Mitigating Disengagement in Higher Music Education:The Role of AI-Assisted Piano Instruction in Sustaining Student Persistence

Mitigating Disengagement in Higher Music Education:The Role of AI-Assisted Piano Instruction in Sustaining Student Persistence | 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 Research Article Mitigating Disengagement in Higher Music Education:The Role of AI-Assisted Piano Instruction in Sustaining Student Persistence LIU CHANG This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9485161/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 This study investigated whether AI-assisted piano instruction enhances engagement, persistence, and satisfaction among 97 undergraduate music students at a Chinese university. Grounded in self-determination theory, the technology acceptance model, and Tinto's integration framework, a 16-week quasi-experimental design compared an AI-assisted group (n = 49) with a traditional instruction group (n = 48). Results revealed significantly higher post-test scores in the AI-assisted group for engagement (d = 1.11), persistence intention (d = 0.88), and satisfaction (d = 1.13), all at p < .001. Autonomy (d = 0.97), competence (d = 0.96), and relatedness (d = 0.72) also improved significantly. Dropout risk was directionally lower in the AI-assisted group (6.1% vs. 18.8%) but non-significant (p = .114). Weekly practice hours were significantly higher in the experimental group (M = 6.82 vs. 5.23, d = 0.80). These findings suggest AI-assisted piano pedagogy can mitigate disengagement by fulfilling psychological needs, supporting retention strategies aligned with SDG 4. Educational Psychology AI-assisted instruction piano pedagogy student persistence engagement self-determination theory higher music education dropout technology-enhanced learning adaptive feedback SDG 4 Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Introduction Higher education institutions worldwide face persistent and multifaceted challenges in retaining students through degree completion. Despite decades of research on student attrition, dropout rates remain stubbornly high across many disciplines, with consequences that extend beyond individual students to affect institutional resources, program viability, and national human capital development (Tinto, 2017 ). The phenomenon of student dropout is particularly complex because it emerges from the intersection of individual characteristics, academic experiences, institutional environments, and broader socioeconomic conditions, making it resistant to simple interventions or one-size-fits-all solutions. Within this landscape, music programs confront distinctive retention pressures that have received comparatively limited empirical attention, despite the demanding nature of conservatory and university-level musical training. Music education at the tertiary level differs fundamentally from most academic disciplines in its reliance on extended, solitary practice as the primary mechanism of skill development. Piano students, in particular, typically spend between 15 and 25 hours per week in independent practice, a largely unsupervised activity that requires sustained concentration, self-regulation, and intrinsic motivation (Jorgensen & Hallam, 2016 ). During these practice sessions, students must diagnose their own technical and interpretive weaknesses, devise strategies for improvement, and maintain the emotional resilience to persist through plateaus and setbacks—all without the immediate guidance of an instructor. This pedagogical structure creates fertile conditions for disengagement, as students who lack adequate feedback mechanisms, clear progress indicators, or a sense of connectedness to their musical community may gradually withdraw their effort, reduce their practice time, and ultimately question their commitment to the degree program (Creech et al., 2008 ). Empirical evidence from European conservatories and North American music programs suggests that dropout rates in music higher education range from 15% to 30%, with attrition concentrated in the first two years of study (Burt-Perkins & Mills, 2019 ). Factors contributing to music student dropout include performance anxiety, burnout from excessive practice demands, perceived mismatch between expectations and reality, financial pressures associated with instrument maintenance and lesson costs, competitive studio cultures that undermine belonging, and insufficient institutional support for students experiencing motivational crises (Bonneville-Roussy et al., 2017 ; Hatfield et al., 2022 ). Importantly, these factors often interact in complex, nonlinear ways: a student experiencing performance anxiety may reduce practice, leading to diminished competence perceptions, which further erodes motivation and ultimately triggers withdrawal consideration. Understanding and interrupting these cascading disengagement trajectories requires interventions that address multiple dimensions of the student experience simultaneously. The rapid integration of artificial intelligence into educational settings has prompted renewed interest in whether technology-mediated instruction can alleviate patterns of disengagement that traditional pedagogical approaches alone have struggled to resolve (Chen et al., 2022 ; Hwang et al., 2020 ). AI-driven tools in music education now offer capabilities that were unimaginable a decade ago: real-time acoustic analysis of pitch accuracy, rhythm precision, and dynamic expression; machine learning algorithms that generate personalized practice