy educators globally, from the United States (Heylings, 2002;Drake et al., 2009;McBride & Drake, 2018), to Australia and New Zealand (Craig et al., 2010;Trautman et al., 2019) and China (Pan et al., 2020).”

• cites Heylings, 2002Not checked yet

  Not checked in the free run.

• cites Drake et al., 2009Not checked yet

  Not checked in the free run.

• cites McBride & Drake, 2018Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

There is a danger that overly prescriptive AI diagnostics may further hinder students' uncertainty tolerance, giving them a flattened sense of the ambiguities and uncertainties intrinsic in human anatomy.

“As such, there is a danger that overly prescriptive AI diagnostics may further hinder students' uncertainty tolerance, giving them a flattened sense of the ambiguities and uncertainties intrinsic in human anatomy.”

May combine two claims.

Existing AI models are trained on mainly European or East Asian populations, hence adoption of such models in AIEd has the potential to train students to poorly identify melanomas in dark-skinned people, reinforcing bias from lack of diversity in anatomy textbooks.

“According to Young and colleagues, existing AI models are trained on mainly European or East Asian populations, hence the adoption of such models in AIEd has the potential to train students to poorly identify melanomas in dark-skinned people, reinforcing the bias that result from the lack of diversity in anatomy textbooks (Young et al., 2020).”

• cites Young et al., 2020Not checked yet

  Not checked in the free run.

May combine two claims.

The uncritical integration of AI into anatomy teaching may influence students' perceptions and treatment of diverse patients when they become clinicians, which could lead to unequal healthcare provision.

“The broader concern here is that the uncritical integration of AI into anatomy teaching may influence students' perceptions and treatment of diverse patients when they become clinicians, which could lead to unequal healthcare provision.”

Classified as probably not a substantial claim.

Automation bias is the tendency for human agents to rely on automated cues to the point of failing to notice anomalies undetected by the automated system, or by ignoring contradictory information.

“This is supported by the well-documented concern with automation bias, which is the tendency for human agents to rely on automated cues to the point of failing to notice anomalies undetected by the automated system, or by ignoring contradictory information (Mosier & Skitka, 2018).”

• cites Mosier & Skitka, 2018Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable).

Expert human educators have a range of qualities that AIEd does not.

“Importantly, expert human educators have a range of qualities that AIEd does not.”

May combine two claims.

Automation bias is reported in healthcare, where use of clinical decision-support systems leads to clinicians failing to notice problems (unless alerted) and complying with incorrect recommendations.

“Automation bias is reported in a range of domains, including healthcare, where the use of clinical decision-support systems leads to clinicians failing to notice problems (unless alerted to them), as well as complying with incorrect recommendations (Lyell et al., 2017).”

• cites Lyell et al., 2017Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable).

Sometimes hidden within tasks such as responding to students' routine questions or identifying absent students is a prompt for more extended student-teacher interactions.

“Sometimes hidden within these tasks, such as responding to students' routine questions or identifying absent students, is a prompt for more extended student-teacher interactions.”

Probably not a substantial claim: may not pass S2 (checkable).

Integrating AIEd without considering these limitations may lead to a de-professionalization of the anatomical educator.

“Integrating AIEd without considering these limitations may also lead to a de-professionalization of the anatomical educator.”

Probably not a substantial claim: may not pass S2 (checkable).

Relying on AI for these aspects of anatomy education will require human deliberation to determine whether the data collected by the AI warrants further action.

“Thus, relying on AI for these aspects of anatomy education will require human deliberation to determine whether the data collected by the AI warrants further action.”

Probably not a substantial claim: may not pass S2 (checkable).

AI systems that claim to analyze student learning vastly reduce the act of studying (let alone learning) to a very limited set of variables.

“Artificial intelligence systems that claim to analyze student learning vastly reduce the act of studying (let alone learning) to a very limited set of variables.”

May combine two claims.

Human teachers can make social connections with students, improvise when needed, and use interpersonal skills to motivate, cajole, and enthuse.

“Human teachers can also make social connections with students, improvise when needed and use their interpersonal skills to motivate, cajole, and enthuse.”

Probably not a substantial claim: may not pass S2 (checkable).

Education requires acts of creativity, innovation, and spontaneity.

“Education requires acts of creativity, innovation, and spontaneity (Selwyn & Gallo Cordoba, 2021).”

• cites Selwyn & Gallo Cordoba, 2021Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Automated essay grading is based on text pattern matching, and so is not capable of discerning meaning and nuance in written expression, which disadvantages students who use nonconventional phrasing and vocabulary.

“Automated essay grading is based on text pattern matching, and so is not capable of discerning meaning and nuance in written expression which disadvantages students who use nonconventional phrasing and vocabulary.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

Recommending which piece of work to tackle next or assessing learners' written work requires holistic awareness of both the student and the disciplinary content being taught.

