Intelligent Tutoring Systems in Open-ended, Multimodal Domains Reviewing the Evidence from Arts, Music, and Sports

Intelligent Tutoring Systems in Open-ended, Multimodal Domains Reviewing the Evidence from Arts, Music, and Sports | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Systematic Review Intelligent Tutoring Systems in Open-ended, Multimodal Domains Reviewing the Evidence from Arts, Music, and Sports Lucas Lörch, Garvin Brod This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8797891/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Intelligent tutoring systems (ITS) have been popular in well-formalized educational domains such as math and physics. The domains of arts, music, and sports, however, have been challenging for ITS due to their open-ended, multimodal nature. We conducted a systematic, narrative literature review to investigate how ITS in arts, music, and sports address these challenges. How do ITS in arts, music, and sports individualize instruction in the face of open-ended tasks? How do they use technology to handle multimodal data? We searched eight databases (Google Scholar, PsychInfo, PsychArticles, Education Research Complete, The New Republic Archive, ProQuest, ERIC, Psyndex) using a search string that combined the term “intelligent tutoring system” and its synonyms with the terms “arts,” “music,” and “sports”. We included 42 publications describing 29 ITS in our review. We found that all of the reviewed ITS addressed skills such as playing a musical instrument that were complex and multifaceted but rather well-formalized than open-ended. Most ITS were focused on providing feedback. However, some systems achieved individualized task selection by decomposing complex skills into component skills and mapping them to tasks. Regarding technology use, we identified two overarching issues: measuring or presenting body movements with hardware such as depth cameras, digital styluses, and robots; and optimizing feedback or task selection with machine learning algorithms. We consider the lack of ITS that address open-ended, creative skills to be a gap in the literature that should be addressed by future research and discuss the potential of artificial intelligence in tackling this challenge. Educational Psychology intelligent tutoring systems arts music sports machine learning large-language models artificial intelligence Figures Figure 1 Highlights • We reviewed intelligent tutoring systems in arts, music, and sports. • We found no systems that address open-ended, creative skills. • Most of the reviewed systems were focused on providing individualized feedback. • Some systems individualized task selection by mapping component skills to tasks. • Innovative hardware was used to measure or present body movements. Introduction Intelligent Tutoring Systems (ITS) are computer-assisted instruction systems that adapt their teaching strategies to individual students (Corbett et al., 1997). The effectiveness of ITS has been demonstrated in multiple meta-analyses (Ma et al., 2014; Steenbergen-Hu & Cooper, 2014; Tamim et al., 2011; VanLehn, 2011). ITS typically select tasks for individual learners based on their previous performance (Corbalan et al., 2006). Learners can decide the time and pace at which they practicing (Martin et al., 2007) and they receive immediate response-specific feedback (Azevedo & Bernard, 1995; Sosa et al., 2011). While such individualized task selection and feedback are cumbersome and time-intensive in human-led tutoring, they come with no extra costs in ITS. Given the widespread availability of computers in Western societies, ITS can thus play an important role in today's education. Implementing ITS is relatively straight-forward in well-formalized, text-based domains (Holland, 2000). Accordingly, ITS have been especially popular in domains such as algebra (Anderson et al., 1995; Koedinger et al., 1997; Ritter et al., 2007), physics (Albacete & VanLehn, 2000; VanLehn et al., 2002, 2007), and language (McNamara et al., 2004; Rowley et al., 1998; Tsiriga & Virvou, 2004). In rather open-ended, multimodal domains such as arts, music, and sports, the implementation of ITS is rather challenging (Phon-Amnuaisuk & Siong, 2008). However, the rise of artificial intelligence (AI) and other technologies, such as motion sensors and virtual reality, has opened new possibilities for ITS in these domains. In the present paper, we review ITS in the domains of arts, music, and sports, focusing on how they address these domains’ characteristic challenges. We emphasize the role of AI and other technological advances in existing ITS and provide impulses how such technologies may be used in future systems. What do we mean by ITS? We define ITS as computer systems that perform tutoring functions, such as presenting information, posing tasks, and providing feedback, and that automatically individualize at least one of these functions based on a model of the student (Ma et al., 2014 ). According to the dynamic framework of personalized education (Tetzlaff et al., 2021 ), the personalization of instruction occurs on three levels. At the macro level, an individual learning goal is selected and its achievement is measured by summative assessment. At the meso level, a task that fits the learner’s prerequisites is selected and the learning progress is measured by formative assessment. On the micro level, assistance is selectively provided or withheld during task completion based on an on-line assessment of the learner’s performance. In ITS, instruction is typically individualized via three system components: the domain model (also called the expert model), the student model, and the tutor model (Ma et al., 2014 ; Sottilare et al., 2013 ). The domain model represents the knowledge that the student is intended to learn. It contains the relevant skills and pieces of knowledge in a given domain. The student model represents the student’s current knowledge, containing a subset of the skills or knowledge in the domain model. The tutor model represents the instructional strategies. For example, it may contain mechanisms for updating the student model and suggesting tasks. Once a student has completed a task with sufficient accuracy, the tutor model may add the associated skills or pieces of knowledge to the student model (Ma et al., 2014 ). When suggesting new tasks, the tutoring model may target a skill not yet included in the student model (Phon-Amnuaisuk & Siong, 2008 ). Through these mechanisms, the domain, student, and tutoring models collaboratively steer the individualization of instruction. Besides the domain, student, and tutor models, another conceptual component of ITS is the system interface (Ma et al., 2014 ). The interface enables the basic system-learner interaction by receiving information input from the learners and presenting information output from the system. However, the interface is also a highly relevant aspect of the instructional design. The hardware components of the interface, such as a computer screen and mouse, determine what information can be received and presented. These restrictions affect the system-learner interaction and thus the tutoring functions. The software components of the interface, such as the graphical user interface, guide users’ attention, determine their actions, and influence the clarity, content, and timing of the tasks. In summary, both the hardware and software components are crucial for the instructional design of ITS. What’s special about arts, music, and sports? We argue that there are two characteristics of arts, music, and sports that make them challenging domains for ITS. First, their subdomains lie on a continuum between rather well-formalized (e.g., sketching, music theory, yoga) and rather open-ended (e.g., abstract painting, music composition, team sports). Open-ended subdomains are especially challenging, because tasks have a large number of possible solutions and lack clear performance benchmarks. This raises the question of how instruction can be individualized in these open-ended subdomains. If there are myriad possible solutions, how should the system select a learning goal? If there are no clear performance benchmarks, how should the system determine whether a task has been completed or whether a skill can be added to the student model? Second, arts, music, and sports are multimodal domains. They involve multimodal information, such as the sound of a musical instrument, a sketched image, or the swing of a bat. Therefore, the system interface must be able to receive, process, and present these types of information. While this has long been expensive or impossible, recent technological advances have brought about many new types of interfaces, such as motion sensors, virtual reality, and AI agents. However, it remains unclear how ITS in multimodal domains use and combine innovative technologies and how technology use is guided by instructional design considerations. Present Study The present work presents a systematic literature review of ITS in arts, music, and sports. An initial reading of the literature revealed that ITS studies in these fields were methodologically diverse and often lacked an effectiveness test. Therefore, we decided to focus solely on how these systems function and on the innovative technologies they use rather than on their efficacy. Consequently, we conducted a narrative review (Baumeister & Leary, 1997 ). Such reviews synthesize the results of a set of studies by identifying overarching topics or reinterpreting them, but they do not analyze the statistical significance of the findings (Siddaway et al., 2019 ). Our analyses were guided by two main research questions. First, how do ITS in arts, music, and sports individualize instruction given the open-ended nature of tasks? Second, how do ITS in arts, music, and sports leverage modern technologies to receive and process relevant multimodal information? Our goal was thus to provide a systematic overview of existing ITS in these uncommon domains to inform and inspire researchers and practitioners in these fields, as well as ITS researchers in other fields. Method Search string The general structure of the search string in our systematic literature search was to combine the term “intelligent tutoring system” with the terms “arts,” “music,” and “sports.” We searched previous reviews on the topic for synonyms of the term "intelligent tutoring system." We found a suitable list of synonyms in Steenbergen-Hu & Cooper ( 2014 ): artificial tutor*, computer tutor*, computer-assisted tutor, computer-based tutor*, intelligent learning environment*, computer coach*, online-tutor*, keyboard tutor*, e-tutor*, electronic tutor*, web-based tutor*. We decided not to use the terms “keyboard tutor” and “artificial tutor” as an initial literature search revealed that they are not common terms for digital learning systems. To derive our search string, we then removed the truncation symbols and placed the individual terms in quotation marks. This was done to avoid searching for overly general terms such as “computer,” “online,” or “intelligent.” Moreover, we searched the APA Thesaurus of psychological index terms and the Education Thesaurus for synonyms for the terms “arts,” “music,” and “sports”. This search yielded only specific activities as synonyms (e.g., "exercise" for sports or "improvisation" for music). Especially in the case of sports and music, there are many specific activities associated with the broader terms. Using these activities as synonyms would have resulted in an overly inclusive search string or an unwarranted focus on single activities. Therefore, we decided not to include any synonyms for the terms “arts,” “music,” and “sports.” Table 1 Search string used in each database. The first line shows the string that was used as a root in all databases. Google Scholar does not allow for parenthesis. Google Scholar and ERIC do not allow for truncation symbols (also called wildcard operators). Database Search string “intelligent tutoring system” OR “AI tutor” OR “computer tutor” OR “computer-assisted tutor” OR “computer-based tutor” OR “computer-based tutoring system” OR “intelligent learning environment” OR “computer coach” OR “online tutor” OR “e-tutor” OR “web-based tutor” Google Scholar AND “art” OR “arts” AND sport AND music APA PsycInfo, APA PsycArticles, Education Research Complete, The New Republic Archive (DFG), ProQuest Dissertation & Thesis Global AND (arts OR music* OR sport*) ERIC AND (arts OR music OR sports) PSYNDEX AND (Kunst OR arts OR musi* OR sport) Finally, the search string was modified to account for differences in database functionality. For example, Google Scholar does not allow parentheses. Thus, we performed three searches, combining “intelligent tutoring system” and its synonyms with each of the three terms „arts,“ „music,“ and „sports“ individually. In all other databases, we used parentheses and Boolean operators to combine the terms. For databases that allowed truncation symbols, we used „music*“ and „sport*“ as search terms to find results for „musical,“ „musician,“ „sportive,“ „sports,“ etc. We refrained from using „art*“ as a search term to avoid finding results related to artificial intelligence. PSYNDEX also contains German-language literature. Thus, we adapted the search string to cover the German terms for arts, music, and sports (Kunst, Musik, and Sport, respectively). The final search strings used in each database can be found in Table 1 . Inclusion criteria Our inclusion criterion was that the study in question described a tutoring system in arts, music, or sports in sufficient detail to understand how the system functions, and was written in English or German. It was not an inclusion criterion that studies performed a controlled test of the effect of the system in comparison to a human tutor. Our focus was on the systems’ functioning and use of innovative technologies, not on their efficacy. We excluded papers that only described how ITS in arts, music, or sports could be designed in general (“unspecific”), did not explain the ITS in sufficient detail (“unclear”), referred to a different type of system or to an ITS in a different domain than arts, music, or sports (“off-topic”), were literature reviews themselves (“review”), described merely a technology that could be used in an ITS (“technology”), or described a commercial, closed-source system (“commercial”). Figure 1 shows how many studies were excluded due to each criterion. Screening procedure Since Google Scholar does not allow for the bulk export of metadata, the initial screening was performed within the web browser. Both reviewers performed the search independently, reading titles (and abstracts, if necessary) and adding items to their personal libraries if they deemed them relevant. They continued the search until they reached a maximum of 1,000 results or until they found no relevant items among 100 consecutive results. After completing the screening, both reviewers exported the metadata from their Google Scholar personal libraries and imported it into Zotero. Then, they searched the other databases, exported the metadata of all the results, and imported them into Zotero. Duplicates were removed, and the remaining results were screened in Zotero. Next, the first author screened all results considered relevant by the second reviewer. Subsequently, the first author sought to retrieve all relevant articles. If an article was not accessible, the first author requested the full text via ResearchGate or contacted the authors via email. Finally, the first author performed an in-depth screening of all 146 papers selected during the initial screening by skimming or reading their main texts. Figure 1 shows the Prism flowchart of the entire screening process. Results Individualization of instruction Open-endedness of the subdomains Our first research question was how ITS in arts, music, and sports individualize instruction, given the open-ended nature of these domains. To address this question, we first analyzed the domains and subdomains of the reviewed ITS. As mentioned earlier, the subdomains of arts, music, and sports lie on a continuum ranging from rather well-formalized to rather open-ended. Thus, we analyzed which subdomains the reviewed ITS address. Table 2 lists all of the reviewed ITS and their respective domains and sub-domains. Note that some of the 39 publications resulting from the systematic literature search refer to the same ITS. Moreover, for each ITS that was described in a publication, we searched for additional literature that was not yet included in our dataset. This resulted in 29 ITS and 42 associated publications. With a look on the domain and subdomain column in Table 2 , it becomes clear that the reviewed ITS avoid overly open-ended subdomains. The most common subdomains, musical instrument performance and music theory, are rather well-formalized than open-ended as a single correct solution can be defined. While more advanced levels of instrument performance allow for some freedom of musical expression, the fundamentals of playing an instrument involve producing the correct tones at the right time. Music theory is governed by clear, logical, and often even mathematical rules. In addition, the four reviewed sports ITS and the two reviewed arts ITS addressed well-formalized subdomains as well. In dancing, yoga, and strength training, there is typically one desired way to perform movements and poses. Sketching involves creating a realistic copy of a model, and there are clear benchmarks for a good sketch. Table 2 The ITS that were part of the present review together with the publications that presented them, their domains and subdomains. Name of ITS Publications Domain Subdomain Piano Tutor (Capell & Dannenberg, 1993 ; Dannenberg et al., 1990 , 1993 ; Sanchez et al., 1987 ) Music Instrument performance Boland Piano Tutor (Boland, 2003 ) Music Instrument performance Digital Violine Tutor (Percival et al., 2007 ) Music Instrument performance MEAWS (Musician Evaluation and Audition for Winds and Strings) (Percival et al., 2007 ) Music Instrument performance iDVT (interactive digital violin tutor) (Huanhuan, 2009 ) Music Instrument performance BACh (Brain Automated Chorales) (Yuksel et al., 2016 ) Music Instrument performance guitMaster (Grigutis, 2018 ) Music Instrument performance instruMentor (Bagga et al., 2019 ) Music Instrument performance Affective Cognitive Tutor for Rhythm (Sanz, 2022 ) Music Instrument performance Interactive Rainbow Score (Chin et al., 2020 ; Chin & Xia, 2022 ) Music Instrument performance Pianobot (Della Ventura, 2022 ) Music Instrument performance HoloMusic XP (Molero et al., 2021 ) Music Instrument performance IMUTUS (Interactive Music Tuition System) (Fober et al., 2004 ; Raptis et al., 2005 ; Schoonderwaldt et al., 2005 ) Music Instrument performance & music theory KANT (Kritical Argument Negotiated Tutoring) (M. Baker, 1990 ; M. J. Baker, 1992 ) Music Music theory Sonata (Angelides & Tong, 1995 ; C Angelides & KY Tong, 1994; Tong & Angelides, 2000 ) Music Music theory ChordTeacher (Greiner, 2011 ) Music Music theory Chat Melody (Jin et al., 2025 ) Music Music theory Maestoso (Taele et al., 2015 ) Music Music theory mySolfeggio (Debevc et al., 2020 ) Music Music theory FEEL-ME (Feedback Learning of Musical Expressivity) (Juslin et al., 2006 ) Music Emotional expressivity ImproScales (Borgogno & Turchet, 2022 ) Music Improvisation Raagang (Gajjar & Patel, 2020 ) Music Improvisation Y-system (Chen et al., 2018 ) Sports Yoga e-YogaGuru (Kale et al., 2021 ) Sports Yoga Dancing Coach (Romano et al., 2019 ) Sports Dancing Selfit (Guarnieri et al., 2023 ; L. Neagu, 2022 ; L.-M. Neagu et al., 2021 ; L.