recommendations based on individual performance trajectories; adaptive difficulty calibration that ensures students are consistently challenged without being overwhelmed; and social features that connect learners to peer communities for collaborative feedback and mutual encouragement (Huang & Chen, 2023 ; Xie et al., 2022 ). Unlike conventional self-assessment methods such as metronome-based practice or self-recording, AI-powered platforms provide immediate, detailed, and objective diagnostics that approximate elements of expert mentorship during the many hours when students practice independently. Despite growing enthusiasm for AI applications in education, the empirical base linking AI-assisted instrumental instruction to student persistence outcomes remains limited, particularly within higher education contexts. The majority of existing studies on AI in music education focus on skill acquisition outcomes—such as improved pitch accuracy or sight-reading speed—rather than motivational, affective, or retention-related variables (Zhai et al., 2021 ). Furthermore, most technology-enhanced learning research has been conducted with K–12 populations or in general academic subjects such as mathematics and language learning, leaving a significant gap in understanding how AI integration functions within the specialized pedagogical environment of university-level music training. The one-to-one lesson format, the centrality of practice as a learning modality, and the high degree of autonomy required of music students create a context that may respond to AI-assisted instruction in ways that differ meaningfully from classroom-based disciplines (Gaunt, 2010 ). A further gap in the literature concerns the theoretical mechanisms through which technology-enhanced instruction influences persistence. Relatively few investigations have situated AI-assisted music instruction within established motivational and retention frameworks capable of explaining how technological support translates into sustained engagement and reduced attrition (Kahu & Nelson, 2018 ). Without such theoretical grounding, empirical findings risk remaining descriptive rather than explanatory, limiting their utility for designing principled interventions. The present study addresses this gap by integrating three complementary theoretical perspectives: self-determination theory (SDT; Ryan & Deci, 2020 ), which explains how environments that support autonomy, competence, and relatedness foster intrinsic motivation; the technology acceptance model (TAM; Venkatesh & Davis, 2000 ), which identifies the cognitive conditions under which students adopt and sustain use of technological tools; and Tinto’s student integration model (Tinto, 1993 , 2017 ), which contextualizes how academic and social integration within institutional environments promotes persistence to degree completion. This study therefore investigates the effects of a 16-week AI-assisted piano instruction intervention on undergraduate music students’ engagement, persistence intention, satisfaction, and dropout risk, while examining whether AI-mediated instruction enhances the fulfillment of basic psychological needs as theorized by SDT. The central research questions are: (RQ1) Does AI-assisted piano instruction produce significantly higher post-intervention scores on engagement, persistence intention, and satisfaction compared with traditional instruction? (RQ2) Does AI-assisted instruction enhance students’ perceived autonomy, competence, and relatedness? (RQ3) Is dropout risk lower among students receiving AI-assisted instruction? These questions carry practical significance for institutions seeking evidence-based strategies to support music student retention, and for the broader conversation around equitable, technology-enhanced quality education aligned with Sustainable Development Goal 4. 2. Literature Review and Theoretical Framework 2.1 Student Disengagement and Dropout in Music Higher Education Research on student attrition in music programs has identified a distinctive combination of academic, psychosocial, and institutional factors that set music education apart from other disciplines. Performance-based assessment, where students’ competence is publicly evaluated in recitals and jury examinations, creates high-stakes evaluative environments that can produce debilitating anxiety and undermine the intrinsic enjoyment that initially drew students to music (Kenny, 2011 ). The intense practice demands of conservatory-style training, often exceeding 20 hours per week of solitary instrumental work, can lead to physical injuries such as repetitive strain and focal dystonia, as well as psychological burnout characterized by emotional exhaustion, depersonalization, and reduced personal accomplishment (Ginsborg et al., 2009 ). Competitive studio cultures, in which students are ranked against their peers for performance opportunities and teacher attention, can erode the sense of belonging and relatedness that is critical for sustained engagement (Bonneville-Roussy et al., 2017 ). Longitudinal data from European conservatories and university music departments indicate that between 15% and 30% of enrolled music students leave their programs before completion, with the highest attrition rates occurring during the transition from the first to second year of study (Burt-Perkins & Mills, 2019 ). Students who withdraw often report a combination of diminished motivation, feelings of isolation during practice, inadequate or infrequent feedback from instructors, financial strain, and a growing disconnect