“For example, recommending which piece of work to tackle next or assessing learners' written work requires holistic awareness of both the student and the disciplinary content being taught.”

Probably not a substantial claim: may not pass S2 (checkable).

Anatomy educators are well-placed to support AI development that is fit-for-purpose.

“Anatomy educators are, thus, well-placed to support AI development that is fit-for-purpose, and are strongly encouraged to engage with this process.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

How AI is engaged in anatomical teaching can model how students will interact with AI as professionals.

“How AI is engaged in anatomical teaching can model how students will interact with AI as professionals.”

Probably not a substantial claim: may not pass S2 (checkable).

Integrating AI without also engaging in discourse about how it perpetuates these historical biases undermines the important work being done.

“Integrating AI without also engaging in discourse about how it perpetuates these historical biases undermines the important work being done.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

The AI community suffers from not seeing how its work fits into a long history of science being used to legitimize violence against marginalized people, and to stratify and separate people.

“the”

• cites Van Noorden, 2020Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable), S4 (based). May combine two claims.

AI tools have value in a number of areas of anatomical education, especially where they can replicate diagnostic and identification functions that AI plays in contemporary clinical practice, and where they genuinely free educators up from nonpedagogical tasks in ways that do not diminish social contact with students.

“Artificial intelligence tools have value in a number of areas of anatomical education-especially where they can replicate the diagnostic and identification functions that AI plays in contemporary clinical practice, and also where they genuinely free educators up from nonpedagogical tasks in ways that do not diminish social contact with students.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

Encouraging human educators to lead the teaching, with AIEd in a supportive role, will help generate an effective balance of the efficiencies of AI and the uncertainties of the very human clinical anatomy context.

“Encouraging human educators to”

Classified as probably not a substantial claim.

Effective education requires pedagogical content knowledge, which represents teaching proficiency and course content knowledge expertise.

“Effective education requires pedagogical content knowledge, which represents teaching proficiency and course content knowledge expertise (Shulman, 1987).”

• cites Shulman, 1987Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S3 (load-bearing).

Another recent publication describes an alternate example wherein AI curriculum was built around a series of lectures designed to help radiology residents gain insight and understanding of AI algorithms.

“Another recent publication describes an alternate example wherein AI curriculum was built around a series of lectures designed to help radiology residents gain insight and understanding of AI algorithms (Lindqwister et al., 2021).”

• cites Lindqwister et al., 2021Not checked yet

  Not checked in the free run.

Classified as probably not a substantial claim.

This study revealed participant's had an increased confidence in interpreting and working alongside AI in the workplace after partaking in these professional development opportunities.

“This study revealed participant's had an increased confidence in interpreting and working alongside AI in the workplace after partaking in these professional development opportunities.”

• cites Lindqwister et al., 2021Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

The opportunities provided by AI in terms of scaling, speed, consistency, and precision are the result of the same factors that challenge anatomy educators' capacity to highlight and communicate this anatomic uncertainty to their healthcare students.

“The opportunities provided by AI in terms of scaling, speed, consistency, and precision are the result of the same factors that challenge anatomy educators' capacity to highlight and communicate this anatomic”

May combine two claims.

This type of AI surveillance can impact negatively on the trust necessary for effective educational environments, and can also hinder transformative learning that typically occurs in un-assessed components of curriculum.

“This type of AI surveillance can impact negatively on the trust necessary for effective educational environments (Gubbins & MacCurtain, 2008), and can also hinder the type of transformative learning that typically occurs in un-assessed components of curriculum.”

• cites Gubbins & MacCurtain, 2008Not checked yet

  Not checked in the free run.

May combine two claims.

These programs provided opportunities to work with AI developers or solve problems using AI.

“These programs provided opportunities to work with AI developers or solve problems using AI.”

• cites Paranjape et al., 2019Not checked yet

  Not checked in the free run.

Probably not a substantial claim: may not pass S2 (checkable).

Anatomy education, through increased transparency and discussions about AI frailties, can help students challenge the presumption that the technology is always right, a necessary competency for an AI-laden healthcare future.

“Anatomy education, through increased transparency and discussions about AI frailties, can help students challenge the presumption that the technology is”

May combine two claims.

A recent report by Human Rights Watch reviewed 164 EdTech products deployed across 49 countries, finding that 89% of them had monitored or had the capacity to monitor students and harvest their personal data without consent, as well as tracking students across the internet and over time.

“For example, a recent report by Human Rights Watch (2022) reviewed 164 EdTech products deployed across 49 countries, finding that 89% of them had monitored or had the capacity to monitor students and harvest their personal data without consent, as well as tracking students across the internet and over time.”

• cites Human Rights Watch, 2022Not checked yet

  Not checked in the free run.