-M. Neagu, Rigaud, Guarnieri, Dascalu, et al., 2022) Sports Exercise SketchTivity (Hammond et al., 2018 ; Keshavabhotla et al., 2017 ; Williford, 2017 ; Williford et al., 2020 ) Arts Sketching TAYouKi (Tutoring Assistant for Young Kids) (Vides Ceron, 2012 ) Arts Sketching We found only three ITS in rather open-ended subdomains, namely improvisation and emotional expression. In ImproScales (Borgogno & Turchet, 2022 ) and Raagang (Gajjar & Patel, 2020 ), users can practice improvising melodies in the Western and in the Hindustani tonal system, respectively. However, a closer look reveals that these ITS also address the well-formalized aspects of this creative activity. In both Western and Hindustani musical traditions, the tones that are used during improvisation are restricted to a certain musical scale. That is, there is a set of “allowed” tones that can be used during improvisation in a given song. Both ImproScales and Raagang address the skill of sticking to the tones of the musical scale during improvisation. In the ImproScales system, users can improvise either freely or with a backing track from Youtube. They can indicate the employed scale or the system can automatically detect the scale of the backing track. The system then provides feedback on whether the performed notes are part of the scale. Similarly, in Raagang, users can select a Hindustani scale (Raga), upload an audio recording of their improvisation in that scale, and the system provides feedback on which notes of the improvisation were from the selected scale. Thus, both ImproScales and Raagang exclusively address a well-formalized aspect of improvisation. Another seemingly open-ended activity supported by the Feel-ME system (Juslin et al., 2006 ) is emotional expression in musical instrument performance. However, even this rather vague and subjective aspect of musical performance is broken down into objective rules. The basic idea behind the system is that musicians use musical parameters, such as tempo or loudness, to express emotions, and a certain usage of these parameters trigger emotions in listeners. For emotional expression to be successful, musicians should use musical parameters in a way that triggers the desired emotions in listeners. Formally, this means that the correlation between musical parameters and emotions should be similar for the performance of a musician and the evaluations of the listeners. Based on these ideas, the Feel-ME system uses correlations between musical parameters (tempo, timbre, articulation) and emotions from a training dataset in which listeners evaluated the emotional expression in a large set of musical performances. Users who want to practice their emotional expression with Feel-ME, they perform several melodies and indicate the emotion they intended to express with each performance. Using the data from all the performances, the program then calculates a point-biserial correlation between the user’s intention to express an emotion (0/1) and the musical parameters (tempo, sound level, articulation, timbre). It further calculates the fit between the correlations in the user’s performance and the correlations in the training dataset. Thereby, the program is able to provide feedback such as, “If you want to express sadness, you should play more legato.” It becomes clear that, although emotional musical expression seems to be an open-ended domain, the system is based on well-formalized, statistical rules. In summary, this first step of analysis revealed that none of the reviewed ITS address truly open-ended and creative activities such as expressive painting, musical composition, or decision making in team sports. This is a clear gap in the ITS landscape. Individualized feedback and human-led task selection However, individualizing instruction can be challenging in well-formalized domains as well. Activities such as playing a musical instrument, sketching, and strength training are complex and multifaceted. Thus, the next step in our analyses was to investigated how instruction was individualized in the reviewed ITS. We found that several ITS did not individualize the instructional activity or content but merely provided feedback for a fixed set of tasks. The flute tutoring system Interactive Rainbow Score (Chin et al., 2020 ; Chin & Xia, 2022 ) provides performance feedback for 16 folk songs. The yoga tutors Y-system (Chen et al., 2018 ) and e-YogaGuru (Kale et al., 2021 ) both contain a set of built-in yoga poses. Users can perform one of these yoga poses in front of a camera and receive feedback on their posture. The Dance Coach (Romano et al., 2019 ) provides feedback on basic salsa steps. In addition to these feedback-based ITS, several systems circumvented the challenge of automated task selection by having learners or a human tutors steer the individualization of the instruction. In MEAWS (Percival et al., 2007 ), Maestoso (Taele et al., 2015 ), and mySolfeggio (Debevc et al., 2020 ), teachers or students can create their own lessons. In HoloMusic XP (Molero et al., 2021 ), students must select a lesson from a predetermined set. Likewise, in the sketch tutoring system SketchTivity (Hammond et al., 2018 ; Keshavabhotla et al., 2017 ; Williford, 2017 ; Williford et al., 2020 ), exercises have a fixed sequence of increasing difficulty, and learners can decide whether to retry an exercise or proceed to the next one. A rather advanced way of combining human and virtual tutoring is implemented in the Boland piano tutor (Boland, 2003 ). A human tutor can create a lesson plan with different objectives and associated success criteria. For instance, a human tutor may want a student to learn to play a song at a fast tempo. Then, the teacher can define a sequence of lessons, each consisting of a certain tempo and a target accuracy. Once the student performs the song at the given tempo with sufficient accuracy, they can proceed to the next lesson, which increases the tempo. Some of the reviewed ITS systems even put responsibility for aspects other than lesson content in the hands of students. In the flute tutoring system IMUTUS (Fober et al., 2004 ; Raptis et al., 2005 ; Schoonderwaldt et al., 2005 ), for example, students must decide how to proceed after receiving feedback on their performance from the system. They can choose to get more information about the error, repeat the phrase containing the error, complete a special exercise addressing the error, or repeat the entire piece. Similarly, learners using the instruMentor (Bagga et al., 2019 ), a robot flute tutor, need to decide on their own about the practice mode: playing alone, playing together with the robot, or observing the robot to play. In summary, many of the reviewed ITS lack automated individualized task selection based on a student model. Instead, these systems provide individualized feedback or put task selection in the hands of learners or human tutors. Automatic individualized task selection However, we also found some systems with automatic individualized task selection. Two of the reviewed systems automatically adapted task difficulty to individual learners. In the guitMaster system (Grigutis, 2018 ), users can practice to perform unknown melodies on the guitar at first sight. The system generates the melodies and analyzes the accuracy. The difficulty of the melodies is defined by the number of strings used in combination with the performance tempo. Once an accuracy level of at least 70% is achieved, users are instructed to increase the difficulty. If the accuracy is below 50%, the system instructs users to decrease the difficulty. In the BACh system (Yuksel et al., 2016 ) in which users practice to perform Bach chorales on the piano, the difficulty of the next chorale increases if cognitive load is low. Cognitive load is measured using functional near-infrared spectroscopy (fNIRS). The classical instrument tutoring system Piano Tutor (Capell & Dannenberg, 1993 ; Dannenberg et al., 1990 , 1993 ; Sanchez et al., 1987 ) is based on an explicitly defined expert, student, and tutor model. The expert model consists of a set of lessons, i.e., single instructional activities that teach a new concept or skill. For example, a lesson may aim to teach students how to perform a new rhythm, such as a 3/4 measure (see (Capell & Dannenberg, 1993 ). Once a learner completes this lesson, the associated skill is added to their student model. Each lesson has prerequisite skills. The system can only suggest a lesson to a learner if the prerequisite skill is already part of their student model. For example, to complete the 3/4 rhythm lesson, a learner must have mastered the skills to detect and perform different notes and to perform the 2/4 rhythm. The ChordTeacher system (Greiner, 2011 ) is based on a similar principle. The expert model consists of the theory of musical chords. It defines root notes, musical intervals, and the structure of different types of chords. It also explains how chords can be displayed, such as on piano keys, as staff notation, or in a Harmony Space matrix. There are various tasks that involve building or recognizing intervals or chords. Once a learner completes a task, their proficiency score increases for the interval, chord type, and representation involved in the task. Each task has a prerequisite proficiency score and can only be selected by the system once the learner has reached it. Another system with a sophisticated expert, student, and tutor model is the Selfit strength training system in sports (Guarnieri et al., 2023 , 2023 ; L.-M. Neagu et al., 2021 ; L.-M. Neagu, Rigaud, Guarnieri, Radu, et al., 2022 ). The authors first developed the OntoStrength ontology to serve as a basis for the system (L.-M. Neagu, Rigaud, Guarnieri, Radu, et al., 2022 ). This ontology contains formal definitions of terms and concepts, as well as their relations within the domain of strength training. One subdomain of the ontology describes strength skills by combining specific muscles with contraction modes (eccentric, concentric, isometric, plyometric) and strength properties (power, maximum strength, endurance). A skill in this subdomain may be, for example, “Biceps Eccentric Maximum Strength.” Each strength skill is associated with a movement that trains it. Another subdomain of the ontology describes training programs by defining templates of training cycles and exercise blocks. Selfit’s domain model then contains more than 1,000 exercises, which are defined by a number of parameters such as the movements or number of joints involved. The student model contains general information about the trainee, such as age, weight, and sex, as well as self-reports on specific events, such as pain, injury, or surgery, and physical condition, such as fatigue level, motivation, sleep quality, and stress level. In addition, after each exercise, users must indicate how many more repetitions they could perform (repetitions in reserve). This information is also added to the student model and is used during task selection by a machine learning algorithm, which we will describe in greater detail in the next section. At the beginning of a training session, the trainee can select which muscle to train. Using the OntoStrength ontology, the system first derives which movements train the selected muscle. Then, based on the domain model, it derives which exercises involve these movements. Considering the information in the student model and the training templates in the ontology, the system then creates an individualized training plan. Usage of modern technologies Our second research question was how ITS in arts, music, and sports leverage modern technologies to accommodate their inherently multimodal nature. We distinguished three classes of technologies in our analyses: input technologies for capturing user behavior (e.g., sensors and detectors), data-processing technologies for interpreting this input (e.g., machine learning algorithms), and output technologies for delivering feedback and instruction (e.g., virtual reality headsets). Table 4 provides an overview of the technologies employed in the reviewed ITS and their purpose. Table 3 Overview of the different types of individualization in the reviewed ITS. Individualization Explanation ITS Individualized feedback The system provides feedback for a fixed set of tasks. Instructional activities, explanations or contents are not individualized. Y-system e-YogaGuru Dance Coach Rainbow Score Human-controlled individualization Learners or human tutors can decide about the instructional activities or contents. MEAWS Maestoso mySolfeggio HoloMusic XP Boland piano tutor IMUTUS instruMentor SketchTivity Individualized task difficulty Task difficulty is adapted based on performance accuracy or cognitive load. guitMaster BACh Individualized task selection Each task is associated with required skills and skills it targets. Targeted skills are added to the student model upon task completion. New tasks are selected whose required skills are already part of the student model and whose targeted skills are not. PianoTutor ChordTeacher Selfit Analyzing the input technologies, it became clear that most of the reviewed ITS relied on rather classical technologies, such as microphones, cameras and MIDI (Musical Instrument Digital Interface). However, some ITS indeed used more advanced technologies to receive multimodal input. The sketch tutoring systems SketchTivity (Williford, 2017 ) and TaYouKi (Vides Ceron, 2012 ), as well as the music theory tutoring system Maestoso (Taele et al., 2015 ), receive user input via a digital stylus, also known as ink digitizer. With a digital stylus, users can write or draw on a computer screen or on a special pad and their strokes are transmitted directly to a computer. All three ITS all used digital styluses from Wacom ( www.wacom.com ), which measure pressure, tilt, and stroke speed using electromagnetic resonance and sensors. The advantage of using a digital stylus is that it enables real-time feedback on activities such as writing musical notation or drawing, while preserving the natural feeling of the activity. Its intuitive design allows allows even young children to use a digital stylus. Another input technology—the Microsoft Kinect depth camera—is used by the dance tutoring system Dancing Coach (Romano et al., 2019 ) and by the yoga tutoring systems Y-system (Chen et al., 2018 ) and e-YogaGuru (Kale et al., 2021 ). Depth cameras can capture body poses or movements at a fine-grained level by combining cameras, infrared transmitters, and infrared detectors (D. Johnson et al., 2019 ; Mat Sanusi et al., 2021 ). In addition to tracking human bodies, depth camera software can create skeleton representations based on joint locations. For the Dancing Coach, Romano et al. ( 2019 ) enabled the system to detect and evaluate basic salsa steps by providing recordings of these movements to the Kinect Visual Gesture Builder. For the Y-system, Chen et al. ( 2018 ) created explicit posture description models for each of the twelve implemented yoga poses. For these posture models, reference points were defined on the body contours of each yoga pose, which were extracted from the images of the depth camera. Then, a correct posture was defined by defining angles between lines connecting some of the reference points. The e-YogaGuru uses skeleton representations from Kinect recordings of eight experts as references to provide feedback. In summary, depth cameras are an important technology for ITS in sports domains as they enable real-time evaluation of body movements. Table 4 Overview of technologies employed in the reviewed ITS and their purpose Technology Purpose ITS Digital stylus and sketch recognition Receive and interpret hand-drawn sketches and hand-written musical notation TaYouKi SketchTivity Maestoso Depth camera and movement recognition Receive and interpret body movements and postures Dancing Coach Y-system eYogaGuru Reinforcement learning Determine the most effective strength exercise Selfit Large language model Receive responses and questions and provide feedback and instruction in natural language Chat Melody Robots Demonstrate movements on instrument or provide motivating interaction InstruMentor Pianobot Augmented reality Show movement triggers on instrument HoloMusic XP Our second category of technologies, data-processing technologies, mainly refers to software and algorithms, such as machine learning algorithms. The sketch tutoring systems TaYouKi, SketchTivity, and Maestoso use sketch recognition algorithms which enable computers to interpret hand-drawn sketches or writing. The reviewed systems use two types of these algorithms: geometric and pattern matching. Geometric algorithms, such as PaleoSketch (Paulson & Hammond, 2008 ), derive basic shapes from the properties of a sketch. These properties may include the number of lines or the number of sudden changes in direction. Pattern matching algorithms require a correct reference sketch for comparison. First, the sketch is pre-processed to match the size and orientation of the reference. Then, equally spaced points are defined on the lines of the sketch and the reference, and a resemblance measure is calculated based on the distances between the points. The reviewed systems use these algorithms to recognize sketched objects or handwritten musical symbols and provide corrective feedback. The AI yoga tutor uses algorithms to automatically provide feedback. First, a large set of photos of yoga poses was collected, and the OpenPose algorithm was used to extract skeleton representations of the bodies in the photos. Then, three yoga experts rated the quality of the poses. The skeleton representations and the associated quality evaluations were provided to a machine learning algorithm. Thereby, the algorithm was trained to evaluate the quality of yoga poses in new photos. Besides providing feedback, machine learning algorithms can be used to determine which instructional activities the system should suggest. The Selfit strength training tutor uses a special type of reinforcement learning called a multi-armed bandit algorithm. These algorithms address the problem of choosing among fixed options (in this case, exercises) when the properties of each option (in this case, their training effect) are only partially known. In simple terms, the algorithm is based on three principles: (1) each option is associated with a reward, (2) the reward associated with the chosen option is updated after a choice (i.e., increased if the choice led to a reward or decreased if the option did not lead to a reward), and (3) choices should maximize the overall reward. In Selfit, exercises are considered optimal if they are completed with a small number of repetitions in reserve. For example, if an exercise is planned with 12 repetitions, it is considered optimal if the trainee completes these 12 repetitions but indicates that they could not have performed any further ones. Thus, Selfit’s algoritgm uses the trainee’s self-reported repetitions in reserve to adjust the reward associated with an exercise. A small number of repetitions in reserve increases the reward, while a large number of repetitions in reserve, whether positive or negative, decreases the reward, indicating that the exercise was too difficult or too easy. The reward affects the probability that an exercise will be suggested again, optimizing the system’s suggestions. Another way of individualizing instructions is implemented in the music theory tutor Chat Melody (Jin et al., 2025 ). The system uses a large language model (LLM) to enable chat-based interactions between the user and the system. The user is given a notated melody and is asked to perform some harmonic analysis on it. The user can respond and ask questions in the chat, and the system provides corrective feedback, hints, or reflective questions. While many aspects of the system remain unclear, Chat Melody offers an initial glimpse into the implementation of ITS in arts, music, and sports. Lastly, we found two types of output technologies: robots and augmented reality (AR). The flute tutoring system InstruMentor (Bagga et al., 2019 ) uses a robot with two humanoid hands that can play the flute. The robot can demonstrate the learner how to play a song. In contrast, the Pianobot (Ritschel et al., 2020 ) is a social robot that communicates with the user to provide feedback, hints, and advice. In doing so, the Pianobot uses gaze and facial expression. This interaction is intended to increase motivation to exercise between lessons with a human tutor. Lastly, the HolowMusic XP piano tutoring system (Molero et al., 2021 ) uses Microsoft Hololens augmented reality glasses to present graphics overlaying the surrounding environment. Through the glasses, users see note symbols falling down on their fingers on the piano, which they must “catch” by pushing the corresponding keys. While it is problematic that users of this system do not really learn to perform musical notation, the system exemplifies how AR can be used to provide instructions referencing external objects, such as musical instruments or sports equipment. Discussion ITS play an increasingly important role in today’s education. Their four main conceptual components are the domain, student, and tutor models, which jointly steer the individualization of instruction, and the interface that enables and restricts the system-learner interaction (Ma et al., 2014 ). Arts, music, and sports are challenging domains for ITS due to their open-ended and multimodal nature. We conducted a systematic, narrative review (Baumeister & Leary, 1997 ; Siddaway et al., 2019 ) of ITS in arts, music, and sports, analyzing how these systems address the challenges posed by these domains. Specifically, we analyzed how these systems individualize instruction in the face of open-ended tasks and use innovative technologies to receive, process, and present multimodal information. Regarding the individualization of instruction, we found that ITS in arts, music, and sports avoid open-ended domains. They exclusively address complex, multifaceted skills in rather well-formalized domains. Moreover, automated, individualized task selection based on a student model—often considered a hallmark feature of ITS (Steenbergen-Hu & Cooper, 2014 )—was implemented in only a few systems. Instead, tasks had a fixed sequence or had to be selected manually. Systems that individualized instruction did so at the meso level of the dynamic framework of personalized education (Tetzlaff et al., 2021 ). That is, generally speaking, performance is assessed upon completion of an instructional unit, and the next unit is selected based on the learner’s individual learning prerequisites. Future systems should extend individualization to the overarching learning goal (macro level) and the assistance during tasks (micro level). Overall, ITS in arts, music, and sports seem to focus more on providing feedback than on individualized task selection. Regarding the use of innovative technologies—our second research question—we found that ITS in arts, music, and sports use new types of interfaces, such as digital styluses, depth cameras, robots, and AR, to capture or display multimodal information. Machine learning algorithms are used to optimize individualized task selection, and chatbots to enable system-user interaction in natural language. A major focus of ITS in arts, music, and sports appears to be finding innovative and domain-specific ways to capture body movements. This comprises not only the movement of the body in itself but also its movement in relation to other objects, such as pens or musical instruments. Future systems should also consider further technologies such as special sensors that detect pressure on the fret of a string instrument (Grosshauser & Tröster, 2014 ), the force, position, and tilt of fingers on piano keys (Grosshauser et al., 2012 ), or the pressure of trombone players’ lips (Grosshauser et al., 2013 ), and suites (R. M. G. Johnson et al., 2010 ; Lieberman & Breazeal, 2007 ) and exoskeletons (Moringen et al., 2021 ), which can guide movements through vibrotactile stimulation. Using LLMs to support creative skills The lack of ITS that support truly open-ended, creative skills is a gap in the ITS landscape that should be addressed by future research. Closing this gap requires an understanding of the relevant tutoring functions in such a context. Tutoring in open-ended domains is typically dialogical and encourages creative, metacognitive, and critical thinking rather than simply stating facts (Cook, 1998 ). In the domain of musical composition, it has recently been found that teachers aim to cultivate originality and the clarity of artistic intentions (Lörch & Huovinen, 2025 ). To achieve this, teachers reported using students’ own compositional work as the basis of their teaching, asking students self-reflective questions, challenging their compositional decisions, and providing new perspectives. How may these tutoring functions be implemented in ITS? A promising