between their initial musical aspirations and the realities of professional training (Gaunt, 2010 ; Hatfield et al., 2022 ). Notably, these factors frequently operate in mutually reinforcing cycles: insufficient feedback leads to stagnating skill development, which diminishes competence perceptions, which reduces practice motivation, which further impairs progress—a downward spiral that, left unaddressed, culminates in withdrawal. The challenge for music educators and institutions is to identify leverage points within these cycles where targeted interventions can interrupt disengagement trajectories and restore positive motivational dynamics. Recent shifts toward hybrid and online learning, accelerated by the global pandemic, have introduced additional challenges for music students. Virtual instruction often fails to capture the nuances of acoustic performance, creating frustrations for both students and instructors who rely on real-time sonic feedback during lessons (Dye, 2022 ). Digital fatigue, reduced peer interaction, and the absence of shared performance spaces have exacerbated feelings of isolation and disconnection among music students (Crawford, 2021 ). These developments underscore the urgency of exploring how technology can be deployed not merely as a substitute for in-person instruction but as a complement that addresses the specific motivational and relational needs of music learners in contemporary higher education environments. 2.2 Self-Determination Theory and Music Learning Self-determination theory (SDT) provides a comprehensive framework for understanding human motivation as shaped by the satisfaction of three innate and universal psychological needs: autonomy, competence, and relatedness (Ryan & Deci, 2020 ). Autonomy refers to the experience of volitional action and self-endorsement—the sense that one’s behavior originates from the self rather than from external pressures. Competence involves the experience of effectance and mastery—the perception that one is capable of achieving desired outcomes and growing in skill. Relatedness encompasses the feeling of connection, belonging, and care within social environments—the sense that one matters to others and is part of a meaningful community. According to SDT, environments that support the satisfaction of these three needs foster the internalization of motivation, promote deeper engagement with learning activities, enhance well-being, and increase the likelihood of sustained participation over time (Niemiec & Ryan, 2009 ). Within music education, SDT-informed research has demonstrated that studio environments promoting learner choice in repertoire selection, constructive and process-oriented feedback, and opportunities for peer collaboration are associated with higher practice motivation, greater intrinsic interest in musical activities, and reduced attrition (Evans, 2015 ; Miksza et al., 2022 ). Conversely, controlling teaching styles that emphasize compliance, performance outcomes over learning processes, and normative comparison among students tend to undermine need satisfaction and promote external forms of motivation that are associated with anxiety, disengagement, and dropout (Bonneville-Roussy et al., 2017 ). The theoretical prediction of SDT—that need-supportive environments promote persistence while need-thwarting environments promote withdrawal—maps directly onto the observed dynamics of music student retention. AI-assisted instruction has the potential to enhance need satisfaction through several mechanisms. Autonomy may be supported when AI platforms allow learners to set their own practice goals, choose the order and pace of exercises, and receive individualized recommendations rather than standardized assignments. Competence is enhanced when real-time feedback provides precise, actionable information about performance quality, enabling students to identify specific areas for improvement and to perceive tangible progress across practice sessions. Relatedness can be supported through social features such as peer dashboards, collaborative challenges, shared recordings, and community forums that maintain a sense of connection even during solitary practice (Xie et al., 2022 ). The present study tests whether these theorized mechanisms are empirically supported in the context of university-level piano instruction. 2.3 Technology Acceptance and AI in Education The technology acceptance model (TAM), originally proposed by Davis ( 1989 ) and subsequently extended by Venkatesh and Davis ( 2000 ), identifies perceived usefulness and perceived ease of use as the primary cognitive determinants of technology adoption behavior. Perceived usefulness refers to the degree to which a person believes that using a particular technology would enhance their performance, while perceived ease of use refers to the degree to which a person believes that using the technology would be free of effort. When students perceive an AI-assisted practice tool as both useful for improving their musical skills and easy to integrate into their existing practice routines, they are more likely to adopt the technology, use it consistently, and derive motivational and learning benefits from it (Hwang et al., 2020 ). In higher education contexts, TAM has been extensively applied to explain students’ acceptance of learning management systems, intelligent tutoring platforms, and AI-based educational tools across diverse disciplinary settings (Al-Emran et al., 2018 ; Scherer et al., 2019 ). A meta-analytic review by Scherer et al. ( 2019 ) found that perceived