The magnitude of medical information and learning resources available online, while advantageous to the user, can also be overwhelming [1-4]. The sheer volume and varying quality can lead to information overload [5-7]. This can be a stressor, making it difficult for one to evaluate the relevance and quality of the information available, ultimately impacting the individual and inhibiting effective decision-making [8-10]. In addition, there are concerns about the credibility and validity of medical and health-related information found on widely used information-sharing platforms [11,12]. A systematic review by Wang et al [13] showed an increasing trend in health-related misinformation on topics such as vaccination, Ebola, and Zika virus. Coincidentally, the article was published a month before the COVID-19 pandemic started making headlines, which itself garnered attention from online “misinformation superspreaders” [14]. Furthermore, YouTube has been identified as a source of misinformation regarding the human papillomavirus vaccine [15]. Studies have also highlighted the spread of cancer misinformation on Facebook [16,17], misinformation related to e-cigarettes and nicotine on Twitter [18], and misinformation about rheumatoid arthritis across various platforms [19]. These issues are problematic because they can lead to public health risks, such as vaccine hesitancy [20], poor disease management [21], and the spread of harmful health practices [22].

In response to these challenges of information overload and the accuracy of information, recommender systems are emerging as a powerful tool [23-25]. Recommender systems, also known as recommendation systems, platforms, engines, or algorithms, are essential tools in information filtering and personalization [26,27]. Their use can improve user experiences, boost engagement, and aid decision-making [28]. A seminal article by Resnick and Varian [29] characterized a recommender system as one in which recommendations are provided as inputs by people, which the system then aggregates and directs to appropriate recipients. Burke [30] described them as systems that generate tailored recommendations or assist users in a personalized way to choose from an array of options. These systems use algorithms to tailor recommendations by filtering, curating content, and guiding the user toward pertinent resources [31,32]. Thus, recommender systems play a pivotal role in alleviating the burden of information overload.

Recommender systems have already seamlessly integrated into our everyday lives. Serving as intelligent guides, they suggest relevant content to users in contexts ranging from social media to e-commerce and streaming services. According to a report by MacKenzie et al [33], 35% of Amazon purchases and 75% of Netflix viewing choices can be attributed to recommendations by such algorithms. In recent years, approaches such as machine learning, natural language processing, and deep learning [34-37] have brought advancements to recommender systems. These systems are not only prevalent in commercial applications but are also increasingly being adopted in educational settings [38,39]. A recent systematic review by Thongchotchat et al [40] found that while attribute- or characteristic-type recommender systems dominate education, there is an emerging trend of varied systems, including hints, clues, prompts, and tutoring, but these are in the early stages and require exploration and development. Combining theories of learning styles may enhance the flexibility and effectiveness of recommendations [40]. A recent bibliometric analysis [39] noted an increase in research dedicated to exploring recommendation techniques with the objective of enhancing the accuracy of recommendations made by educational recommender systems, driven by the ever-increasing amount of available learning resources. Large language models (LLMs) now enhance conversational recommender systems by leveraging their contextual understanding, potentially improving recommendation quality [41,42]. Recommender systems fall under 1 of 4 types of filtering approaches [43-45]. Table 1 provides a detailed description of each type, including its respective strengths and limitations, as well as examples.

The integration of generative artificial intelligence and LLMs has advanced the development of recommender systems. These technologies enhance natural language processing, facilitate natural-sounding dialogues, and improve recommendation accuracy and user interaction experiences [66-68]. Approaches such as retrieval-augmented generation models combine retrieval models and generative models to integrate specific knowledge sources, aiding complex recommendations while also addressing issues such as hallucinations and cold starts [69,70]. These advances, however, are not just restricted to text-based systems. Multimodal recommender systems are emerging in popularity, integrating various types of input data such as text, image, and audio. By leveraging the capabilities of LLMs to process this multimodal data, more comprehensive and accurate recommendations are possible [71,72]. Despite these advancements, challenges remain. Biases in LLM-based recommendations [73] and ethical concerns continue to pose challenges [74]. As these systems continue to evolve, they will offer more efficient and personalized solutions.

Learners in health care seek reliable materials for their education online [75,76], while others turn to the internet to understand their health conditions, treatment options, and preventive measures [77-81]. YouTube, a popular educational resource among health profession students [82,83], raises concerns for health care educators regarding the quality of information presented to students. This concern is supported by findings from Helming et al [84], who found that many YouTube videos on medical education are of low quality, even those created by academic physicians. Metrics such as views and likes, while popular among learners for navigating their choices, can be superficial and unreliable indicators of quality or accuracy [11,83,85]. Similarly, a study by Camm et al [86] found no correlation between video quality as assessed by a scoring tool and commonly used preference metrics (ie, hits, likes, dislikes, or search page rankings). When it comes to the videos recommended by the platform, the rationale behind the algorithm for suggesting specific content remains unclear? By developing a proprietary recommender system curated by experts in the subject matter, the suggested information and learning resources could be moderated, ensuring the quality and relevance of the materials recommended to learners.