approach is using chatbots based on large language models (LLM). Users could upload one of their artistic works. The chatbot could then compare the uploaded piece to works in the training dataset or found online. Thereby, the chatbot could evaluate the originality of the work and recommend composers or artists with similar approaches. The chatbot could also ask users about their artistic intentions, aesthetic ideals, and the reasons behind their artistic decisions. Users’ responses to these questions could feed into a student model which guides subsequent interactions. For example, if a painter indicated to dislike strong colors, the chatbot may suggest that they look into the works of an artist who used strong colors to help them reflect on the origins of their aesthetic preference. While such a system cannot replace a human tutor, it could support users between regular lessons and encourage self-reflection and self-criticism. Using machine learning to provide feedback for complex skills Many ITS in arts, music, and sports merely provide feedback rather than individualizing task selection or content. According to some more restrictive definitions, these systems would not even qualify as ITS. However, according to our definition of ITS which is based on Ma et al. ( 2014 ), these systems can indeed be considered ITS because they perform one tutoring function (providing feedback) in an individualized manner. In addition to the importance of feedback in computerized instruction in general (Azevedo & Bernard, 1995 ), feedback may be especially important in arts, music, and sports, since the skills in these domains are typically complex and multifaceted. When performing a yoga pose, for example, many aspects are relevant, such as the overall body posture, the position of the limbs, and the angles between limbs. It can be difficult to monitor all these aspects during performance. Therefore, providing individualized feedback may be especially relevant for acquiring complex, multifaceted skills. Our findings revealed the important role of machine learning algorithms (Campesato, 2020 ) in providing feedback in this context. These algorithms can detect patterns and regularities in large, complex datasets. To provide feedback on a complex, multifaceted skill, one can provide the algorithm with a training dataset containing recordings of expert performances (or of performances rated by experts). The algorithm can then provide feedback by comparing a learner’s performance with the patterns and regularities found in the training dataset (Nguyen et al., 2025 ; Owusu, 2007 ; Santos, 2019 ; Wesely et al., 2025 ; Yang et al., 2021 ). This approach is especially powerful when combined with other algorithms that extract specific performance parameters. Examples include body and movement recognition algorithms (Chiang et al., 2019 ; Hui, 2023 ; Ji, 2020 ; Kotte et al., 2023 ; Lin et al., 2023 ), sketch recognition algorithms (Iarussi et al., 2013 ; Mittal et al., 2015 ; Zhang et al., 2020 ), and algorithms that detect pitches, chords, and harmonies (Brandao, 2005 ; Della Ventura, 2022 ; Jamshidi et al., 2021 ; Zhou & Gong, 2020 ). Incorporating the outputs of these algorithms into training datasets alongside expert ratings will enhance the specificity of the feedback. That is, the machine learning algorithm will not only be able to categorize new performances based on their quality but will also be able to point out specific problems in certain parameters. Domain, student, and tutor models for complex skills Some ITS in arts, music, and sports contain sophisticated domain, student, and tutor models that enable individualized instruction for complex, multifaceted skills. After synthesizing the mechanisms of the reviewed ITS, we identified a general approach for individualized instruction in this context. This approach is based on the idea that subject didactics and pedagogical expertise enable breaking down a complex skill into component skills and to map these components to tasks. In this approach, the domain model contains all the component skills, and the student model contains the subset of these skills that the student has learned already. The tutor model contains the tasks that are implemented in the ITS, the benchmarks for completing them, their prerequisite and target skills, and mechanisms to update the student model and suggest new tasks. A typical mechanism for updating the student model is to increase proficiency in the skills associated with a task after a student completes it. A typical mechanism for suggesting new tasks is to suggest tasks whose prerequisite skills are part of the student model, but whose target skills are not. The strengths of this approach are its transparency and its clear grounding in subject didactics. Its weakness is that decomposing a complex skill into components and mapping them to tasks can be cumbersome and time-intensive. Limitations Although the present work offers valuable insights into ITS in arts, music, and sports, several limitations should be pointed out. During the screening process, it was sometimes difficult to distinguish studies that presented proper ITS from those that merely presented technologies that could be used for ITS. Since we considered the latter to be potentially relevant, we screened their content and included it in the discussion whenever possible. Moreover, many systems were not described in sufficient detail to understand how they functioned, and 18% of the potentially relevant papers were not available. Thus, promising approaches that may have been in these works were not included in this review. Many of the presented studies were published as theses or in conference proceedings and thus, it is unclear whether they were peer-reviewed. Lastly, our analytical approach—the systematic, narrative review—is inherently subjective. It does not involve any quantitative analyses. Instead, the content of the reviewed studies is summarized qualitatively, overarching topics and recurring themes are identified. This process naturally depends on the researchers and their background, posing the risk that important issues are overlooked. However, despite these weaknesses, we believe that a systematic, narrative review was the most appropriate approach for this small, methodologically diverse research field. Conclusion ITS in arts, music, and sports is a young and evolving research field. It profits from the current technological developments, such as VR, sensors, machine learning, and AI, as these developments enable computers to harness new types of information and open new possibilities for system-user interaction. Researchers have developed sophisticated methods for handling complex, multifaceted skills, including providing individualized feedback, combining human and computerized tutoring, and breaking down complex skills into components. However, one major challenge still lies ahead—addressing truly open-ended, creative skills. We believe our work has provided important impulses for the field to finally face this challenge. Declarations Author Note Correspondence concerning this article should be addressed to Lucas Lörch, DIPF - Leibniz Institute for Research and Information in Education, Rostocker Straße 6, 60323 Frankfurt a. M.. Email: [email protected] . We want to thank Franziska Cremer for her help during the screening of the articles. This study was funded by the German Federal Ministry of Education and Research (BMBF). The views expressed in this study are those of the authors and may not necessarily reflect those of the funding institution. The study is part of the KuMuS-ProNeD project, which is one of eight project networks in the Music/Arts/Sports Competence Center of the lernen:digital Competence Network. The funding agency played no role in the design of the study, the analysis of the data, the reporting of the results, or the decision to submit the study for publication. References Albacete P, VanLehn K (2000) Evaluating the Effectiveness of a Cognitive Tutor for Fundamental Physics Concepts Anderson JR, Corbett AT, Koedinger KR, Pelletier R (1995) Cognitive Tutors: Lessons Learned. J Learn Sci 4(2):167–207 Angelides MC, Tong AKY (1994) Using hypertext to implement multiple tutoring strategies in an intelligent tutoring system for music learning. J Comput Inform Technol 2(2):123–138 Angelides MC, Tong AKY (1995) Implementing multiple tutoring strategies in an intelligent tutoring system for music learning. J Inform Technol 10(1):52–62. https://doi.org/10.1057/jit.1995.7 Azevedo R, Bernard RM (1995) A Meta-Analysis of the Effects of Feedback in Computer-Based Instruction. J Educational Comput Res 13(2):111–127. https://doi.org/10.2190/9LMD-3U28-3A0G-FTQT Bagga S, Maurer B, Miller T, Quinlan L, Silvestri L, Wells D, Winqvist R, Zolotas M, Demiris Y (2019) instruMentor: An Interactive Robot for Musical Instrument Tutoring. In K. Althoefer, J. Konstantinova, & K. Zhang (Eds.), Towards Autonomous Robotic Systems (pp. 303–315). Springer International Publishing. https://doi.org/10.1007/978-3-030-23807-0_25 Baker M (1990) Negotiated tutoring: An approach to interaction in intelligent tutoring systems. [Doctoral dissertation, The Open University]. https://www.semanticscholar.org/paper/Negotiated-tutoring-%3A-an-approach-to-interaction-in-Baker/d6ede52bb412c28d20379753eda333dbf3a818c3 Baker MJ (1992) Design of an intelligent tutoring system for musical structure and interpretation. Understanding music with AI: Perspectives on music cognition (pp. 466–489) Baumeister R, Leary M (1997) Writing Narrative Literature Reviews. Rev Gen Psychol 1:311–320. https://doi.org/10.1037//1089-2680.1.3.311 Boland RJ (2003) The Problems Associated with the Design, Implementation and Evaluation of an Intelligent Piano Tutor. [Doctoral dissertation, University of Canberra]. https://researchsystem.canberra.edu.au/ws/portalfiles/portal/57342228/Boland_Robert_P.pdf Borgogno T, Turchet L (2022) ImproScales: A self-tutoring web system for using scales in improvisations. Proceedings of the 17th International Audio Mostly Conference (pp. 219–225) Brandao M (2005) Guided discovery tutoring and schoenberg’s harmony teaching method: An investigation Campesato O (2020) Artificial Intelligence, Machine Learning, and Deep Learning. Mercury Learning and Information Capell P, Dannenberg RB (1993) Instructional Design and Intelligent Tutoring: Theory and the Precision of Design. J Artif Intell Educ 4(1):95–121 Chen H-T, He Y-Z, Hsu C-C (2018) Computer-assisted yoga training system. Multimedia Tools Appl 77:23969–23991 Chiang H-H, Chen W-M, Chao H-C, Tsai D-L (2019) A virtual tutor movement learning system in eLearning. Multimedia Tools Appl 78:4835–4850 Chin D, Xia G (2022) A Computer-aided Multimodal Music Learning System with Curriculum: A Pilot Study. Proceedings of the International Conference on New Interfaces for Musical Expression (NIME) 2022. https://doi.org/10.21428/92fbeb44.c6910363 Chin D, Zhang Y, Zhang T, Zhao J, Xia GG (2020) Interactive Rainbow Score: A Visual-centered Multimodal Flute Tutoring System. Proceedings of the International Conference on New Interfaces for Musical Expression (NIME) 2020 Cook J (1998) Mentoring, Metacognition and Music: Interaction Analyses and Implications for Intelligent Learning Environments Corbalan G, Kester L, Van Merriënboer JJG (2006) Towards a personalized task selection model with shared instructional control. Instr Sci 34(5):399–422. https://doi.org/10.1007/s11251-005-5774-2 Corbett AT, Koedinger KR, Anderson JR (1997) Intelligent Tutoring Systems. In M. G. Helander, T. K. Landauer, & P. V. Prabhu (Eds.), Handbook of Human-Computer Interaction (Second Edition) (pp. 849–874). North-Holland. https://doi.org/10.1016/B978-044481862-1.50103-5 Dannenberg RB, Sanchez M, Joseph A, Capell P, Joseph R, Saul R (1990) A computer-based multi‐media tutor for beginning piano students. J New Music Res 19(2–3):155–173 Dannenberg RB, Sanchez M, Joseph A, Joseph R, Saul R, Capell P (1993) Results from the piano tutor project. Proceedings of the Fourth Biennial Arts and Technology Symposium, 143–150 Debevc M, Weiss J, Šorgo A, Kožuh I (2020) Solfeggio learning and the influence of a mobile application based on visual, auditory and tactile modalities. Br J Edu Technol 51(1):177–193. https://doi.org/10.1111/bjet.12792 Della Ventura MD (2022) A Self-adaptive Learning Music Composition Algorithm as Virtual Tutor. In I. Maglogiannis, L. Iliadis, J. Macintyre, & P. Cortez (Eds.), Artificial Intelligence Applications and Innovations (pp. 16–26). Springer International Publishing. https://doi.org/10.1007/978-3-031-08333-4_2 Fober D, Letz S, Orlarey Y, Askenfelt A, Falkenberg-Hansen K, Schoonderwaldt E (2004) Imutus-an interactive music tuition system. Proceedings of the Sound and Music Computing Conference, 97–103 Gajjar K, Patel M (2020) RAAGANG—a proposed model of tutoring system for novice learners of Hindustani classical music. ICT Systems and Sustainability: Proceedings of ICT4SD 2019, Volume 1, 509–517 Greiner E (2011) A prototype Intelligent Tutoring System for musical chord theory. [Bachelor thesis, University of Amsterdam]. https://staff.fnwi.uva.nl/b.bredeweg/pdf/BSc/20102011/Greiner.pdf Grigutis L (2018) Investigation into aspects of feedback and gamification in computer-assisted musical instrument tutoring. [Master’s thesis, University of Huddersfield]. https://eprints.hud.ac.uk/id/eprint/34633/1/Laurynas%20Grigutis%20FINAL%20THESIS.PDF Grosshauser T, Hufnagl B, Tröster G (2013) Sensor Based Hand and Lip Weight and Pressure Measurements in Trombone Playing. Proceedings of SMAC 2013: 4th Stockholm Music Acoustics Conference, 509–513 Grosshauser T, Tessendorf B, Tröster G, Hildebrandt H, Candia V Sensor setup for force and finger position and tilt measurements for pianists. Proceedings of the 9th Sound and Music Computing, Conference (2012) SMC 2012 Grosshauser T, Tröster G (2014) Musical instrument interaction: Development of a sensor fingerboard for string instruments. Proceedings of the 8th International Conference on Tangible, Embedded and Embodied Interaction, 177–180. https://doi.org/10.1145/2540930.2540956 Guarnieri V, Neagu L-M, Rigaud E, Travadel S, Dascalu M (2023) Enhancing Load Evaluation in Intelligent Tutoring Systems Through Velocity-Based Training. Proceedings of the Conference on Smart Learning Ecosystems and Regional Development, 21–37 Hammond T, Kumar SPA, Runyon M, Cherian J, Williford B, Keshavabhotla S, Valentine S, Li W, Linsey J (2018) It’s not just about accuracy: Metrics that matter when modeling expert sketching ability. ACM Trans Interact Intell Syst (TiiS) 8(3):1–47 Holland S (2000), August 8 Artificial Intelligence in Music Education: A Critical Review. Readings in Music and Artificial Intelligence. https://www.semanticscholar.org/paper/Artificial-Intelligence-in-Music-Education%3A-A-Holland/a78fa73c5d04169c34b2a945a0a38e5a210773b7 Huanhuan L (2009) Computer assisted music instrument tutoring applied to violin practice. [Master thesis, National University of Singapore]. https://www.comp.nus.edu.sg/~leowwk/thesis/luhuanhuan.pdf Hui B (2023) Visualization system of martial arts training action based on artificial intelligence algorithm. Soft Comput, 1–12 Iarussi E, Bousseau A, Tsandilas T (2013) The drawing assistant: Automated drawing guidance and feedback from photographs. Proceedings of the 26th Annual ACM Symposium on User Interface Software and Technology, 183–192. https://doi.org/10.1145/2501988.2501997 Jamshidi F, Marghitu D, Chapman R (2021) Developing an Online Music Teaching and Practicing Platform via Machine Learning: A Review Paper. In M. Antona & C. Stephanidis (Eds.), Universal Access in Human-Computer Interaction. Access to Media, Learning and Assistive Environments (Vol. 12769, pp. 95–108). Springer International Publishing. https://doi.org/10.1007/978-3-030-78095-1_9 Ji R (2020) Research on Basketball Shooting Action Based on Image Feature Extraction and Machine Learning. IEEE Access, 8, 138743–138751. IEEE Access. https://doi.org/10.1109/ACCESS.2020.3012456 Jin L, Lin B, Hong M, Zhang K, So H-J (2025) Exploring the Impact of an LLM-Based Teachable Agent on Learning Gains and Cognitive Load in Music Education Johnson D, Damian D, Tzanetakis G (2019) Detecting Hand Posture in Piano Playing Using Depth Data. Comput Music J 43(1):59–78. https://doi.org/10.1162/COMJ_a_00500 Johnson RMG, Van Der Linden J, Rogers Y (2010) MusicJacket: The efficacy of real-time vibrotactile feedback for learning to play the violin. CHI ’10 Extended Abstracts on Human Factors in Computing Systems. 3475–3480. https://doi.org/10.1145/1753846.1754004 Juslin PN, Karlsson J, Lindström E, Friberg A, Schoonderwaldt E (2006) Play it again with feeling: Computer feedback in musical communication of emotions. J Experimental Psychology: Appl 12(2):79–95. https://doi.org/10.1037/1076-898X.12.2.79 Kale G, Patil V, Munot M (2021) A novel and intelligent vision-based tutor for Yogāsana: E-YogaGuru. Mach Vis Appl 32:1–17 Keshavabhotla S, Williford B, Kumar S, Hilton E, Taele P, Li W, Linsey J, Hammond T (2017) Conquering the cube: Learning to sketch primitives in perspective with an intelligent tutoring system. Proceedings of the Symposium on Sketch-Based Interfaces and Modeling, 1–11 Koedinger K, Anderson J, Hadley W, Mark M (1997) Intelligent Tutoring Goes To School in the Big City. Int J Artif Intell Educ 8:30–43 Kotte H, Kravcik M, Duong-Trung N (2023) Real-Time Posture Correction in Gym Exercises: A Computer Vision-Based Approach for Performance Analysis, Error Classification and Feedback. MILeS@ EC-TEL, pp 64–70 Lieberman J, Breazeal C (2007) TIKL: Development of a wearable vibrotactile feedback suit for improved human motor learning. IEEE Trans Robot 23(5):919–926 Lin K-C, Ko C-W, Hung H-C, Chen N-S (2023) The effect of real-time pose recognition on badminton learning performance. Interact Learn Environ 31(8):4772–4786. https://doi.org/10.1080/10494820.2021.1981396 Lörch L, Huovinen E (2025) Fostering Composer Voice in Tertiary Teaching of Contemporary Music Composition. J Res Music Educ 73(1):87–108. https://doi.org/10.1177/00224294241255579 Ma W, Adesope OO, Nesbit JC, Liu Q (2014) Intelligent tutoring systems and learning outcomes: A meta-analysis. J Educ Psychol 106(4):901–918. https://doi.org/10.1037/a0037123 Martin F, Klein JD, Sullivan H (2007) The impact of instructional elements in computer-based instruction. Br J Edu Technol 38(4):623–636. https://doi.org/10.1111/j.1467-8535.2006.00670.x Mat Sanusi KA, Mitri DD, Limbu B, Klemke R (2021) Table tennis tutor: Forehand strokes classification based on multimodal data and neural networks. Sensors 21(9):3121 McNamara DS, Levinstein IB, Boonthum C (2004) iSTART: Interactive strategy training for active reading and thinking. Behav Res Methods Instruments Computers 36(2):222–233. https://doi.org/10.3758/BF03195567 Mittal P, Vatsa M, Singh R (2015) Composite sketch recognition via deep network—A transfer learning approach. 2015 International Conference on Biometrics (ICB), 251–256. https://doi.org/10.1109/ICB.2015.7139092 Molero D, Schez-Sobrino S, Vallejo D, Glez-Morcillo C, Albusac J (2021) A novel approach to learning music and piano based on mixed reality and gamification. Multimedia Tools Appl 80(1):165–186. https://doi.org/10.1007/s11042-020-09678-9 Moringen A, Rüttgers S, Zintgraf L, Friedman J, Ritter H (2021) Optimizing piano practice with a utility-based scaffold (No. arXiv:2106.12937). https://doi.org/10.48550/arXiv.2106.12937 . arXiv Neagu L (2022) Intelligent Tutoring Systems for Psychomotor Development in Open Environments. [Doctoral dissertation, Université Paris sciences et lettres & Universitatea politehnica Bucarest]. https://pastel.hal.science/tel-04751910/document Neagu L, Rigaud E, Guarnieri V, Dascalu M, Travadel S (2022) Selfit v2 – Challenges Encountered in Building a Psychomotor Intelligent Tutoring System. In S. Crossley & E. Popescu (Eds.), Intelligent Tutoring Systems (Vol. 13284, pp. 350–361). Springer International Publishing. https://doi.org/10.1007/978-3-031-09680-8_33 Neagu L, Rigaud E, Guarnieri V, Radu EI, Travadel S, Dascalu M, Rughinis R (2022) OntoStrength: An Ontology for Psychomotor Strength Development. Interact Des Architecture(s) 52:101–108. https://doi.org/10.55612/s-5002-052-006 Neagu L, Rigaud E, Guarnieri V, Travadel S, Dascalu M (2021) Selfit–an intelligent tutoring system for psychomotor development. Intelligent Tutoring Systems: 17th International Conference, ITS 2021, Virtual Event, June 7–11, 2021, Proceedings 17, 291–295 Nguyen SM, Devanne M, Remy-Neris O, Lempereur M, Thepaut A (2025) A Medical Low-Back Pain Physical Rehabilitation Dataset for Human Body Movement Analysis (No. arXiv:2407.00521). arXiv. https://doi.org/10.48550/arXiv.2407.00521 Owusu G (2007) AI and computer-based methods in performance evaluation of sporting feats: An overview. Artif Intell Rev 27(1):57–70. https://doi.org/10.1007/s10462-008-9068-3 Paulson B, Hammond T (2008) PaleoSketch: Accurate primitive sketch recognition and beautification. Proceedings of the 13th International Conference on Intelligent User Interfaces, 1–10. https://doi.org/10.1145/1378773.1378775 Percival G, Wang Y, Tzanetakis G (2007) Effective use of multimedia for computer-assisted musical instrument tutoring. Proceedings of the International Workshop on Educational Multimedia and Multimedia Education, 67–76 Phon-Amnuaisuk S, Siong CK (2008) Web-Based Music Intelligent Tutoring Systems. In Interactive Multimedia Music Technologies. IGI Global 231–248. https://doi.org/10.4018/978-1-59904-150-6.ch011 Raptis S, Baxevanis A, Askenfelt A, Schoonderwaldt E, Falkenberg K, Fober D, Letz S, Orlarey Y (2005) IMUTUS - An Effective Practicing Environment for Music Tuition Ritschel H, Seiderer A, Andre E (2020) Pianobot: An Adaptive Robotic Piano Tutor Ritter S, Kulikowich J, Lei P-W, Mcguire C, Morgan P (2007) What Evidence Matters? A randomized field trial of Cognitive Tutor Algebra I. (p. 20) Romano G, Schneider J, Drachsler H (2019) Dancing Salsa with Machines—Filling the Gap of Dancing Learning Solutions. Sensors 19(17) Article 17. https://doi.org/10.3390/s19173661 Rowley K, Carlson P, Miller T (1998) A Cognitive Technology to Teach Composition Skills: Four Studies with the R-Wise Writing Tutor. J Educational Comput Res 18(3):259–296. https://doi.org/10.2190/KW4V-FJKD-L7J1-EFK0 Sanchez M, Joseph