usefulness consistently exhibited the strongest relationship with technology acceptance among educational stakeholders, while perceived ease of use operated both directly and indirectly through its influence on perceived usefulness. For AI-assisted music instruction, these findings suggest that the design of the technological interface matters: platforms that are intuitive, responsive, and clearly connected to musical learning goals are more likely to be adopted and sustained by students, thereby maximizing the motivational benefits of AI-mediated feedback. TAM complements SDT by addressing a prerequisite condition for need satisfaction: if students do not accept and consistently use the AI tool, the potential benefits for autonomy, competence, and relatedness cannot be realized. In this sense, technology acceptance serves as a gateway mechanism that must be satisfied before the motivational dynamics predicted by SDT can operate. The integration of TAM with SDT in the present study provides a more complete account of the pathway from technology introduction to motivational outcomes to persistence behavior. 2.4 Tinto’s Integration Model and Institutional Persistence Tinto’s model of student integration (Tinto, 1993 , 2017 ) has been among the most influential frameworks in higher education retention research for over four decades. The model posits that students persist in their educational programs to the extent that they achieve sufficient academic and social integration within their institutional environment. Academic integration encompasses intellectual development, satisfactory academic performance, and meaningful interactions with academic systems and personnel, including faculty, advisors, and curricula. Social integration involves the development of interpersonal connections, peer relationships, and a sense of belonging within the broader campus community. When students experience high levels of both academic and social integration, they develop stronger institutional commitment and goal commitment, which in turn increase the likelihood of persistence to degree completion. In music departments, the primary sites of academic and social integration include the one-to-one lesson with a studio teacher, ensemble rehearsals and performances, peer interactions in practice rooms and common areas, and participation in masterclasses and departmental events (Gaunt, 2010 ). The studio teacher relationship is often the single most important factor in music students’ institutional integration, serving simultaneously as a source of technical instruction, artistic mentorship, emotional support, and professional socialization (Creech et al., 2008 ). However, this relationship is typically limited to one or two contact hours per week, leaving students to navigate the majority of their practice time without direct guidance or social connection. AI-assisted tools may contribute to academic integration by structuring independent practice in ways that deepen skill acquisition, by providing data-rich performance summaries that enrich teacher–student dialogue during lessons, and by helping students set and track meaningful learning goals that connect practice to broader curricular objectives. They may also support social integration when collaborative features enable students to share recordings, comment on peers’ progress, participate in group challenges, and maintain ongoing communication with fellow learners within the platform. In Tinto’s ( 2017 ) updated framework, which emphasizes the importance of self-efficacy, belonging, and perceived relevance of the curriculum, AI-assisted instruction may function as a tool that enhances all three of these conditions by making practice more productive, more connected, and more clearly aligned with students’ evolving musical goals. 2.5 Hypotheses Building on the integrated theoretical framework depicted in Fig. 1 , the following hypotheses were formulated to guide the empirical investigation. These hypotheses reflect the theorized mechanisms through which AI-assisted piano instruction influences motivational processes and retention-relevant outcomes: Hypothesis 1 (H1a–H1c) : Students receiving AI-assisted piano instruction will report significantly higher post-intervention levels of perceived autonomy (H1a), competence (H1b), and relatedness (H1c) compared with students receiving traditional instruction, after controlling for pre-test scores. Hypothesis 2 (H2a–H2c) : AI-assisted instruction will produce significantly greater improvements in composite engagement (H2a), persistence intention (H2b), and course satisfaction (H2c) relative to traditional instruction, after controlling for pre-test scores. Hypothesis 3 (H3) : The proportion of students classified as at risk of dropout will be significantly lower in the AI-assisted group than in the control group at the end of the 16-week intervention period. 