Educational recommender systems are information systems designed for use in educational settings to recommend various resources to different stakeholders, including learners, educators, researchers, and others [87]. Leveraging advanced algorithms and data analytics, educational recommender systems provide tailored recommendations that enhance the educational experience by addressing individual learner needs and preferences [1,38]. They facilitate efficient resource discovery, aiding personalized learning and informed decision-making in academic applications [38,40]. These systems are used in a variety of ways, with the goal of assisting students with academic decisions. They are used to recommend suitable institutes, study topics, and courses [88-90]. They can be used to propose personalized syllabi and scheduling [91-93], learning materials [94-97], scholarships [98], supervisors [99], and career paths [100]. These systems can also be used to offer learners feedback on summative assessments [101]. In education, personalized learning centers on learners, acknowledging their inherent differences—variations in strengths, weaknesses, and learning preferences [102]. This recognition of differences suggests that a uniform or “one-size-fits-all” approach may not be optimal for all learners. Koestner et al [103] conducted a meta-analysis that revealed that learners are more likely to achieve their goals when those goals are (1) self-concordant and (2) accompanied by an implementation plan. The concept of agency, as defined by Bandura [104], refers to the capacity to initiate actions with intention and purpose. In education, this concept manifests as learner agency—a framework that embodies self-directed learning [105,106]. Learners actively pursue knowledge aligned with personal goals and interests, facilitated by strategic guidance from educators [107,108]. Educators play a crucial role in this process by providing the necessary guidance and support to ensure that learners have access to high-quality and relevant resources. In addition to the information overload highlighted earlier, learners may sometimes be unaware of what to search for [109]. Deschênes [110] highlighted that recommender systems are valuable for curating educational resources tailored to the needs of individual learners, empowering informed choices, guiding learners to resources (that they may not have encountered unprompted), and fostering an environment conducive to autonomous yet guided education.

While the topic of recommender systems has been extensively studied in both health care [111,112] and education [113-115], to the best of the authors’ knowledge, there appears to be a gap in the literature regarding their application in health professions education (HPE). For example, a topic search (covering the title, abstract, and keywords) on Web of Science using the terms “recommender system*” and “health professions education” returned only 2 articles. This scoping review aims to comprehensively map the extent of recommender systems in the HPE domain.

In reviewing the literature for applications of such systems in HPE, several innovative implementations with diverse purposes have been identified. One system provides adaptive, competence-based recommendations tailored to each student’s unique needs [116]. Another is used to evaluate nursing competencies and to recommend targeted learning materials to address skill gaps [117]. Liou and Chen [118] presented an e-learning platform that recommended personalized learning activities for medical interns. Another system assists second-language nurses in patient care charting by suggesting optimal terms during documentation [119]. Liou [120] introduced a recommender system for an online discussion forum that personalized article recommendations based on individual preferences.

This scoping review aims to identify the different use cases of recommender systems in HPE. In addition, this review aims to determine the filtering techniques used by these systems, analyze their attributes or features, and explore the user experiences of both learners and educators. Furthermore, this review seeks to identify gaps in the current literature and provide suggestions for future research on the use of recommender systems in HPE.

This study is based on the analysis of published literature, both scientific and gray. It does not involve patients, medical research, or any type of personal information, and, as such, no ethics approval is required. The results of this scoping review will be submitted for publication in a peer-reviewed international journal and presented at scientific meetings and conferences related to HPE.

The anticipated findings of this scoping review are expected to provide valuable insights into the diverse applications and potential benefits of recommender systems in HPE. These systems can significantly enhance learning experiences by personalizing the content to the unique needs of individual learners. By providing tailored recommendations, these systems can help bridge knowledge gaps, reinforce learning, and support continuous professional development. The expected findings will provide insights into the experiences of both learners and educators, identifying applications and implementation challenges. The scoping review aims to evaluate existing evidence and identify gaps in the literature concerning the use of recommender systems in HPE.

Previous studies have highlighted the benefits of recommender systems in various educational contexts of HPE (116-20). A continuously growing number of learning resources has created a need for these systems in education [23-25,39]. Digel et al [38] reported that engaging users in the development process of such systems is crucial to ensure their acceptance and use. It is anticipated that the applications and implementation challenges identified through this scoping review will help health profession educators in developing future systems. This review will build on these findings by specifically focusing on their use with HPE and identifying unique challenges and opportunities in this field.

A strength of this review is its comprehensive methodology, which includes a thorough search across multiple scientific databases and gray literature sources. This approach ensures a broad and inclusive overview of the existing literature. However, some limitations should be acknowledged. First, because the screening processes for scoping review do not assess the quality of the included studies, the robustness of the evidence presented may be impacted [132]. In addition, by focusing exclusively on studies published from 2000 onward, the review may overlook earlier research that could offer valuable insights and increase the potential for systematic bias [133]. Furthermore, while the aim is to cover a broad range of studies related to recommender systems, this breadth may come at the cost of a lack of depth in the analysis [134], making it difficult to draw specific conclusions. Publication bias, in which the likelihood of a study being published is influenced by the nature and direction of its results, creates a systematic difference between published and unpublished studies. Although incorporating gray literature in our study helps mitigate this bias, it does not fully eliminate it, as studies with positive or significant outcomes are still more likely to be published [135].

The dissemination strategies for the findings of this scoping review will include publication in a peer-reviewed journal and presentations at national and international conferences. The dissemination activities will commence in the winter of 2025 with the manuscript submission.

While recommender systems have been extensively studied in health care and education, there is a significant gap in the literature regarding their application in HPE. To the best of our knowledge, this is the first comprehensive exploration of recommender systems specifically focused on HPE. This scoping review aims to thoroughly map the extent and various applications of these systems within this domain. By identifying effective practices and highlighting existing gaps, this review will serve as a valuable resource for health professions educators, enriching academic discourse and providing practical guidance for effective implementation in HPE.