A, Dannenberg R, Miller P, Capell P, Joseph R (1987) Piano tutor: An Intelligent Keyboard Instruction System. Proceedings of the First International Workshop on Artificial Intelligence and Music,. St. Paul, Minnesota: AAAI Santos OC (2019) Artificial Intelligence in Psychomotor Learning: Modeling Human Motion from Inertial Sensor Data. Int J Artif Intell Tools 28(04):1940006. https://doi.org/10.1142/S0218213019400062 Sanz J (2022) Improving music education using biofeedback-based cognitive tutoring Schoonderwaldt E, Askenfelt A, Falkenberg K (2005) Design and implementation of automatic evaluation of recorder performance in IMUTUS Siddaway AP, Wood AM, Hedges LV (2019) How to Do a Systematic Review: A Best Practice Guide for Conducting and Reporting Narrative Reviews, Meta-Analyses, and Meta-Syntheses. Annual Review of Psychology, 70(Volume 70, 2019), 747–770. https://doi.org/10.1146/annurev-psych-010418-102803 Sosa GW, Berger DE, Saw AT, Mary JC (2011) Effectiveness of Computer-Assisted Instruction in Statistics: A Meta-Analysis. Rev Educ Res 81(1):97–128. https://doi.org/10.3102/0034654310378174 Sottilare R, Graesser A, Hu X, Holden H (eds) (2013) Design Recommendations for Intelligent Tutoring Systems. U.S. Army Research Laboratory Steenbergen-Hu S, Cooper H (2014) A meta-analysis of the effectiveness of intelligent tutoring systems on college students’ academic learning. J Educ Psychol 106(2):331–347. https://doi.org/10.1037/a0034752 Taele P, Barreto L, Hammond T (2015) Maestoso: An intelligent educational sketching tool for learning music theory. Proceedings of the AAAI Conference on Artificial Intelligence, 29(2), 3999–4005 Tamim RM, Bernard RM, Borokhovski E, Abrami PC, Schmid RF (2011) What Forty Years of Research Says About the Impact of Technology on Learning: A Second-Order Meta-Analysis and Validation Study. Rev Educ Res 81(1):4–28. https://doi.org/10.3102/0034654310393361 Tetzlaff L, Schmiedek F, Brod G (2021) Developing personalized education. A dynamic framework. Educational Psychol Rev 33(3):863–882. https://doi.org/10.25656/01:25237 Tong AKY, Angelides MC (2000) An empirical model for tutoring strategy selection in multimedia tutoring systems. Decis Support Syst 29(1):31–45. https://doi.org/10.1016/S0167-9236(00)00061-0 Tsiriga V, Virvou M (2004) Evaluating the intelligent features of a web-based intelligent computer assisted language learning system. Int J Artif Intell Tools 13(02):411–425. https://doi.org/10.1142/S0218213004001600 VanLehn K (2011) The Relative Effectiveness of Human Tutoring, Intelligent Tutoring Systems, and Other Tutoring Systems. Educational Psychol 46(4):197–221. https://doi.org/10.1080/00461520.2011.611369 VanLehn K, Graesser AC, Jackson GT, Jordan P, Olney A, Rosé CP (2007) Than Reading? Cogn Sci 31(1):3–62. https://doi.org/10.1080/03640210709336984 . When Are Tutorial Dialogues More Effective VanLehn K, Jordan PW, Rosé CP, Bhembe D, Böttner M, Gaydos A, Makatchev M, Pappuswamy U, Ringenberg M, Roque A, Siler S, Srivastava R (2002) The Architecture of Why2-Atlas: A Coach for Qualitative Physics Essay Writing. In S. A. Cerri, G. Gouardères, & F. Paraguaçu (Eds.), Intelligent Tutoring Systems (pp. 158–167). Springer. https://doi.org/10.1007/3-540-47987-2_20 Vides Ceron F (2012) TAYouKi: A Sketch-Based Tutoring System for Young Kids Wesely S, Hofer E, Curth R, Paryani S, Mills N, Ueberschär O, Westermayr J (2025) Artificial intelligence for objective assessment of acrobatic movements: How to apply machine learning for identifying tumbling elements in cheer sports (No. arXiv:2503.04764). arXiv. https://doi.org/10.48550/arXiv.2503.04764 Williford B SketchTivity: Improving Creativity by Learning Sketching with an Intelligent Tutoring System. Proceedings of the 2017 ACM SIGCHI Conference on Creativity and, Cognition (2017) 477–483. https://doi.org/10.1145/3059454.3078695 Williford B, Runyon M, Li W, Linsey J, Hammond T (2020) Exploring the potential of an intelligent tutoring system for sketching fundamentals. Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, 1–13 Yang C, Chiang F-K, Cheng Q, Ji J (2021) Machine Learning-Based Student Modeling Methodology for Intelligent Tutoring Systems. J Educational Comput Res 59(6):1015–1035. https://doi.org/10.1177/0735633120986256 Yuksel BF, Oleson KB, Harrison L, Peck EM, Afergan D, Chang R, Jacob RJ (2016) Learn piano with BACh: An adaptive learning interface that adjusts task difficulty based on brain state. Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, 5372–5384 Zhang X, Huang Y, Zou Q, Pei Y, Zhang R, Wang S (2020) A Hybrid convolutional neural network for sketch recognition. Pattern Recognit Lett 130:73–82. https://doi.org/10.1016/j.patrec.2019.01.006 Zhou M, Gong T (2020) Optimization of multimedia computer-aided interaction system of vocal music teaching based on voice recognition. Comput-Aided Des Appl 8:113–122 Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8797891","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Systematic Review","associatedPublications":[],"authors":[{"id":586406187,"identity":"3270b4da-479e-4530-bbbd-c2cb910a46c6","order_by":0,"name":"Lucas Lörch","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAwElEQVRIiWNgGAWjYBACPgbmBhCdwM+Q2MBQwcBMWAsbAyNEi2QDUMsZkrQYHEhgIFpL44OPe+zyjI8ntz04wGAtR4yWZsMZz5KLzc48bDc4wJBuTIyWNmmeAwcSt91IbJP+wHA4sYEoLX+AWjbPSGyTOMBwuJ44LQxALRskIFoSCDuMGeiXngPJiTPAfjFINyRoCz9788EHPw7YJfa3pz97cKDCWp6gLcgRwcbAYEBYA6ojSVQ/CkbBKBgFIwUAANQDPud3ylCoAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0002-5665-3313","institution":"DIPF | Leibniz Institute for Research and Information in Education","correspondingAuthor":true,"prefix":"","firstName":"Lucas","middleName":"","lastName":"Lörch","suffix":""},{"id":586406370,"identity":"88912176-04c1-472c-b759-8f6ade3193f8","order_by":1,"name":"Garvin Brod","email":"","orcid":"https://orcid.org/0000-0002-7976-5609","institution":"DIPF | Leibniz Institute for Research and Information in Education","correspondingAuthor":false,"prefix":"","firstName":"Garvin","middleName":"","lastName":"Brod","suffix":""}],"badges":[],"createdAt":"2026-02-05 13:55:01","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-8797891/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8797891/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":102021555,"identity":"67ca5f3f-0cb1-46d0-aad2-223c8e941ec8","added_by":"auto","created_at":"2026-02-06 08:51:55","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":219746,"visible":true,"origin":"","legend":"\u003cp\u003ePrism flow chart diagram of the screening procedure.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8797891/v1/16e5c8df6647d7f79ac5ef45.png"},{"id":102295411,"identity":"f7eff7f5-48ed-4042-a6df-ea2bfcc15225","added_by":"auto","created_at":"2026-02-10 10:11:03","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1072846,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8797891/v1/395a7e8e-390d-4914-97d4-dbdf8d953988.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eIntelligent Tutoring Systems in Open-ended, Multimodal Domains Reviewing the Evidence from Arts, Music, and Sports\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"Highlights","content":"\u003cp\u003e\u0026bull; We reviewed intelligent tutoring systems in arts, music, and sports.\u003c/p\u003e\u003cp\u003e\u0026bull; We found no systems that address open-ended, creative skills.\u003c/p\u003e\u003cp\u003e \u0026bull; Most of the reviewed systems were focused on providing individualized feedback.\u003c/p\u003e\u003cp\u003e\u0026bull; Some systems individualized task selection by mapping component skills to tasks.\u003c/p\u003e\u003cp\u003e\u0026bull; Innovative hardware was used to measure or present body movements.\u003c/p\u003e"},{"header":"Introduction","content":"\u003cp\u003eIntelligent Tutoring Systems (ITS) are computer-assisted instruction systems that adapt their teaching strategies to individual students\u0026nbsp;(Corbett et al., 1997). The effectiveness of ITS has been demonstrated in multiple meta-analyses\u0026nbsp;(Ma et al., 2014; Steenbergen-Hu \u0026amp; Cooper, 2014; Tamim et al., 2011; VanLehn, 2011). ITS typically select tasks for individual learners based on their previous performance\u0026nbsp;(Corbalan et al., 2006). Learners can decide the time and pace at which they practicing\u0026nbsp;(Martin et al., 2007)\u0026nbsp;and they receive immediate response-specific feedback\u0026nbsp;(Azevedo \u0026amp; Bernard, 1995; Sosa et al., 2011). While such individualized task selection and feedback are cumbersome and time-intensive in human-led tutoring, they come with no extra costs in ITS. Given the widespread availability of computers in Western societies, ITS can thus play an important role in today\u0026apos;s education.\u003c/p\u003e\n\u003cp\u003eImplementing ITS is relatively straight-forward in well-formalized, text-based domains (Holland, 2000). Accordingly, ITS have been especially popular in domains such as algebra (Anderson et al., 1995; Koedinger et al., 1997; Ritter et al., 2007), physics (Albacete \u0026amp; VanLehn, 2000; VanLehn et al., 2002, 2007), and language (McNamara et al., 2004; Rowley et al., 1998; Tsiriga \u0026amp; Virvou, 2004). In rather open-ended, multimodal domains such as arts, music, and sports, the implementation of ITS is rather challenging (Phon-Amnuaisuk \u0026amp; Siong, 2008). However, the rise of artificial intelligence (AI) and other technologies, such as motion sensors and virtual reality, has opened new possibilities for ITS in these domains. In the present paper, we review ITS in the domains of arts, music, and sports, focusing on how they address these domains\u0026rsquo; characteristic challenges. We emphasize the role of AI and other technological advances in existing ITS and provide impulses how such technologies may be used in future systems.\u003c/p\u003e\n\u003ch3\u003eWhat do we mean by ITS?\u003c/h3\u003e\n\u003cp\u003eWe define ITS as computer systems that perform tutoring functions, such as presenting information, posing tasks, and providing feedback, and that automatically individualize at least one of these functions based on a model of the student (Ma et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). According to the dynamic framework of personalized education (Tetzlaff et al., \u003cspan citationid=\"CR83\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), the personalization of instruction occurs on three levels. At the macro level, an individual learning goal is selected and its achievement is measured by summative assessment. At the meso level, a task that fits the learner\u0026rsquo;s prerequisites is selected and the learning progress is measured by formative assessment. On the micro level, assistance is selectively provided or withheld during task completion based on an on-line assessment of the learner\u0026rsquo;s performance.\u003c/p\u003e \u003cp\u003eIn ITS, instruction is typically individualized via three system components: the domain model (also called the expert model), the student model, and the tutor model (Ma et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2014\u003c/span\u003e; Sottilare et al., \u003cspan citationid=\"CR79\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). The domain model represents the knowledge that the student is intended to learn. It contains the relevant skills and pieces of knowledge in a given domain. The student model represents the student\u0026rsquo;s current knowledge, containing a subset of the skills or knowledge in the domain model. The tutor model represents the instructional strategies. For example, it may contain mechanisms for updating the student model and suggesting tasks. Once a student has completed a task with sufficient accuracy, the tutor model may add the associated skills or pieces of knowledge to the student model (Ma et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). When suggesting new tasks, the tutoring model may target a skill not yet included in the student model (Phon-Amnuaisuk \u0026amp; Siong, \u003cspan citationid=\"CR67\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). Through these mechanisms, the domain, student, and tutoring models collaboratively steer the individualization of instruction.\u003c/p\u003e \u003cp\u003eBesides the domain, student, and tutor models, another conceptual component of ITS is the system interface (Ma et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). The interface enables the basic system-learner interaction by receiving information input from the learners and presenting information output from the system. However, the interface is also a highly relevant aspect of the instructional design. The hardware components of the interface, such as a computer screen and mouse, determine what information can be received and presented. These restrictions affect the system-learner interaction and thus the tutoring functions. The software components of the interface, such as the graphical user interface, guide users\u0026rsquo; attention, determine their actions, and influence the clarity, content, and timing of the tasks. In summary, both the hardware and software components are crucial for the instructional design of ITS.\u003c/p\u003e\n\u003ch3\u003eWhat’s special about arts, music, and sports?\u003c/h3\u003e\n\u003cp\u003eWe argue that there are two characteristics of arts, music, and sports that make them challenging domains for ITS. First, their subdomains lie on a continuum between rather well-formalized (e.g., sketching, music theory, yoga) and rather open-ended (e.g., abstract painting, music composition, team sports). Open-ended subdomains are especially challenging, because tasks have a large number of possible solutions and lack clear performance benchmarks. This raises the question of how instruction can be individualized in these open-ended subdomains. If there are myriad possible solutions, how should the system select a learning goal? If there are no clear performance benchmarks, how should the system determine whether a task has been completed or whether a skill can be added to the student model?\u003c/p\u003e \u003cp\u003eSecond, arts, music, and sports are multimodal domains. They involve multimodal information, such as the sound of a musical instrument, a sketched image, or the swing of a bat. Therefore, the system interface must be able to receive, process, and present these types of information. While this has long been expensive or impossible, recent technological advances have brought about many new types of interfaces, such as motion sensors, virtual reality, and AI agents. However, it remains unclear how ITS in multimodal domains use and combine innovative technologies and how technology use is guided by instructional design considerations.\u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003ePresent Study\u003c/h2\u003e \u003cp\u003eThe present work presents a systematic literature review of ITS in arts, music, and sports. An initial reading of the literature revealed that ITS studies in these fields were methodologically diverse and often lacked an effectiveness test. Therefore, we decided to focus solely on how these systems function and on the innovative technologies they use rather than on their efficacy. Consequently, we conducted a narrative review (Baumeister \u0026amp; Leary, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e1997\u003c/span\u003e). Such reviews synthesize the results of a set of studies by identifying overarching topics or reinterpreting them, but they do not analyze the statistical significance of the findings (Siddaway et al., \u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Our analyses were guided by two main research questions. First, how do ITS in arts, music, and sports individualize instruction given the open-ended nature of tasks? Second, how do ITS in arts, music, and sports leverage modern technologies to receive and process relevant multimodal information? Our goal was thus to provide a systematic overview of existing ITS in these uncommon domains to inform and inspire researchers and practitioners in these fields, as well as ITS researchers in other fields.\u003c/p\u003e \u003c/div\u003e"},{"header":"Method","content":"\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003eSearch string\u003c/h2\u003e \u003cp\u003eThe general structure of the search string in our systematic literature search was to combine the term \u0026ldquo;intelligent tutoring system\u0026rdquo; with the terms \u0026ldquo;arts,\u0026rdquo; \u0026ldquo;music,\u0026rdquo; and \u0026ldquo;sports.\u0026rdquo; We searched previous reviews on the topic for synonyms of the term \"intelligent tutoring system.\" We found a suitable list of synonyms in Steenbergen-Hu \u0026amp; Cooper (\u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e2014\u003c/span\u003e): artificial tutor*, computer tutor*, computer-assisted tutor, computer-based tutor*, intelligent learning environment*, computer coach*, online-tutor*, keyboard tutor*, e-tutor*, electronic tutor*, web-based tutor*. We decided not to use the terms \u0026ldquo;keyboard tutor\u0026rdquo; and \u0026ldquo;artificial tutor\u0026rdquo; as an initial literature search revealed that they are not common terms for digital learning systems. To derive our search string, we then removed the truncation symbols and placed the individual terms in quotation marks. This was done to avoid searching for overly general terms such as \u0026ldquo;computer,\u0026rdquo; \u0026ldquo;online,\u0026rdquo; or \u0026ldquo;intelligent.\u0026rdquo;\u003c/p\u003e \u003cp\u003eMoreover, we searched the APA Thesaurus of psychological index terms and the Education Thesaurus for synonyms for the terms \u0026ldquo;arts,\u0026rdquo; \u0026ldquo;music,\u0026rdquo; and \u0026ldquo;sports\u0026rdquo;. This search yielded only specific activities as synonyms (e.g., \"exercise\" for sports or \"improvisation\" for music). Especially in the case of sports and music, there are many specific activities associated with the broader terms. Using these activities as synonyms would have resulted in an overly inclusive search string or an unwarranted focus on single activities. Therefore, we decided not to include any synonyms for the terms \u0026ldquo;arts,\u0026rdquo; \u0026ldquo;music,\u0026rdquo; and \u0026ldquo;sports.\u0026rdquo;\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSearch string used in each database. The first line shows the string that was used as a root in all databases. Google Scholar does not allow for parenthesis. Google Scholar and ERIC do not allow for truncation symbols (also called wildcard operators).\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDatabase\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSearch string\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026ldquo;intelligent tutoring system\u0026rdquo; OR \u0026ldquo;AI tutor\u0026rdquo; OR \u0026ldquo;computer tutor\u0026rdquo; OR \u0026ldquo;computer-assisted tutor\u0026rdquo; OR \u0026ldquo;computer-based tutor\u0026rdquo; OR \u0026ldquo;computer-based tutoring system\u0026rdquo; OR \u0026ldquo;intelligent learning environment\u0026rdquo; OR \u0026ldquo;computer coach\u0026rdquo; OR \u0026ldquo;online tutor\u0026rdquo; OR \u0026ldquo;e-tutor\u0026rdquo; OR \u0026ldquo;web-based tutor\u0026rdquo;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGoogle Scholar\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAND \u0026ldquo;art\u0026rdquo; OR \u0026ldquo;arts\u0026rdquo;\u003c/p\u003e \u003cp\u003eAND sport\u003c/p\u003e \u003cp\u003eAND music\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAPA PsycInfo, APA PsycArticles, Education Research Complete,\u003c/p\u003e \u003cp\u003eThe New Republic Archive (DFG),\u003c/p\u003e \u003cp\u003eProQuest Dissertation \u0026amp; Thesis Global\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAND (arts OR music* OR sport*)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eERIC\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAND (arts OR music OR sports)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePSYNDEX\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAND (Kunst OR arts OR musi* OR sport)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eFinally, the search string was modified to account for differences in database functionality. For example, Google Scholar does not allow parentheses. Thus, we performed three searches, combining \u0026ldquo;intelligent tutoring system\u0026rdquo; and its synonyms with each of the three terms \u0026bdquo;arts,\u0026ldquo; \u0026bdquo;music,\u0026ldquo; and \u0026bdquo;sports\u0026ldquo; individually. In all other databases, we used parentheses and Boolean operators to combine the terms. For databases that allowed truncation symbols, we used \u0026bdquo;music*\u0026ldquo; and \u0026bdquo;sport*\u0026ldquo; as search terms to find results for \u0026bdquo;musical,\u0026ldquo; \u0026bdquo;musician,\u0026ldquo; \u0026bdquo;sportive,\u0026ldquo; \u0026bdquo;sports,\u0026ldquo; etc. We refrained from using \u0026bdquo;art*\u0026ldquo; as a search term to avoid finding results related to artificial intelligence. PSYNDEX also contains German-language literature. Thus, we adapted the search string to cover the German terms for arts, music, and sports (Kunst, Musik, and Sport, respectively). The final search strings used in each database can be found in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eInclusion criteria\u003c/h3\u003e\n\u003cp\u003eOur inclusion criterion was that the study in question described a tutoring system in arts, music, or sports in sufficient detail to understand how the system functions, and was written in English or German. It was not an inclusion criterion that studies performed a controlled test of the effect of the system in comparison to a human tutor. Our focus was on the systems\u0026rsquo; functioning and use of innovative technologies, not on their efficacy. We excluded papers that only described how ITS in arts, music, or sports could be designed in general (\u0026ldquo;unspecific\u0026rdquo;), did not explain the ITS in sufficient detail (\u0026ldquo;unclear\u0026rdquo;), referred to a different type of system or to an ITS in a different domain than arts, music, or sports (\u0026ldquo;off-topic\u0026rdquo;), were literature reviews themselves (\u0026ldquo;review\u0026rdquo;), described merely a technology that could be used in an ITS (\u0026ldquo;technology\u0026rdquo;), or described a commercial, closed-source system (\u0026ldquo;commercial\u0026rdquo;). Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e shows how many studies were excluded due to each criterion.\u003c/p\u003e\n\u003ch3\u003eScreening procedure\u003c/h3\u003e\n\u003cp\u003eSince Google Scholar does not allow for the bulk export of metadata, the initial screening was performed within the web browser. Both reviewers performed the search independently, reading titles (and abstracts, if necessary) and adding items to their personal libraries if they deemed them relevant. They continued the search until they reached a maximum of 1,000 results or until they found no relevant items among 100 consecutive results. After completing the screening, both reviewers exported the metadata from their Google Scholar personal libraries and imported it into Zotero. Then, they searched the other databases, exported the metadata of all the results, and imported them into Zotero. Duplicates were removed, and the remaining results were screened in Zotero. Next, the first author screened all results considered relevant by the second reviewer. Subsequently, the first author sought to retrieve all relevant articles. If an article was not accessible, the first author requested the full text via ResearchGate or contacted the authors via email. Finally, the first author performed an in-depth screening of all 146 papers selected during the initial screening by skimming or reading their main texts. Figure\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e shows the Prism flowchart of the entire screening process.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eIndividualization of instruction\u003c/h2\u003e \u003cdiv id=\"Sec10\" class=\"Section3\"\u003e \u003ch2\u003eOpen-endedness of the subdomains\u003c/h2\u003e \u003cp\u003eOur first research question was how ITS in arts, music, and sports individualize instruction, given the open-ended nature of these domains. To address this question, we first analyzed the domains and subdomains of the reviewed ITS. As mentioned earlier, the subdomains of arts, music, and sports lie on a continuum ranging from rather well-formalized to rather open-ended. Thus, we analyzed which subdomains the reviewed ITS address. Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e lists all of the reviewed ITS and their respective domains and sub-domains. Note that some of the 39 publications resulting from the systematic literature search refer to the same ITS. Moreover, for each ITS that was described in a publication, we searched for additional literature that was not yet included in our dataset. This resulted in 29 ITS and 42 associated publications.\u003c/p\u003e \u003cp\u003eWith a look on the domain and subdomain column in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, it becomes clear that the reviewed ITS avoid overly open-ended subdomains. The most common subdomains, musical instrument performance and music theory, are rather well-formalized than open-ended as a single correct solution can be defined. While more advanced levels of instrument performance allow for some freedom of musical expression, the fundamentals of playing an instrument involve producing the correct tones at the right time. Music theory is governed by clear, logical, and often even mathematical rules. In addition, the four reviewed sports ITS and the two reviewed arts ITS addressed well-formalized subdomains as well. In dancing, yoga, and strength training, there is typically one desired way to perform movements and poses. Sketching involves creating a realistic copy of a model, and there are clear benchmarks for a good sketch.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eThe ITS that were part of the present review together with the publications that presented them, their domains and subdomains.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eName of ITS\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePublications\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDomain\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSubdomain\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePiano Tutor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Capell \u0026amp; Dannenberg, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e1993\u003c/span\u003e; Dannenberg et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e1990\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e1993\u003c/span\u003e; Sanchez et al., \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e1987\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInstrument performance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBoland Piano Tutor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Boland, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2003\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInstrument performance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDigital Violine Tutor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Percival et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2007\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInstrument performance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMEAWS (Musician Evaluation and Audition for Winds and Strings)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Percival et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2007\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInstrument performance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eiDVT (interactive digital violin tutor)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Huanhuan, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2009\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInstrument performance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBACh (Brain Automated Chorales)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Yuksel et al., \u003cspan citationid=\"CR94\" class=\"CitationRef\"\u003e2016\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInstrument performance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eguitMaster\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Grigutis, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2018\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInstrument performance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003einstruMentor\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Bagga et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2019\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInstrument performance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAffective Cognitive Tutor for Rhythm\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Sanz, \u003cspan citationid=\"CR75\" class=\"CitationRef\"\u003e2022\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInstrument performance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInteractive Rainbow Score\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Chin et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Chin \u0026amp; Xia, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInstrument performance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePianobot\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Della Ventura, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2022\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInstrument performance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHoloMusic XP\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Molero et al., \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInstrument performance\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIMUTUS (Interactive Music Tuition System)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Fober et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2004\u003c/span\u003e; Raptis et al., \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Schoonderwaldt et al., \u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e2005\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInstrument performance \u0026amp; music theory\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eKANT (Kritical Argument Negotiated Tutoring)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(M. Baker, \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e1990\u003c/span\u003e; M. J. Baker, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e1992\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMusic theory\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSonata\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Angelides \u0026amp; Tong, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1995\u003c/span\u003e; C Angelides \u0026amp; KY Tong, 1994; Tong \u0026amp; Angelides, \u003cspan citationid=\"CR84\" class=\"CitationRef\"\u003e2000\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMusic theory\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChordTeacher\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Greiner, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2011\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMusic theory\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChat Melody\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Jin et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2025\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMusic theory\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMaestoso\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Taele et al., \u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e2015\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMusic theory\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003emySolfeggio\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Debevc et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2020\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eMusic theory\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFEEL-ME (Feedback Learning of Musical Expressivity)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Juslin et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2006\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eEmotional expressivity\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eImproScales\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Borgogno \u0026amp; Turchet, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2022\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eImprovisation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRaagang\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Gajjar \u0026amp; Patel, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2020\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMusic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eImprovisation\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eY-system\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Chen et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2018\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSports\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYoga\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ee-YogaGuru\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Kale et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2021\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSports\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eYoga\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDancing Coach\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Romano et al., \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2019\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSports\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDancing\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSelfit\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Guarnieri et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; L. Neagu, \u003cspan citationid=\"CR59\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; L.-M. Neagu et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; L.-M. Neagu, Rigaud, Guarnieri, Dascalu, et al., 2022)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSports\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eExercise\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSketchTivity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Hammond et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Keshavabhotla et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Williford, \u003cspan citationid=\"CR91\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Williford et al., \u003cspan citationid=\"CR92\" class=\"CitationRef\"\u003e2020\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eArts\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSketching\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTAYouKi (Tutoring Assistant for Young Kids)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e(Vides Ceron, \u003cspan citationid=\"CR89\" class=\"CitationRef\"\u003e2012\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eArts\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSketching\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eWe found only three ITS in rather open-ended subdomains, namely improvisation and emotional expression. In ImproScales (Borgogno \u0026amp; Turchet, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) and Raagang (Gajjar \u0026amp; Patel, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), users can practice improvising melodies in the Western and in the Hindustani tonal system, respectively. However, a closer look reveals that these ITS also address the well-formalized aspects of this creative activity. In both Western and Hindustani musical traditions, the tones that are used during improvisation are restricted to a certain musical scale. That is, there is a set of \u0026ldquo;allowed\u0026rdquo; tones that can be used during improvisation in a given song. Both ImproScales and Raagang address the skill of sticking to the tones of the musical scale during improvisation. In the ImproScales system, users can improvise either freely or with a backing track from Youtube. They can indicate the employed scale or the system can automatically detect the scale of the backing track. The system then provides feedback on whether the performed notes are part of the scale. Similarly, in Raagang, users can select a Hindustani scale (Raga), upload an audio recording of their improvisation in that scale, and the system provides feedback on which notes of the improvisation were from the selected scale. Thus, both ImproScales and Raagang exclusively address a well-formalized aspect of improvisation.\u003c/p\u003e \u003cp\u003eAnother seemingly open-ended activity supported by the Feel-ME system (Juslin et al., \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2006\u003c/span\u003e) is emotional expression in musical instrument performance. However, even this rather vague and subjective aspect of musical performance is broken down into objective rules. The basic idea behind the system is that musicians use musical parameters, such as tempo or loudness, to express emotions, and a certain usage of these parameters trigger emotions in listeners. For emotional expression to be successful, musicians should use musical parameters in a way that triggers the desired emotions in listeners. Formally, this means that the correlation between musical parameters and emotions should be similar for the performance of a musician and the evaluations of the listeners. Based on these ideas, the Feel-ME system uses correlations between musical parameters (tempo, timbre, articulation) and emotions from a training dataset in which listeners evaluated the emotional expression in a large set of musical performances. Users who want to practice their emotional expression with Feel-ME, they perform several melodies and indicate the emotion they intended to express with each performance. Using the data from all the performances, the program then calculates a point-biserial correlation between the user\u0026rsquo;s intention to express an emotion (0/1) and the musical parameters (tempo, sound level, articulation, timbre). It further calculates the fit between the correlations in the user\u0026rsquo;s performance and the correlations in the training dataset. Thereby, the program is able to provide feedback such as, \u0026ldquo;If you want to express sadness, you should play more legato.\u0026rdquo; It becomes clear that, although emotional musical expression seems to be an open-ended domain, the system is based on well-formalized, statistical rules. In summary, this first step of analysis revealed that none of the reviewed ITS address truly open-ended and creative activities such as expressive painting, musical composition, or decision making in team sports. This is a clear gap in the ITS landscape.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eIndividualized feedback and human-led task selection\u003c/h2\u003e \u003cp\u003eHowever, individualizing instruction can be challenging in well-formalized domains as well. Activities such as playing a musical instrument, sketching, and strength training are complex and multifaceted. Thus, the next step in our analyses was to investigated how instruction was individualized in the reviewed ITS. We found that several ITS did not individualize the instructional activity or content but merely provided feedback for a fixed set of tasks. The flute tutoring system Interactive Rainbow Score (Chin et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Chin \u0026amp; Xia, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) provides performance feedback for 16 folk songs. The yoga tutors Y-system (Chen et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) and e-YogaGuru (Kale et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) both contain a set of built-in yoga poses. Users can perform one of these yoga poses in front of a camera and receive feedback on their posture. The Dance Coach (Romano et al., \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) provides feedback on basic salsa steps.\u003c/p\u003e \u003cp\u003eIn addition to these feedback-based ITS, several systems circumvented the challenge of automated task selection by having learners or a human tutors steer the individualization of the instruction. In MEAWS (Percival et al., \u003cspan citationid=\"CR66\" class=\"CitationRef\"\u003e2007\u003c/span\u003e), Maestoso (Taele et al., \u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), and mySolfeggio (Debevc et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), teachers or students can create their own lessons. In HoloMusic XP (Molero et al., \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), students must select a lesson from a predetermined set. Likewise, in the sketch tutoring system SketchTivity (Hammond et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Keshavabhotla et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Williford, \u003cspan citationid=\"CR91\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Williford et al., \u003cspan citationid=\"CR92\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), exercises have a fixed sequence of increasing difficulty, and learners can decide whether to retry an exercise or proceed to the next one. A rather advanced way of combining human and virtual tutoring is implemented in the Boland piano tutor (Boland, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2003\u003c/span\u003e). A human tutor can create a lesson plan with different objectives and associated success criteria. For instance, a human tutor may want a student to learn to play a song at a fast tempo. Then, the teacher can define a sequence of lessons, each consisting of a certain tempo and a target accuracy. Once the student performs the song at the given tempo with sufficient accuracy, they can proceed to the next lesson, which increases the tempo.\u003c/p\u003e \u003cp\u003eSome of the reviewed ITS systems even put responsibility for aspects other than lesson content in the hands of students. In the flute tutoring system IMUTUS (Fober et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2004\u003c/span\u003e; Raptis et al., \u003cspan citationid=\"CR68\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Schoonderwaldt et al., \u003cspan citationid=\"CR76\" class=\"CitationRef\"\u003e2005\u003c/span\u003e), for example, students must decide how to proceed after receiving feedback on their performance from the system. They can choose to get more information about the error, repeat the phrase containing the error, complete a special exercise addressing the error, or repeat the entire piece. Similarly, learners using the instruMentor (Bagga et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), a robot flute tutor, need to decide on their own about the practice mode: playing alone, playing together with the robot, or observing the robot to play. In summary, many of the reviewed ITS lack automated individualized task selection based on a student model. Instead, these systems provide individualized feedback or put task selection in the hands of learners or human tutors.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eAutomatic individualized task selection\u003c/h2\u003e \u003cp\u003eHowever, we also found some systems with automatic individualized task selection. Two of the reviewed systems automatically adapted task difficulty to individual learners. In the guitMaster system (Grigutis, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2018\u003c/span\u003e), users can practice to perform unknown melodies on the guitar at first sight. The system generates the melodies and analyzes the accuracy. The difficulty of the melodies is defined by the number of strings used in combination with the performance tempo. Once an accuracy level of at least 70% is achieved, users are instructed to increase the difficulty. If the accuracy is below 50%, the system instructs users to decrease the difficulty. In the BACh system (Yuksel et al., \u003cspan citationid=\"CR94\" class=\"CitationRef\"\u003e2016\u003c/span\u003e) in which users practice to perform Bach chorales on the piano, the difficulty of the next chorale increases if cognitive load is low. Cognitive load is measured using functional near-infrared spectroscopy (fNIRS).