3. Methods 3.1 Research Design and Context A quasi-experimental pre-test–post-test design with a non-equivalent control group was employed to evaluate the effects of AI-assisted piano instruction on student engagement, persistence, satisfaction, and dropout risk. This design was selected because random assignment at the individual level was not feasible within the existing course structure of the participating institution, where students were pre-enrolled in designated sections of the piano performance course based on scheduling availability and prior year of study (see Fig. 3 ). Two sections of an undergraduate piano performance course offered at a comprehensive public university in central China during the Fall 2024 semester were designated as the experimental and control conditions. The study spanned the full 16-week semester, with baseline data collected during Week 0 (prior to the start of the intervention) and post-test data collected during Week 16 (following the completion of the intervention period in Week 14, with a two-week buffer for post-testing). Ethical approval was obtained from the university’s institutional review board (IRB Protocol No. [to be inserted]), and all participants provided written informed consent prior to data collection. Students were informed that participation was voluntary and that their decision to participate or withdraw would have no impact on their course grades or academic standing. 3.2 Participants The sample comprised 97 undergraduate students enrolled in piano performance courses at the participating university. The experimental group (n = 49) received AI-assisted piano instruction supplemented with a commercial adaptive practice platform, while the control group (n = 48) received traditional piano instruction following the same curricular objectives and syllabus without AI supplementation. An a priori power analysis conducted using G*Power (version 3.1) indicated that a sample size of 90 participants (45 per group) would provide 80% power to detect a medium-to-large effect size (d = 0.60) at an alpha level of .05 using independent-samples t-tests. The obtained sample of 97 exceeded this threshold, ensuring adequate statistical power for the primary analyses. Participants ranged in age from 18 to 23 years (M = 19.67, SD = 1.36). The sample included 43 male students (44.3%) and 54 female students (55.7%), a gender distribution consistent with enrollment patterns in Chinese music programs. Participants represented all four undergraduate years: first-year (n = 29, 29.9%), second-year (n = 34, 35.1%), third-year (n = 21, 21.6%), and fourth-year (n = 13, 13.4%). Prior piano experience ranged from 0.5 to 8.0 years (M = 4.33, SD = 2.37), reflecting the diversity typical of university music programs that admit students with varying levels of pre-collegiate training. Cumulative grade point averages ranged from 2.00 to 4.00 (M = 3.21, SD = 0.43). Baseline comparisons using independent-samples t-tests confirmed that the two groups did not differ significantly on any demographic variable or pre-test measure (all p values > .05), supporting the assumption of baseline equivalence necessary for valid quasi-experimental inference. 3.3 Intervention Both the experimental and control groups received two 50-minute individual piano lessons per week from the same pool of four qualified piano instructors, all of whom held master’s degrees or higher in piano performance and had a minimum of five years of university-level teaching experience. Instructors were randomly assigned to teach both experimental and control students to minimize instructor effects. The instructional syllabus, shared across both conditions, covered technical exercises including scales, arpeggios, and études; repertoire development spanning Baroque, Classical, Romantic, and contemporary periods; sight-reading proficiency; and expressive performance skills including dynamics, phrasing, and pedaling technique. Lesson content, sequencing, and assessment criteria were standardized through a jointly developed syllabus document and biweekly instructor meetings to ensure fidelity of implementation across both conditions. The experimental group additionally used an AI-powered piano practice application during their independent practice sessions throughout the 14-week intervention period (Weeks 1–14). The application, a commercially available platform widely used in Chinese music education institutions, incorporated the following core functionalities: (a) real-time pitch and rhythm analysis through acoustic signal processing that detected deviations from the score and provided immediate visual feedback on a tablet display; (b) technique-specific diagnostics generated by machine learning algorithms trained on a large corpus of expert performance data, offering targeted suggestions for finger positioning, hand shape, wrist relaxation, and articulation; (c) adaptive practice scheduling that automatically adjusted exercise difficulty, tempo, and duration based on the learner’s performance trajectory across sessions, ensuring an appropriate balance between challenge and mastery; and (d) a social dashboard feature that enabled students to share practice clips, view anonymized peer progress summaries, post encouraging comments, and participate in weekly practice challenges organized by the research team. Students in the experimental group were instructed to use the AI platform during at least 60% of their weekly independent practice hours and to log all practice sessions through the application’s built-in tracking system. Compliance was monitored through weekly review of platform usage data by the research team, and students who fell below the 60% threshold for two consecutive weeks received a brief motivational reminder. The average compliance rate across the 14-week intervention period was 78.3% (SD = 11.6%), indicating that most students substantially exceeded the minimum usage requirement. Control-group students practiced independently using conventional methods, including metronomes, self-recording on personal devices, written instructor notes, and commercially available score-following applications that did not incorporate AI-based adaptive feedback. Control-group students were not prohibited from using any technology other than the specific AI platform employed in the experimental condition. 