\u003c/p\u003e \u003cp\u003eThe classical instrument tutoring system Piano Tutor (Capell \u0026amp; Dannenberg, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e1993\u003c/span\u003e; Dannenberg et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e1990\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e1993\u003c/span\u003e; Sanchez et al., \u003cspan citationid=\"CR73\" class=\"CitationRef\"\u003e1987\u003c/span\u003e) is based on an explicitly defined expert, student, and tutor model. The expert model consists of a set of lessons, i.e., single instructional activities that teach a new concept or skill. For example, a lesson may aim to teach students how to perform a new rhythm, such as a 3/4 measure (see (Capell \u0026amp; Dannenberg, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e1993\u003c/span\u003e). Once a learner completes this lesson, the associated skill is added to their student model. Each lesson has prerequisite skills. The system can only suggest a lesson to a learner if the prerequisite skill is already part of their student model. For example, to complete the 3/4 rhythm lesson, a learner must have mastered the skills to detect and perform different notes and to perform the 2/4 rhythm.\u003c/p\u003e \u003cp\u003eThe ChordTeacher system (Greiner, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2011\u003c/span\u003e) is based on a similar principle. The expert model consists of the theory of musical chords. It defines root notes, musical intervals, and the structure of different types of chords. It also explains how chords can be displayed, such as on piano keys, as staff notation, or in a Harmony Space matrix. There are various tasks that involve building or recognizing intervals or chords. Once a learner completes a task, their proficiency score increases for the interval, chord type, and representation involved in the task. Each task has a prerequisite proficiency score and can only be selected by the system once the learner has reached it.\u003c/p\u003e \u003cp\u003eAnother system with a sophisticated expert, student, and tutor model is the Selfit strength training system in sports (Guarnieri et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2023\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; L.-M. Neagu et al., \u003cspan citationid=\"CR62\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; L.-M. Neagu, Rigaud, Guarnieri, Radu, et al., \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The authors first developed the OntoStrength ontology to serve as a basis for the system (L.-M. Neagu, Rigaud, Guarnieri, Radu, et al., \u003cspan citationid=\"CR61\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). This ontology contains formal definitions of terms and concepts, as well as their relations within the domain of strength training. One subdomain of the ontology describes strength skills by combining specific muscles with contraction modes (eccentric, concentric, isometric, plyometric) and strength properties (power, maximum strength, endurance). A skill in this subdomain may be, for example, \u0026ldquo;Biceps Eccentric Maximum Strength.\u0026rdquo; Each strength skill is associated with a movement that trains it. Another subdomain of the ontology describes training programs by defining templates of training cycles and exercise blocks. Selfit\u0026rsquo;s domain model then contains more than 1,000 exercises, which are defined by a number of parameters such as the movements or number of joints involved. The student model contains general information about the trainee, such as age, weight, and sex, as well as self-reports on specific events, such as pain, injury, or surgery, and physical condition, such as fatigue level, motivation, sleep quality, and stress level. In addition, after each exercise, users must indicate how many more repetitions they could perform (repetitions in reserve). This information is also added to the student model and is used during task selection by a machine learning algorithm, which we will describe in greater detail in the next section. At the beginning of a training session, the trainee can select which muscle to train. Using the OntoStrength ontology, the system first derives which movements train the selected muscle. Then, based on the domain model, it derives which exercises involve these movements. Considering the information in the student model and the training templates in the ontology, the system then creates an individualized training plan.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eUsage of modern technologies\u003c/h2\u003e \u003cp\u003eOur second research question was how ITS in arts, music, and sports leverage modern technologies to accommodate their inherently multimodal nature. We distinguished three classes of technologies in our analyses: input technologies for capturing user behavior (e.g., sensors and detectors), data-processing technologies for interpreting this input (e.g., machine learning algorithms), and output technologies for delivering feedback and instruction (e.g., virtual reality headsets). Table\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e provides an overview of the technologies employed in the reviewed ITS and their purpose.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eOverview of the different types of individualization in the reviewed ITS.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIndividualization\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExplanation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eITS\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIndividualized feedback\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eThe system provides feedback for a fixed set of tasks. Instructional activities, explanations or contents are not individualized.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eY-system \u003c/p\u003e \u003cp\u003ee-YogaGuru\u003c/p\u003e \u003cp\u003eDance Coach\u003c/p\u003e \u003cp\u003eRainbow Score\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHuman-controlled individualization\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLearners or human tutors can decide about the instructional activities or contents.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eMEAWS \u003c/p\u003e \u003cp\u003eMaestoso \u003c/p\u003e \u003cp\u003emySolfeggio \u003c/p\u003e \u003cp\u003eHoloMusic XP \u003c/p\u003e \u003cp\u003eBoland piano tutor \u003c/p\u003e \u003cp\u003eIMUTUS\u003c/p\u003e \u003cp\u003einstruMentor\u003c/p\u003e \u003cp\u003eSketchTivity\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIndividualized task difficulty\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTask difficulty is adapted based on performance accuracy or cognitive load.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eguitMaster\u003c/p\u003e \u003cp\u003eBACh\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIndividualized task selection\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eEach task is associated with required skills and skills it targets. Targeted skills are added to the student model upon task completion. New tasks are selected whose required skills are already part of the student model and whose targeted skills are not.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePianoTutor\u003c/p\u003e \u003cp\u003eChordTeacher\u003c/p\u003e \u003cp\u003eSelfit\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eAnalyzing the input technologies, it became clear that most of the reviewed ITS relied on rather classical technologies, such as microphones, cameras and MIDI (Musical Instrument Digital Interface). However, some ITS indeed used more advanced technologies to receive multimodal input. The sketch tutoring systems SketchTivity (Williford, \u003cspan citationid=\"CR91\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) and TaYouKi (Vides Ceron, \u003cspan citationid=\"CR89\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), as well as the music theory tutoring system Maestoso (Taele et al., \u003cspan citationid=\"CR81\" class=\"CitationRef\"\u003e2015\u003c/span\u003e), receive user input via a digital stylus, also known as ink digitizer. With a digital stylus, users can write or draw on a computer screen or on a special pad and their strokes are transmitted directly to a computer. All three ITS all used digital styluses from Wacom (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e\u003ca href=\"http://www.wacom.com\" target=\"_blank\"\u003ewww.wacom.com\u003c/a\u003e\u003c/span\u003e\u003cspan address=\"http://www.wacom.com\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e), which measure pressure, tilt, and stroke speed using electromagnetic resonance and sensors. The advantage of using a digital stylus is that it enables real-time feedback on activities such as writing musical notation or drawing, while preserving the natural feeling of the activity. Its intuitive design allows allows even young children to use a digital stylus.\u003c/p\u003e \u003cp\u003eAnother input technology\u0026mdash;the Microsoft Kinect depth camera\u0026mdash;is used by the dance tutoring system Dancing Coach (Romano et al., \u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) and by the yoga tutoring systems Y-system (Chen et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) and e-YogaGuru (Kale et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Depth cameras can capture body poses or movements at a fine-grained level by combining cameras, infrared transmitters, and infrared detectors (D. Johnson et al., \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Mat Sanusi et al., \u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). In addition to tracking human bodies, depth camera software can create skeleton representations based on joint locations. For the Dancing Coach, Romano et al. (\u003cspan citationid=\"CR71\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) enabled the system to detect and evaluate basic salsa steps by providing recordings of these movements to the Kinect Visual Gesture Builder. For the Y-system, Chen et al. (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2018\u003c/span\u003e) created explicit posture description models for each of the twelve implemented yoga poses. For these posture models, reference points were defined on the body contours of each yoga pose, which were extracted from the images of the depth camera. Then, a correct posture was defined by defining angles between lines connecting some of the reference points. The e-YogaGuru uses skeleton representations from Kinect recordings of eight experts as references to provide feedback. In summary, depth cameras are an important technology for ITS in sports domains as they enable real-time evaluation of body movements.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eOverview of technologies employed in the reviewed ITS and their purpose\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTechnology\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePurpose\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eITS\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDigital stylus and sketch recognition\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eReceive and interpret hand-drawn sketches and hand-written musical notation\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTaYouKi\u003c/p\u003e \u003cp\u003eSketchTivity\u003c/p\u003e \u003cp\u003eMaestoso\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDepth camera and movement recognition\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eReceive and interpret body movements and postures\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eDancing Coach\u003c/p\u003e \u003cp\u003eY-system\u003c/p\u003e \u003cp\u003eeYogaGuru\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eReinforcement learning\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDetermine the most effective strength exercise\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSelfit\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLarge language model\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eReceive responses and questions and provide feedback and instruction in natural language\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eChat Melody\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRobots\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDemonstrate movements on instrument or provide motivating interaction\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eInstruMentor\u003c/p\u003e \u003cp\u003ePianobot\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAugmented reality\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eShow movement triggers on instrument\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHoloMusic XP\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eOur second category of technologies, data-processing technologies, mainly refers to software and algorithms, such as machine learning algorithms. The sketch tutoring systems TaYouKi, SketchTivity, and Maestoso use sketch recognition algorithms which enable computers to interpret hand-drawn sketches or writing. The reviewed systems use two types of these algorithms: geometric and pattern matching. Geometric algorithms, such as PaleoSketch (Paulson \u0026amp; Hammond, \u003cspan citationid=\"CR65\" class=\"CitationRef\"\u003e2008\u003c/span\u003e), derive basic shapes from the properties of a sketch. These properties may include the number of lines or the number of sudden changes in direction. Pattern matching algorithms require a correct reference sketch for comparison. First, the sketch is pre-processed to match the size and orientation of the reference. Then, equally spaced points are defined on the lines of the sketch and the reference, and a resemblance measure is calculated based on the distances between the points. The reviewed systems use these algorithms to recognize sketched objects or handwritten musical symbols and provide corrective feedback.\u003c/p\u003e \u003cp\u003eThe AI yoga tutor uses algorithms to automatically provide feedback. First, a large set of photos of yoga poses was collected, and the OpenPose algorithm was used to extract skeleton representations of the bodies in the photos. Then, three yoga experts rated the quality of the poses. The skeleton representations and the associated quality evaluations were provided to a machine learning algorithm. Thereby, the algorithm was trained to evaluate the quality of yoga poses in new photos.\u003c/p\u003e \u003cp\u003eBesides providing feedback, machine learning algorithms can be used to determine which instructional activities the system should suggest. The Selfit strength training tutor uses a special type of reinforcement learning called a multi-armed bandit algorithm. These algorithms address the problem of choosing among fixed options (in this case, exercises) when the properties of each option (in this case, their training effect) are only partially known. In simple terms, the algorithm is based on three principles: (1) each option is associated with a reward, (2) the reward associated with the chosen option is updated after a choice (i.e., increased if the choice led to a reward or decreased if the option did not lead to a reward), and (3) choices should maximize the overall reward. In Selfit, exercises are considered optimal if they are completed with a small number of repetitions in reserve. For example, if an exercise is planned with 12 repetitions, it is considered optimal if the trainee completes these 12 repetitions but indicates that they could not have performed any further ones. Thus, Selfit\u0026rsquo;s algoritgm uses the trainee\u0026rsquo;s self-reported repetitions in reserve to adjust the reward associated with an exercise. A small number of repetitions in reserve increases the reward, while a large number of repetitions in reserve, whether positive or negative, decreases the reward, indicating that the exercise was too difficult or too easy. The reward affects the probability that an exercise will be suggested again, optimizing the system\u0026rsquo;s suggestions.\u003c/p\u003e \u003cp\u003eAnother way of individualizing instructions is implemented in the music theory tutor Chat Melody (Jin et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The system uses a large language model (LLM) to enable chat-based interactions between the user and the system. The user is given a notated melody and is asked to perform some harmonic analysis on it. The user can respond and ask questions in the chat, and the system provides corrective feedback, hints, or reflective questions. While many aspects of the system remain unclear, Chat Melody offers an initial glimpse into the implementation of ITS in arts, music, and sports.\u003c/p\u003e \u003cp\u003eLastly, we found two types of output technologies: robots and augmented reality (AR). The flute tutoring system InstruMentor (Bagga et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) uses a robot with two humanoid hands that can play the flute. The robot can demonstrate the learner how to play a song. In contrast, the Pianobot (Ritschel et al., \u003cspan citationid=\"CR69\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) is a social robot that communicates with the user to provide feedback, hints, and advice. In doing so, the Pianobot uses gaze and facial expression. This interaction is intended to increase motivation to exercise between lessons with a human tutor. Lastly, the HolowMusic XP piano tutoring system (Molero et al., \u003cspan citationid=\"CR57\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) uses Microsoft Hololens augmented reality glasses to present graphics overlaying the surrounding environment. Through the glasses, users see note symbols falling down on their fingers on the piano, which they must \u0026ldquo;catch\u0026rdquo; by pushing the corresponding keys. While it is problematic that users of this system do not really learn to perform musical notation, the system exemplifies how AR can be used to provide instructions referencing external objects, such as musical instruments or sports equipment.\u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eITS play an increasingly important role in today\u0026rsquo;s education. Their four main conceptual components are the domain, student, and tutor models, which jointly steer the individualization of instruction, and the interface that enables and restricts the system-learner interaction (Ma et al., \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Arts, music, and sports are challenging domains for ITS due to their open-ended and multimodal nature. We conducted a systematic, narrative review (Baumeister \u0026amp; Leary, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e1997\u003c/span\u003e; Siddaway et al., \u003cspan citationid=\"CR77\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) of ITS in arts, music, and sports, analyzing how these systems address the challenges posed by these domains. Specifically, we analyzed how these systems individualize instruction in the face of open-ended tasks and use innovative technologies to receive, process, and present multimodal information.\u003c/p\u003e \u003cp\u003eRegarding the individualization of instruction, we found that ITS in arts, music, and sports avoid open-ended domains. They exclusively address complex, multifaceted skills in rather well-formalized domains. Moreover, automated, individualized task selection based on a student model\u0026mdash;often considered a hallmark feature of ITS (Steenbergen-Hu \u0026amp; Cooper, \u003cspan citationid=\"CR80\" class=\"CitationRef\"\u003e2014\u003c/span\u003e)\u0026mdash;was implemented in only a few systems. Instead, tasks had a fixed sequence or had to be selected manually. Systems that individualized instruction did so at the meso level of the dynamic framework of personalized education (Tetzlaff et al., \u003cspan citationid=\"CR83\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). That is, generally speaking, performance is assessed upon completion of an instructional unit, and the next unit is selected based on the learner\u0026rsquo;s individual learning prerequisites. Future systems should extend individualization to the overarching learning goal (macro level) and the assistance during tasks (micro level). Overall, ITS in arts, music, and sports seem to focus more on providing feedback than on individualized task selection.\u003c/p\u003e \u003cp\u003eRegarding the use of innovative technologies\u0026mdash;our second research question\u0026mdash;we found that ITS in arts, music, and sports use new types of interfaces, such as digital styluses, depth cameras, robots, and AR, to capture or display multimodal information. Machine learning algorithms are used to optimize individualized task selection, and chatbots to enable system-user interaction in natural language. A major focus of ITS in arts, music, and sports appears to be finding innovative and domain-specific ways to capture body movements. This comprises not only the movement of the body in itself but also its movement in relation to other objects, such as pens or musical instruments. Future systems should also consider further technologies such as special sensors that detect pressure on the fret of a string instrument (Grosshauser \u0026amp; Tr\u0026ouml;ster, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), the force, position, and tilt of fingers on piano keys (Grosshauser et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2012\u003c/span\u003e), or the pressure of trombone players\u0026rsquo; lips (Grosshauser et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2013\u003c/span\u003e), and suites (R. M. G. Johnson et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Lieberman \u0026amp; Breazeal, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2007\u003c/span\u003e) and exoskeletons (Moringen et al., \u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), which can guide movements through vibrotactile stimulation.