3.4 Instruments All instruments employed five-point Likert response scales (1 = strongly disagree to 5 = strongly agree) unless otherwise noted. Instruments were originally developed in English and translated into Mandarin Chinese following standard back-translation procedures (Brislin, 1970 ). The translated instruments were piloted with a separate sample of 30 music students (not included in the study sample) to verify clarity, cultural appropriateness, and psychometric adequacy. Minor wording adjustments were made based on pilot feedback. Instruments were administered online via a secure survey platform at both the pre-test (Week 0) and post-test (Week 16) time points, with completion typically requiring 15–20 minutes. Student Engagement. Engagement was measured using an adapted version of the Student Course Engagement Questionnaire (SCEQ; Handelsman et al., 2005 ), modified for music education contexts through consultation with three music education researchers and two piano instructors. The adapted instrument comprised four subscales: intrinsic motivation (6 items; sample item: I find the practice activities genuinely interesting; pre-test α = .90), behavioral engagement (6 items; sample item: I consistently put effort into my piano practice sessions; α = .89), emotional engagement (6 items; sample item: I feel enthusiastic when practicing piano; α = .92), and cognitive engagement (5 items; sample item: I try to connect new techniques with what I already know; α = .85). A composite engagement score was computed as the unweighted mean of the four subscale scores. Basic Psychological Need Satisfaction. Perceived autonomy (5 items; sample item: I feel I have choices in how I approach my piano practice; α = .81), competence (6 items; sample item: I feel capable of mastering challenging passages; α = .87), and relatedness (5 items; sample item: I feel connected to other students in the piano program; α = .83) were assessed using the Basic Psychological Need Satisfaction Scale adapted for educational technology contexts (Chen et al., 2015 ). A composite SDT score was computed as the mean of the three subscale scores. Persistence Intention. Persistence was measured using a 6-item scale adapted from the College Persistence Questionnaire (Davidson et al., 2009 ), modified to reflect the specific context of music degree programs. Items assessed students’ intention to continue their music studies through degree completion, their commitment to the program in the face of difficulties, and their confidence in finishing the degree. A sample item reads: I am committed to completing my music degree even when coursework feels overwhelming. Internal consistency was excellent (α = .91). Course Satisfaction. Satisfaction was measured using an 8-item instrument adapted from the Student Satisfaction Inventory (Schreiner & Juillerat, 1994 ), tailored to reflect music course contexts. Items addressed satisfaction with instructional quality, feedback adequacy, practice resources, course organization, and overall learning experience. Internal consistency was good (α = .88). Dropout Risk. Dropout risk was operationalized as a binary variable (0 = not at risk, 1 = at risk) indicating whether a student exhibited two or more of the following behavioral indicators by the end of the semester: (a) cumulative attendance below 75% of scheduled lessons; (b) failure to submit two or more required practice logs during the intervention period; (c) documented expression of intention to withdraw, transfer, or take a leave of absence as recorded in academic advising records or instructor communications. This multi-indicator approach was adopted to increase the ecological validity of the dropout risk measure beyond single-item self-report scales, which may be subject to social desirability bias. Weekly Practice Logs. Students in both groups maintained weekly practice logs documenting total practice hours and practice activities for each of the 16 weeks. Experimental-group students’ practice hours were automatically recorded by the AI platform, while control-group students submitted self-reported logs through the course management system. Additionally, the number of AI feedback interactions per practice session was automatically recorded for experimental-group students, providing a measure of technology engagement intensity. 3.5 Data Analysis Data were analyzed using IBM SPSS Statistics (Version 28) and R (Version 4.3.2). The analytical strategy proceeded in three stages. First, independent-samples t-tests and chi-square tests were conducted to verify baseline equivalence between the experimental and control groups on all demographic variables and pre-test outcome measures. Second, the primary hypotheses (H1–H3) were tested using independent-samples t-tests to compare post-test means between groups, with Cohen’s d calculated as the standardized mean difference using the pooled standard deviation. Effect sizes were interpreted according to conventional benchmarks: small (d = 0.20), medium (d = 0.50), and large (d = 0.80; Cohen, 1988 ). Ninety-five percent confidence intervals for all effect sizes were computed using the noncentral t-distribution method. The chi-square test of independence was used to compare the proportion of students classified as at risk of dropout between groups. Third, supplementary analyses examined weekly practice hours and AI interaction patterns across the 16-week period to provide additional context for interpreting the primary findings. All tests used a two-tailed significance threshold of α = .05. Given the family-wise error rate concerns associated with multiple comparisons across outcome variables, a Bonferroni-corrected threshold of p < .007 (for seven primary comparisons) was also applied as a sensitivity check, and all significant results except one remained significant under this more conservative criterion. 