\u003c/p\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003eUsing LLMs to support creative skills\u003c/h2\u003e \u003cp\u003eThe lack of ITS that support truly open-ended, creative skills is a gap in the ITS landscape that should be addressed by future research. Closing this gap requires an understanding of the relevant tutoring functions in such a context. Tutoring in open-ended domains is typically dialogical and encourages creative, metacognitive, and critical thinking rather than simply stating facts (Cook, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e1998\u003c/span\u003e). In the domain of musical composition, it has recently been found that teachers aim to cultivate originality and the clarity of artistic intentions (L\u0026ouml;rch \u0026amp; Huovinen, \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). To achieve this, teachers reported using students\u0026rsquo; own compositional work as the basis of their teaching, asking students self-reflective questions, challenging their compositional decisions, and providing new perspectives. How may these tutoring functions be implemented in ITS?\u003c/p\u003e \u003cp\u003eA promising approach is using chatbots based on large language models (LLM). Users could upload one of their artistic works. The chatbot could then compare the uploaded piece to works in the training dataset or found online. Thereby, the chatbot could evaluate the originality of the work and recommend composers or artists with similar approaches. The chatbot could also ask users about their artistic intentions, aesthetic ideals, and the reasons behind their artistic decisions. Users\u0026rsquo; responses to these questions could feed into a student model which guides subsequent interactions. For example, if a painter indicated to dislike strong colors, the chatbot may suggest that they look into the works of an artist who used strong colors to help them reflect on the origins of their aesthetic preference. While such a system cannot replace a human tutor, it could support users between regular lessons and encourage self-reflection and self-criticism.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eUsing machine learning to provide feedback for complex skills\u003c/h2\u003e \u003cp\u003eMany ITS in arts, music, and sports merely provide feedback rather than individualizing task selection or content. According to some more restrictive definitions, these systems would not even qualify as ITS. However, according to our definition of ITS which is based on Ma et al. (\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2014\u003c/span\u003e), these systems can indeed be considered ITS because they perform one tutoring function (providing feedback) in an individualized manner. In addition to the importance of feedback in computerized instruction in general (Azevedo \u0026amp; Bernard, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e1995\u003c/span\u003e), feedback may be especially important in arts, music, and sports, since the skills in these domains are typically complex and multifaceted. When performing a yoga pose, for example, many aspects are relevant, such as the overall body posture, the position of the limbs, and the angles between limbs. It can be difficult to monitor all these aspects during performance. Therefore, providing individualized feedback may be especially relevant for acquiring complex, multifaceted skills.\u003c/p\u003e \u003cp\u003eOur findings revealed the important role of machine learning algorithms (Campesato, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2020\u003c/span\u003e) in providing feedback in this context. These algorithms can detect patterns and regularities in large, complex datasets. To provide feedback on a complex, multifaceted skill, one can provide the algorithm with a training dataset containing recordings of expert performances (or of performances rated by experts). The algorithm can then provide feedback by comparing a learner\u0026rsquo;s performance with the patterns and regularities found in the training dataset (Nguyen et al., \u003cspan citationid=\"CR63\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Owusu, \u003cspan citationid=\"CR64\" class=\"CitationRef\"\u003e2007\u003c/span\u003e; Santos, \u003cspan citationid=\"CR74\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Wesely et al., \u003cspan citationid=\"CR90\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Yang et al., \u003cspan citationid=\"CR93\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis approach is especially powerful when combined with other algorithms that extract specific performance parameters. Examples include body and movement recognition algorithms (Chiang et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Hui, \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Ji, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Kotte et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Lin et al., \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), sketch recognition algorithms (Iarussi et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Mittal et al., \u003cspan citationid=\"CR56\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Zhang et al., \u003cspan citationid=\"CR95\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), and algorithms that detect pitches, chords, and harmonies (Brandao, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Della Ventura, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Jamshidi et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Zhou \u0026amp; Gong, \u003cspan citationid=\"CR96\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Incorporating the outputs of these algorithms into training datasets alongside expert ratings will enhance the specificity of the feedback. That is, the machine learning algorithm will not only be able to categorize new performances based on their quality but will also be able to point out specific problems in certain parameters.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003eDomain, student, and tutor models for complex skills\u003c/h2\u003e \u003cp\u003eSome ITS in arts, music, and sports contain sophisticated domain, student, and tutor models that enable individualized instruction for complex, multifaceted skills. After synthesizing the mechanisms of the reviewed ITS, we identified a general approach for individualized instruction in this context. This approach is based on the idea that subject didactics and pedagogical expertise enable breaking down a complex skill into component skills and to map these components to tasks. In this approach, the domain model contains all the component skills, and the student model contains the subset of these skills that the student has learned already. The tutor model contains the tasks that are implemented in the ITS, the benchmarks for completing them, their prerequisite and target skills, and mechanisms to update the student model and suggest new tasks. A typical mechanism for updating the student model is to increase proficiency in the skills associated with a task after a student completes it. A typical mechanism for suggesting new tasks is to suggest tasks whose prerequisite skills are part of the student model, but whose target skills are not. The strengths of this approach are its transparency and its clear grounding in subject didactics. Its weakness is that decomposing a complex skill into components and mapping them to tasks can be cumbersome and time-intensive.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eLimitations\u003c/h2\u003e \u003cp\u003eAlthough the present work offers valuable insights into ITS in arts, music, and sports, several limitations should be pointed out. During the screening process, it was sometimes difficult to distinguish studies that presented proper ITS from those that merely presented technologies that could be used for ITS. Since we considered the latter to be potentially relevant, we screened their content and included it in the discussion whenever possible. Moreover, many systems were not described in sufficient detail to understand how they functioned, and 18% of the potentially relevant papers were not available. Thus, promising approaches that may have been in these works were not included in this review. Many of the presented studies were published as theses or in conference proceedings and thus, it is unclear whether they were peer-reviewed.\u003c/p\u003e \u003cp\u003eLastly, our analytical approach\u0026mdash;the systematic, narrative review\u0026mdash;is inherently subjective. It does not involve any quantitative analyses. Instead, the content of the reviewed studies is summarized qualitatively, overarching topics and recurring themes are identified. This process naturally depends on the researchers and their background, posing the risk that important issues are overlooked. However, despite these weaknesses, we believe that a systematic, narrative review was the most appropriate approach for this small, methodologically diverse research field.\u003c/p\u003e \u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eITS in arts, music, and sports is a young and evolving research field. It profits from the current technological developments, such as VR, sensors, machine learning, and AI, as these developments enable computers to harness new types of information and open new possibilities for system-user interaction. Researchers have developed sophisticated methods for handling complex, multifaceted skills, including providing individualized feedback, combining human and computerized tutoring, and breaking down complex skills into components. However, one major challenge still lies ahead\u0026mdash;addressing truly open-ended, creative skills. We believe our work has provided important impulses for the field to finally face this challenge.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor Note\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eCorrespondence concerning this article should be addressed to Lucas L\u0026ouml;rch, DIPF - Leibniz Institute for Research and Information in Education, Rostocker Stra\u0026szlig;e 6, 60323 Frankfurt a. M.. Email: [email protected]. We want to thank Franziska Cremer for her help during the screening of the articles. This study was funded by the German Federal Ministry of Education and Research (BMBF). The views expressed in this study are those of the authors and may not necessarily reflect those of the funding institution. The study is part of the KuMuS-ProNeD project, which is one of eight project networks in the Music/Arts/Sports Competence Center of the lernen:digital Competence Network. The funding agency played no role in the design of the study, the analysis of the data, the reporting of the results, or the decision to submit the study for publication.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAlbacete P, VanLehn K (2000) Evaluating the Effectiveness of a Cognitive Tutor for Fundamental Physics Concepts\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnderson JR, Corbett AT, Koedinger KR, Pelletier R (1995) Cognitive Tutors: Lessons Learned. J Learn Sci 4(2):167\u0026ndash;207\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAngelides MC, Tong AKY (1994) Using hypertext to implement multiple tutoring strategies in an intelligent tutoring system for music learning. J Comput Inform Technol 2(2):123\u0026ndash;138\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAngelides MC, Tong AKY (1995) Implementing multiple tutoring strategies in an intelligent tutoring system for music learning. J Inform Technol 10(1):52\u0026ndash;62. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1057/jit.1995.7\u003c/span\u003e\u003cspan address=\"10.1057/jit.1995.7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAzevedo R, Bernard RM (1995) A Meta-Analysis of the Effects of Feedback in Computer-Based Instruction. J Educational Comput Res 13(2):111\u0026ndash;127. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2190/9LMD-3U28-3A0G-FTQT\u003c/span\u003e\u003cspan address=\"10.2190/9LMD-3U28-3A0G-FTQT\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBagga S, Maurer B, Miller T, Quinlan L, Silvestri L, Wells D, Winqvist R, Zolotas M, Demiris Y (2019) instruMentor: An Interactive Robot for Musical Instrument Tutoring. In K. Althoefer, J. Konstantinova, \u0026amp; K. Zhang (Eds.), Towards Autonomous Robotic Systems (pp. 303\u0026ndash;315). Springer International Publishing. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-030-23807-0_25\u003c/span\u003e\u003cspan address=\"10.1007/978-3-030-23807-0_25\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaker M (1990) Negotiated tutoring: An approach to interaction in intelligent tutoring systems. [Doctoral dissertation, The Open University]. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.semanticscholar.org/paper/Negotiated-tutoring-%3A-an-approach-to-interaction-in-Baker/d6ede52bb412c28d20379753eda333dbf3a818c3\u003c/span\u003e\u003cspan address=\"https://www.semanticscholar.org/paper/Negotiated-tutoring-%3A-an-approach-to-interaction-in-Baker/d6ede52bb412c28d20379753eda333dbf3a818c3\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaker MJ (1992) Design of an intelligent tutoring system for musical structure and interpretation. Understanding music with AI: Perspectives on music cognition (pp. 466\u0026ndash;489)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBaumeister R, Leary M (1997) Writing Narrative Literature Reviews. Rev Gen Psychol 1:311\u0026ndash;320. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1037//1089-2680.1.3.311\u003c/span\u003e\u003cspan address=\"10.1037//1089-2680.1.3.311\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoland RJ (2003) The Problems Associated with the Design, Implementation and Evaluation of an Intelligent Piano Tutor. [Doctoral dissertation, University of Canberra]. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://researchsystem.canberra.edu.au/ws/portalfiles/portal/57342228/Boland_Robert_P.pdf\u003c/span\u003e\u003cspan address=\"https://researchsystem.canberra.edu.au/ws/portalfiles/portal/57342228/Boland_Robert_P.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBorgogno T, Turchet L (2022) ImproScales: A self-tutoring web system for using scales in improvisations. Proceedings of the 17th International Audio Mostly Conference (pp. 219\u0026ndash;225)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrandao M (2005) Guided discovery tutoring and schoenberg\u0026rsquo;s harmony teaching method: An investigation\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCampesato O (2020) Artificial Intelligence, Machine Learning, and Deep Learning. Mercury Learning and Information\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCapell P, Dannenberg RB (1993) Instructional Design and Intelligent Tutoring: Theory and the Precision of Design. J Artif Intell Educ 4(1):95\u0026ndash;121\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen H-T, He Y-Z, Hsu C-C (2018) Computer-assisted yoga training system. Multimedia Tools Appl 77:23969\u0026ndash;23991\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChiang H-H, Chen W-M, Chao H-C, Tsai D-L (2019) A virtual tutor movement learning system in eLearning. Multimedia Tools Appl 78:4835\u0026ndash;4850\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChin D, Xia G (2022) A Computer-aided Multimodal Music Learning System with Curriculum: A Pilot Study. Proceedings of the International Conference on New Interfaces for Musical Expression (NIME) 2022. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.21428/92fbeb44.c6910363\u003c/span\u003e\u003cspan address=\"10.21428/92fbeb44.c6910363\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChin D, Zhang Y, Zhang T, Zhao J, Xia GG (2020) Interactive Rainbow Score: A Visual-centered Multimodal Flute Tutoring System. Proceedings of the International Conference on New Interfaces for Musical Expression (NIME) 2020\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCook J (1998) Mentoring, Metacognition and Music: Interaction Analyses and Implications for Intelligent Learning Environments\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCorbalan G, Kester L, Van Merri\u0026euml;nboer JJG (2006) Towards a personalized task selection model with shared instructional control. Instr Sci 34(5):399\u0026ndash;422. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11251-005-5774-2\u003c/span\u003e\u003cspan address=\"10.1007/s11251-005-5774-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCorbett AT, Koedinger KR, Anderson JR (1997) Intelligent Tutoring Systems. In M. G. Helander, T. K. Landauer, \u0026amp; P. V. Prabhu (Eds.), Handbook of Human-Computer Interaction (Second Edition) (pp. 849\u0026ndash;874). North-Holland. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/B978-044481862-1.50103-5\u003c/span\u003e\u003cspan address=\"10.1016/B978-044481862-1.50103-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDannenberg RB, Sanchez M, Joseph A, Capell P, Joseph R, Saul R (1990) A computer-based multi‐media tutor for beginning piano students. J New Music Res 19(2\u0026ndash;3):155\u0026ndash;173\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDannenberg RB, Sanchez M, Joseph A, Joseph R, Saul R, Capell P (1993) Results from the piano tutor project. Proceedings of the Fourth Biennial Arts and Technology Symposium, 143\u0026ndash;150\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDebevc M, Weiss J, Šorgo A, Kožuh I (2020) Solfeggio learning and the influence of a mobile application based on visual, auditory and tactile modalities. Br J Edu Technol 51(1):177\u0026ndash;193. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/bjet.12792\u003c/span\u003e\u003cspan address=\"10.1111/bjet.12792\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDella Ventura MD (2022) A Self-adaptive Learning Music Composition Algorithm as Virtual Tutor. In I. Maglogiannis, L. Iliadis, J. Macintyre, \u0026amp; P. Cortez (Eds.), Artificial Intelligence Applications and Innovations (pp. 16\u0026ndash;26). Springer International Publishing. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-031-08333-4_2\u003c/span\u003e\u003cspan address=\"10.1007/978-3-031-08333-4_2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFober D, Letz S, Orlarey Y, Askenfelt A, Falkenberg-Hansen K, Schoonderwaldt E (2004) Imutus-an interactive music tuition system. Proceedings of the Sound and Music Computing Conference, 97\u0026ndash;103\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGajjar K, Patel M (2020) RAAGANG\u0026mdash;a proposed model of tutoring system for novice learners of Hindustani classical music. ICT Systems and Sustainability: Proceedings of ICT4SD 2019, Volume 1, 509\u0026ndash;517\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGreiner E (2011) A prototype Intelligent Tutoring System for musical chord theory. [Bachelor thesis, University of Amsterdam]. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://staff.fnwi.uva.nl/b.bredeweg/pdf/BSc/20102011/Greiner.pdf\u003c/span\u003e\u003cspan address=\"https://staff.fnwi.uva.nl/b.bredeweg/pdf/BSc/20102011/Greiner.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrigutis L (2018) Investigation into aspects of feedback and gamification in computer-assisted musical instrument tutoring. [Master\u0026rsquo;s thesis, University of Huddersfield]. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://eprints.hud.ac.uk/id/eprint/34633/1/Laurynas%20Grigutis%20FINAL%20THESIS.PDF\u003c/span\u003e\u003cspan address=\"https://eprints.hud.ac.uk/id/eprint/34633/1/Laurynas%20Grigutis%20FINAL%20THESIS.PDF\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrosshauser T, Hufnagl B, Tr\u0026ouml;ster G (2013) Sensor Based Hand and Lip Weight and Pressure Measurements in Trombone Playing. Proceedings of SMAC 2013: 4th Stockholm Music Acoustics Conference, 509\u0026ndash;513\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrosshauser T, Tessendorf B, Tr\u0026ouml;ster G, Hildebrandt H, Candia V Sensor setup for force and finger position and tilt measurements for pianists. Proceedings of the 9th Sound and Music Computing, Conference (2012) SMC 2012\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrosshauser T, Tr\u0026ouml;ster G (2014) Musical instrument interaction: Development of a sensor fingerboard for string instruments. Proceedings of the 8th International Conference on Tangible, Embedded and Embodied Interaction, 177\u0026ndash;180. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/2540930.2540956\u003c/span\u003e\u003cspan address=\"10.1145/2540930.2540956\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGuarnieri V, Neagu L-M, Rigaud E, Travadel S, Dascalu M (2023) Enhancing Load Evaluation in Intelligent Tutoring Systems Through Velocity-Based Training. Proceedings of the Conference on Smart Learning Ecosystems and Regional Development, 21\u0026ndash;37\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHammond T, Kumar SPA, Runyon M, Cherian J, Williford B, Keshavabhotla S, Valentine S, Li W, Linsey J (2018) It\u0026rsquo;s not just about accuracy: Metrics that matter when modeling expert sketching ability. ACM Trans Interact Intell Syst (TiiS) 8(3):1\u0026ndash;47\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHolland S (2000), August 8 Artificial Intelligence in Music Education: A Critical Review. Readings in Music and Artificial Intelligence. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.semanticscholar.org/paper/Artificial-Intelligence-in-Music-Education%3A-A-Holland/a78fa73c5d04169c34b2a945a0a38e5a210773b7\u003c/span\u003e\u003cspan address=\"https://www.semanticscholar.org/paper/Artificial-Intelligence-in-Music-Education%3A-A-Holland/a78fa73c5d04169c34b2a945a0a38e5a210773b7\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHuanhuan L (2009) Computer assisted music instrument tutoring applied to violin practice. [Master thesis, National University of Singapore]. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.comp.nus.edu.sg/~leowwk/thesis/luhuanhuan.pdf\u003c/span\u003e\u003cspan address=\"https://www.comp.nus.edu.sg/~leowwk/thesis/luhuanhuan.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHui B (2023) Visualization system of martial arts training action based on artificial intelligence algorithm. Soft Comput, 1\u0026ndash;12\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIarussi E, Bousseau A, Tsandilas T (2013) The drawing assistant: Automated drawing guidance and feedback from photographs. Proceedings of the 26th Annual ACM Symposium on User Interface Software and Technology, 183\u0026ndash;192. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/2501988.2501997\u003c/span\u003e\u003cspan address=\"10.1145/2501988.2501997\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJamshidi F, Marghitu D, Chapman R (2021) Developing an Online Music Teaching and Practicing Platform via Machine Learning: A Review Paper. In M. Antona \u0026amp; C. Stephanidis (Eds.), Universal Access in Human-Computer Interaction. Access to Media, Learning and Assistive Environments (Vol. 12769, pp. 95\u0026ndash;108). Springer International Publishing. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-030-78095-1_9\u003c/span\u003e\u003cspan address=\"10.1007/978-3-030-78095-1_9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJi R (2020) Research on Basketball Shooting Action Based on Image Feature Extraction and Machine Learning. IEEE Access, 8, 138743\u0026ndash;138751. IEEE Access. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/ACCESS.2020.3012456\u003c/span\u003e\u003cspan address=\"10.1109/ACCESS.2020.3012456\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJin L, Lin B, Hong M, Zhang K, So H-J (2025) Exploring the Impact of an LLM-Based Teachable Agent on Learning Gains and Cognitive Load in Music Education\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohnson D, Damian D, Tzanetakis G (2019) Detecting Hand Posture in Piano Playing Using Depth Data. Comput Music J 43(1):59\u0026ndash;78. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1162/COMJ_a_00500\u003c/span\u003e\u003cspan address=\"10.1162/COMJ_a_00500\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJohnson RMG, Van Der Linden J, Rogers Y (2010) MusicJacket: The efficacy of real-time vibrotactile feedback for learning to play the violin. CHI \u0026rsquo;10 Extended Abstracts on Human Factors in Computing Systems. 3475\u0026ndash;3480. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/1753846.1754004\u003c/span\u003e\u003cspan address=\"10.1145/1753846.1754004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJuslin PN, Karlsson J, Lindstr\u0026ouml;m E, Friberg A, Schoonderwaldt E (2006) Play it again with feeling: Computer feedback in musical communication of emotions. J Experimental Psychology: Appl 12(2):79\u0026ndash;95. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1037/1076-898X.12.2.79\u003c/span\u003e\u003cspan address=\"10.1037/1076-898X.12.2.79\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKale G, Patil V, Munot M (2021) A novel and intelligent vision-based tutor for Yogāsana: E-YogaGuru. Mach Vis Appl 32:1\u0026ndash;17\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKeshavabhotla S, Williford B, Kumar S, Hilton E, Taele P, Li W, Linsey J, Hammond T (2017) Conquering the cube: Learning to sketch primitives in perspective with an intelligent tutoring system. Proceedings of the Symposium on Sketch-Based Interfaces and Modeling, 1\u0026ndash;11\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKoedinger K, Anderson J, Hadley W, Mark M (1997) Intelligent Tutoring Goes To School in the Big City. Int J Artif Intell Educ 8:30\u0026ndash;43\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKotte H, Kravcik M, Duong-Trung N (2023) Real-Time Posture Correction in Gym Exercises: A Computer Vision-Based Approach for Performance Analysis, Error Classification and Feedback. MILeS@ EC-TEL, pp 64\u0026ndash;70\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLieberman J, Breazeal C (2007) TIKL: Development of a wearable vibrotactile feedback suit for improved human motor learning. IEEE Trans Robot 23(5):919\u0026ndash;926\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLin K-C, Ko C-W, Hung H-C, Chen N-S (2023) The effect of real-time pose recognition on badminton learning performance. Interact Learn Environ 31(8):4772\u0026ndash;4786. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10494820.2021.1981396\u003c/span\u003e\u003cspan address=\"10.1080/10494820.2021.1981396\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eL\u0026ouml;rch L, Huovinen E (2025) Fostering Composer Voice in Tertiary Teaching of Contemporary Music Composition. J Res Music Educ 73(1):87\u0026ndash;108. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/00224294241255579\u003c/span\u003e\u003cspan address=\"10.1177/00224294241255579\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMa W, Adesope OO, Nesbit JC, Liu Q (2014) Intelligent tutoring systems and learning outcomes: A meta-analysis. J Educ Psychol 106(4):901\u0026ndash;918. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1037/a0037123\u003c/span\u003e\u003cspan address=\"10.1037/a0037123\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMartin F, Klein JD, Sullivan H (2007) The impact of instructional elements in computer-based instruction. Br J Edu Technol 38(4):623\u0026ndash;636. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/j.1467-8535.2006.00670.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1467-8535.2006.00670.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMat Sanusi KA, Mitri DD, Limbu B, Klemke R (2021) Table tennis tutor: Forehand strokes classification based on multimodal data and neural networks. Sensors 21(9):3121\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcNamara DS, Levinstein IB, Boonthum C (2004) iSTART: Interactive strategy training for active reading and thinking. Behav Res Methods Instruments Computers 36(2):222\u0026ndash;233. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3758/BF03195567\u003c/span\u003e\u003cspan address=\"10.3758/BF03195567\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMittal P, Vatsa M, Singh R (2015) Composite sketch recognition via deep network\u0026mdash;A transfer learning approach. 2015 International Conference on Biometrics (ICB), 251\u0026ndash;256. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/ICB.2015.7139092\u003c/span\u003e\u003cspan address=\"10.1109/ICB.2015.7139092\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMolero D, Schez-Sobrino S, Vallejo D, Glez-Morcillo C, Albusac J (2021) A novel approach to learning music and piano based on mixed reality and gamification. Multimedia Tools Appl 80(1):165\u0026ndash;186. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11042-020-09678-9\u003c/span\u003e\u003cspan address=\"10.1007/s11042-020-09678-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMoringen A, R\u0026uuml;ttgers S, Zintgraf L, Friedman J, Ritter H (2021) Optimizing piano practice with a utility-based scaffold (No. arXiv:2106.12937). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.2106.12937\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.2106.12937\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. arXiv\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNeagu L (2022) Intelligent Tutoring Systems for Psychomotor Development in Open Environments. [Doctoral dissertation, Universit\u0026eacute; Paris sciences et lettres \u0026amp; Universitatea politehnica Bucarest]. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://pastel.hal.science/tel-04751910/document\u003c/span\u003e\u003cspan address=\"https://pastel.hal.science/tel-04751910/document\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNeagu L, Rigaud E, Guarnieri V, Dascalu M, Travadel S (2022) Selfit v2 \u0026ndash; Challenges Encountered in Building a Psychomotor Intelligent Tutoring System. In S. Crossley \u0026amp; E. Popescu (Eds.), Intelligent Tutoring Systems (Vol. 13284, pp. 350\u0026ndash;361). Springer International Publishing. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/978-3-031-09680-8_33\u003c/span\u003e\u003cspan address=\"10.1007/978-3-031-09680-8_33\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNeagu L, Rigaud E, Guarnieri V, Radu EI, Travadel S, Dascalu M, Rughinis R (2022) OntoStrength: An Ontology for Psychomotor Strength Development. Interact Des Architecture(s) 52:101\u0026ndash;108. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.55612/s-5002-052-006\u003c/span\u003e\u003cspan address=\"10.55612/s-5002-052-006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNeagu L, Rigaud E, Guarnieri V, Travadel S, Dascalu M (2021) Selfit\u0026ndash;an intelligent tutoring system for psychomotor development. Intelligent Tutoring Systems: 17th International Conference, ITS 2021, Virtual Event, June 7\u0026ndash;11, 2021, Proceedings 17, 291\u0026ndash;295\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNguyen SM, Devanne M, Remy-Neris O, Lempereur M, Thepaut A (2025) A Medical Low-Back Pain Physical Rehabilitation Dataset for Human Body Movement Analysis (No. arXiv:2407.00521). arXiv. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.2407.00521\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.2407.00521\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOwusu G (2007) AI and computer-based methods in performance evaluation of sporting feats: An overview. Artif Intell Rev 27(1):57\u0026ndash;70. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10462-008-9068-3\u003c/span\u003e\u003cspan address=\"10.1007/s10462-008-9068-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePaulson B, Hammond T (2008) PaleoSketch: Accurate primitive sketch recognition and beautification. Proceedings of the 13th International Conference on Intelligent User Interfaces, 1\u0026ndash;10. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/1378773.1378775\u003c/span\u003e\u003cspan address=\"10.1145/1378773.1378775\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePercival G, Wang Y, Tzanetakis G (2007) Effective use of multimedia for computer-assisted musical instrument tutoring. Proceedings of the International Workshop on Educational Multimedia and Multimedia Education, 67\u0026ndash;76\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePhon-Amnuaisuk S, Siong CK (2008) Web-Based Music Intelligent Tutoring Systems. In Interactive Multimedia Music Technologies. IGI Global 231\u0026ndash;248. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4018/978-1-59904-150-6.ch011\u003c/span\u003e\u003cspan address=\"10.4018/978-1-59904-150-6.ch011\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRaptis S, Baxevanis A, Askenfelt A, Schoonderwaldt E, Falkenberg K, Fober D, Letz S, Orlarey Y (2005) IMUTUS - An Effective Practicing Environment for Music Tuition\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRitschel H, Seiderer A, Andre E (2020) Pianobot: An Adaptive Robotic Piano Tutor\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRitter S, Kulikowich J, Lei P-W, Mcguire C, Morgan P (2007) What Evidence Matters? A randomized field trial of Cognitive Tutor Algebra I. (p. 20)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRomano G, Schneider J, Drachsler H (2019) Dancing Salsa with Machines\u0026mdash;Filling the Gap of Dancing Learning Solutions. Sensors 19(17) Article 17. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/s19173661\u003c/span\u003e\u003cspan address=\"10.3390/s19173661\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRowley K, Carlson P, Miller T (1998) A Cognitive Technology to Teach Composition Skills: Four Studies with the R-Wise Writing Tutor. J Educational Comput Res 18(3):259\u0026ndash;296. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2190/KW4V-FJKD-L7J1-EFK0\u003c/span\u003e\u003cspan address=\"10.2190/KW4V-FJKD-L7J1-EFK0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSanchez M, Joseph A, Dannenberg R, Miller P, Capell P, Joseph R (1987) Piano tutor: An Intelligent Keyboard Instruction System. Proceedings of the First International Workshop on Artificial Intelligence and Music,. St. Paul, Minnesota: AAAI\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSantos OC (2019) Artificial Intelligence in Psychomotor Learning: Modeling Human Motion from Inertial Sensor Data. Int J Artif Intell Tools 28(04):1940006. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1142/S0218213019400062\u003c/span\u003e\u003cspan address=\"10.1142/S0218213019400062\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSanz J (2022) Improving music education using biofeedback-based cognitive tutoring\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchoonderwaldt E, Askenfelt A, Falkenberg K (2005) Design and implementation of automatic evaluation of recorder performance in IMUTUS\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSiddaway AP, Wood AM, Hedges LV (2019) How to Do a Systematic Review: A Best Practice Guide for Conducting and Reporting Narrative Reviews, Meta-Analyses, and Meta-Syntheses. Annual Review of Psychology, 70(Volume 70, 2019), 747\u0026ndash;770. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1146/annurev-psych-010418-102803\u003c/span\u003e\u003cspan address=\"10.1146/annurev-psych-010418-102803\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSosa GW, Berger DE, Saw AT, Mary JC (2011) Effectiveness of Computer-Assisted Instruction in Statistics: A Meta-Analysis. Rev Educ Res 81(1):97\u0026ndash;128. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3102/0034654310378174\u003c/span\u003e\u003cspan address=\"10.3102/0034654310378174\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSottilare R, Graesser A, Hu X, Holden H (eds) (2013) Design Recommendations for Intelligent Tutoring Systems. U.S. Army Research Laboratory\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSteenbergen-Hu S, Cooper H (2014) A meta-analysis of the effectiveness of intelligent tutoring systems on college students\u0026rsquo; academic learning. J Educ Psychol 106(2):331\u0026ndash;347. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1037/a0034752\u003c/span\u003e\u003cspan address=\"10.1037/a0034752\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTaele P, Barreto L, Hammond T (2015) Maestoso: An intelligent educational sketching tool for learning music theory. Proceedings of the AAAI Conference on Artificial Intelligence, 29(2), 3999\u0026ndash;4005\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTamim RM, Bernard RM, Borokhovski E, Abrami PC, Schmid RF (2011) What Forty Years of Research Says About the Impact of Technology on Learning: A Second-Order Meta-Analysis and Validation Study. Rev Educ Res 81(1):4\u0026ndash;28. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3102/0034654310393361\u003c/span\u003e\u003cspan address=\"10.3102/0034654310393361\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTetzlaff L, Schmiedek F, Brod G (2021) Developing personalized education. A dynamic framework. Educational Psychol Rev 33(3):863\u0026ndash;882. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.25656/01:25237\u003c/span\u003e\u003cspan address=\"10.25656/01:25237\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTong AKY, Angelides MC (2000) An empirical model for tutoring strategy selection in multimedia tutoring systems. Decis Support Syst 29(1):31\u0026ndash;45. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S0167-9236(00)00061-0\u003c/span\u003e\u003cspan address=\"10.1016/S0167-9236(00)00061-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTsiriga V, Virvou M (2004) Evaluating the intelligent features of a web-based intelligent computer assisted language learning system. Int J Artif Intell Tools 13(02):411\u0026ndash;425. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1142/S0218213004001600\u003c/span\u003e\u003cspan address=\"10.1142/S0218213004001600\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVanLehn K (2011) The Relative Effectiveness of Human Tutoring, Intelligent Tutoring Systems, and Other Tutoring Systems. Educational Psychol 46(4):197\u0026ndash;221. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/00461520.2011.611369\u003c/span\u003e\u003cspan address=\"10.1080/00461520.2011.611369\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVanLehn K, Graesser AC, Jackson GT, Jordan P, Olney A, Ros\u0026eacute; CP (2007) Than Reading? Cogn Sci 31(1):3\u0026ndash;62. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/03640210709336984\u003c/span\u003e\u003cspan address=\"10.1080/03640210709336984\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. When Are Tutorial Dialogues More Effective\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVanLehn K, Jordan PW, Ros\u0026eacute; CP, Bhembe D, B\u0026ouml;ttner M, Gaydos A, Makatchev M, Pappuswamy U, Ringenberg M, Roque A, Siler S, Srivastava R (2002) The Architecture of Why2-Atlas: A Coach for Qualitative Physics Essay Writing. In S. A. Cerri, G. Gouard\u0026egrave;res, \u0026amp; F. Paragua\u0026ccedil;u (Eds.), Intelligent Tutoring Systems (pp. 158\u0026ndash;167). Springer. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/3-540-47987-2_20\u003c/span\u003e\u003cspan address=\"10.1007/3-540-47987-2_20\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVides Ceron F (2012) TAYouKi: A Sketch-Based Tutoring System for Young Kids\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWesely S, Hofer E, Curth R, Paryani S, Mills N, Uebersch\u0026auml;r O, Westermayr J (2025) Artificial intelligence for objective assessment of acrobatic movements: How to apply machine learning for identifying tumbling elements in cheer sports (No. arXiv:2503.04764). arXiv. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.48550/arXiv.2503.04764\u003c/span\u003e\u003cspan address=\"10.48550/arXiv.2503.04764\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilliford B SketchTivity: Improving Creativity by Learning Sketching with an Intelligent Tutoring System. Proceedings of the 2017 ACM SIGCHI Conference on Creativity and, Cognition (2017) 477\u0026ndash;483. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/3059454.3078695\u003c/span\u003e\u003cspan address=\"10.1145/3059454.3078695\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilliford B, Runyon M, Li W, Linsey J, Hammond T (2020) Exploring the potential of an intelligent tutoring system for sketching fundamentals. Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, 1\u0026ndash;13\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang C, Chiang F-K, Cheng Q, Ji J (2021) Machine Learning-Based Student Modeling Methodology for Intelligent Tutoring Systems. J Educational Comput Res 59(6):1015\u0026ndash;1035. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/0735633120986256\u003c/span\u003e\u003cspan address=\"10.1177/0735633120986256\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYuksel BF, Oleson KB, Harrison L, Peck EM, Afergan D, Chang R, Jacob RJ (2016) Learn piano with BACh: An adaptive learning interface that adjusts task difficulty based on brain state. Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, 5372\u0026ndash;5384\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang X, Huang Y, Zou Q, Pei Y, Zhang R, Wang S (2020) A Hybrid convolutional neural network for sketch recognition. Pattern Recognit Lett 130:73\u0026ndash;82. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.patrec.2019.01.006\u003c/span\u003e\u003cspan address=\"10.1016/j.patrec.2019.01.006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou M, Gong T (2020) Optimization of multimedia computer-aided interaction system of vocal music teaching based on voice recognition. Comput-Aided Des Appl 8:113\u0026ndash;122\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[{"identity":"3d4b2862-71b0-4762-8dec-6de89eb67e81","identifier":"10.13039/501100002347","name":"Bundesministerium für Bildung und Forschung","awardNumber":"0000","order_by":0}],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"DIPF | Leibniz Institute for Research and Information in Education","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"intelligent tutoring systems, arts, music, sports, machine learning, large-language models, artificial intelligence","lastPublishedDoi":"10.21203/rs.3.rs-8797891/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8797891/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eIntelligent tutoring systems (ITS) have been popular in well-formalized educational domains such as math and physics. The domains of arts, music, and sports, however, have been challenging for ITS due to their open-ended, multimodal nature. We conducted a systematic, narrative literature review to investigate how ITS in arts, music, and sports address these challenges. How do ITS in arts, music, and sports individualize instruction in the face of open-ended tasks? How do they use technology to handle multimodal data? We searched eight databases (Google Scholar, PsychInfo, PsychArticles, Education Research Complete, The New Republic Archive, ProQuest, ERIC, Psyndex) using a search string that combined the term \u0026ldquo;intelligent tutoring system\u0026rdquo; and its synonyms with the terms \u0026ldquo;arts,\u0026rdquo; \u0026ldquo;music,\u0026rdquo; and \u0026ldquo;sports\u0026rdquo;. We included 42 publications describing 29 ITS in our review. We found that all of the reviewed ITS addressed skills such as playing a musical instrument that were complex and multifaceted but rather well-formalized than open-ended. Most ITS were focused on providing feedback. However, some systems achieved individualized task selection by decomposing complex skills into component skills and mapping them to tasks. Regarding technology use, we identified two overarching issues: measuring or presenting body movements with hardware such as depth cameras, digital styluses, and robots; and optimizing feedback or task selection with machine learning algorithms. We consider the lack of ITS that address open-ended, creative skills to be a gap in the literature that should be addressed by future research and discuss the potential of artificial intelligence in tackling this challenge.\u003c/p\u003e","manuscriptTitle":"Intelligent Tutoring Systems in Open-ended, Multimodal Domains Reviewing the Evidence from Arts, Music, and Sports","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-06 08:51:50","doi":"10.21203/rs.3.rs-8797891/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ff1177ee-7688-4a77-b256-26ed8609628d","owner":[],"postedDate":"February 6th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":62438527,"name":"Educational Psychology"}],"tags":[],"updatedAt":"2026-02-06T08:51:50+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-06 08:51:50","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8797891","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8797891","identity":"rs-8797891","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}