4. Results 4.1 Baseline Equivalence Independent-samples t-tests confirmed that the experimental and control groups did not differ significantly at the pre-test time point on any measured variable. Results indicated no significant group differences for age (t(95) = 0.32, p = .747, d = 0.07), cumulative GPA (t(95) = 1.75, p = .084, d = 0.35), prior piano experience in years (t(95) = − 0.35, p = .729, d = 0.07), composite engagement (t(95) = 0.61, p = .541, d = 0.12), SDT composite (t(95) = 1.34, p = .183, d = 0.27), persistence intention (t(95) = 0.86, p = .394, d = 0.17), and course satisfaction (t(95) = 1.06, p = .290, d = 0.22). A chi-square test confirmed no significant difference in gender distribution between groups (χ²(1) = 0.65, p = .421). These results provide strong evidence that any observed post-test differences can be attributed to the intervention rather than to pre-existing group differences. 4.2 Descriptive Statistics Table 1 presents the pre- and post-test means and standard deviations for both groups across all primary outcome variables. Visual inspection of the descriptive statistics reveals a consistent pattern: both groups had comparable scores at baseline, but the experimental group showed substantially larger gains from pre-test to post-test across all engagement subscales, SDT need satisfaction measures, persistence intention, and course satisfaction. The control group showed modest improvements or relative stability across the same period, suggesting that typical instructional exposure produced limited motivational gains in the absence of AI-assisted supplementation. Table 1 Pre- and Post-Test Descriptive Statistics by Group Variable Experimental (n = 49) Control (n = 48) Pre M Pre SD Post M Post SD Pre M Pre SD Post M Post SD Engagement (comp.) 2.93 0.34 3.45 0.42 2.89 0.37 3.00 0.39 Intrinsic Motiv. 3.05 0.67 3.54 0.61 2.93 0.72 3.01 0.68 Behavioral Eng. 2.87 0.63 3.35 0.60 2.83 0.60 2.93 0.62 Emotional Eng. 3.10 0.67 3.66 0.60 3.02 0.71 3.11 0.70 Cognitive Eng. 2.80 0.64 3.28 0.63 2.79 0.62 2.96 0.63 SDT Composite 3.03 0.35 3.62 0.37 2.87 0.38 3.02 0.38 Autonomy 3.04 0.55 3.54 0.52 2.89 0.62 2.98 0.63 Competence 2.82 0.58 3.47 0.60 2.74 0.67 2.85 0.66 Relatedness 3.24 0.52 3.55 0.56 3.00 0.60 3.09 0.57 Persistence 3.20 0.63 3.73 0.68 3.08 0.75 3.08 0.80 Satisfaction 2.99 0.50 3.65 0.62 2.88 0.60 2.94 0.65 Note. Comp. = composite. All scores on a 1–5 Likert scale. Pre = pre-test (Week 0); Post = post-test (Week 16). 4.3 Post-Test Group Comparisons and Effect Sizes Table 2 summarizes the results of post-test comparisons between the experimental and control groups, including t-statistics, significance values, effect sizes (Cohen’s d), and 95% confidence intervals for the effect sizes. All primary outcome variables showed statistically significant differences favoring the experimental group. With the exception of cognitive engagement (p = .014), all results remained significant after applying Bonferroni correction (adjusted α = .007). Table 2 Post-Test Group Comparisons with Effect Sizes Variable Exp M(SD) Ctrl M(SD) t(95) p d 95% CI Engagement 3.45(.42) 3.00(.39) 5.46 < .001 1.11 [0.68,1.53] Intrinsic Mot. 3.54(.61) 3.01(.68) 4.05 < .001 0.82 [0.41,1.23] Behavioral Eng. 3.35(.60) 2.93(.62) 3.39 .001 0.69 [0.28,1.09] Emotional Eng. 3.66(.60) 3.11(.70) 4.00 < .001 0.85 [0.43,1.26] Cognitive Eng. 3.28(.63) 2.96(.63) 2.51 .014 0.51 [0.10,0.91] SDT Composite 3.62(.37) 3.02(.38) 7.82 < .001 1.59 [1.13,2.04] Autonomy 3.54(.52) 2.98(.63) 4.77 < .001 0.97 [0.55,1.38] Competence 3.47(.60) 2.85(.66) 4.71 < .001 0.96 [0.54,1.37] Relatedness 3.55(.56) 3.09(.57) 3.56 < .001 0.72 [0.31,1.13] Persistence 3.73(.68) 3.08(.80) 4.32 < .001 0.88 [0.46,1.29] Satisfaction 3.65(.62) 2.94(.65) 5.55 < .001 1.13 [0.70,1.55] Note. Exp = Experimental; Ctrl = Control. CI = confidence interval for Cohen’s d. All Bonferroni-corrected results (adjusted α = .007) remained significant except cognitive engagement (p = .014). 4.4 Hypothesis Testing Hypotheses H1a through H1c predicted that AI-assisted instruction would enhance the satisfaction of basic psychological needs as conceptualized by SDT. All three hypotheses were fully supported by the data. Students in the experimental group reported significantly higher post-test autonomy (t(95) = 4.77, p < .001, d = 0.97, 95% CI [0.55, 1.38]) than their counterparts in the control group, indicating that the AI platform’s self-directed goal-setting and personalized recommendation features meaningfully enhanced students’ sense of volitional engagement with practice. Competence showed similarly large gains (t(95) = 4.71, p < .001, d = 0.96, 95% CI [0.54, 1.37]), consistent with the theoretical prediction that real-time performance feedback strengthens mastery experiences and reinforces self-efficacy. Relatedness exhibited a medium-to-large effect (t(95) = 3.56, p < .001, d = 0.72, 95% CI [0.31, 1.13]), suggesting that the social dashboard feature contributed to students’ sense of connectedness, though to a lesser extent than the competence and autonomy enhancements. Hypotheses H2a through H2c predicted that AI-assisted instruction would produce significantly greater improvements in engagement, persistence, and satisfaction. All three hypotheses were strongly supported. The experimental group demonstrated significantly higher post-test composite engagement (t(95) = 5.46, p < .001, d = 1.11, 95% CI [0.68, 1.53]), with large effects observed for emotional engagement (d = 0.85) and intrinsic motivation (d = 0.82), and medium effects for behavioral engagement (d = 0.69) and cognitive engagement (d = 0.51). Persistence intention showed a large effect (t(95) = 4.32, p < .001, d = 0.88, 95% CI [0.46, 1.29]), indicating that AI-assisted students expressed substantially stronger intentions to complete their music degree programs. Course satisfaction exhibited the largest effect among the primary outcomes (t(95) = 5.55, p < .001, d = 1.13, 95% CI [0.70, 1.55]), suggesting that the AI-enhanced practice experience fundamentally improved students’ evaluative assessments of their educational experience. Hypothesis H3 predicted that dropout risk would be significantly lower in the AI-assisted group. This hypothesis was partially supported in terms of direction but did not reach conventional statistical significance. The proportion of students classified as at risk was 6.1% in the experimental group (3 out of 49 students) compared with 18.8% in the control group (9 out of 48 students)—a difference of 12.7 percentage points representing a threefold reduction in relative risk. However, the chi-square test did not reach the conventional significance threshold (χ²(1) = 2.50, p = .114, Cramer’s V = .16). A post hoc power analysis indicated that detecting a significant difference in proportions of this magnitude (6% vs. 19%) at α = .05 with 80% power would require approximately 140 participants per group, suggesting that the non-significant result likely reflects insufficient statistical power for detecting differences in low-frequency binary outcomes rather than a true null effect. 4.5 Supplementary Analyses: Practice Behavior and AI Engagement Figure 4 presents a multi-panel visualization of the comparative outcomes. Panel A displays horizontal bar charts of mean gain scores (post-test minus pre-test) for each outcome variable, making the differential improvement between groups immediately visible. Satisfaction showed the largest absolute gain in the AI-assisted group (+ 0.66 points) compared with negligible change in the control group (+ 0.06), while emotional engagement demonstrated the second-largest differential (+ 0.56 vs. +0.09). Notably, persistence in the control group registered zero net change over the semester, underscoring the stagnation of motivation under conventional instruction. Panel B uses a radar chart to compare post-test SDT need satisfaction profiles, illustrating that the AI-assisted group achieved meaningfully higher scores across all three needs, with the most pronounced divergence on competence (3.47 vs. 2.85) and autonomy (3.54 vs. 2.98). Panel C tracks the pre-to-post trajectories for three key outcomes, revealing that the experimental group’s scores diverged sharply upward while the control group’s lines remained nearly flat. This divergence is especially striking for satisfaction, where the experimental trajectory rose from 2.99 to 3.65 while the control trajectory barely moved from 2.88 to 2.94. Panel D presents the dropout risk distribution using stacked bars, showing that only 3 students (6.1%) in the AI-assisted group were classified as at risk compared with 9 students (18.8%) in the control group, though this difference did not reach statistical significance (χ²(1) = 2.50, p = .114). Analysis of weekly practice logs revealed that students in the experimental group practiced a mean of 6.82 hours per week (SD = 1.94) across the 16-week observation period, compared with 5.23 hours per week (SD = 2.01) in the control group. This difference was statistically significant (t(95) = 3.91, p < .001, d = 0.80), representing a large effect that indicates the AI-assisted platform was associated with meaningfully increased practice engagement over time. Importantly, longitudinal examination of weekly practice hours revealed diverging trajectories: experimental-group students showed a gradual increase in practice hours from an average of 5.9 hours in Week 1 to 7.8 hours in Week 16, while control-group students showed relative stability with a slight declining trend from 5.5 hours in Week 1 to 5.0 hours in Week 16. Students in the experimental group generated an average of 9.4 AI feedback interactions per practice session, with interaction frequency increasing from a mean of 8.2 interactions in Week 1 to 12.6 interactions in Week 16. This upward trend suggests that students became increasingly engaged with the AI feedback system over time rather than experiencing novelty effects that dissipated after initial exposure. The correlation between cumulative AI interaction count and post-test engagement composite was positive and moderate (r = .38, p = .007), providing preliminary evidence that more intensive use of the AI platform was associated with greater motivational gains. Internal consistency reliability was assessed for all scales at both time points and was found to be satisfactory to excellent across all measures. Pre-test Cronbach’s alpha values ranged from .81 (autonomy) to .92 (emotional engagement), and post-test values ranged from .83 (relatedness) to .93 (emotional engagement). These values exceed the conventional threshold of .70 recommended for research purposes and support the psychometric adequacy of the measurement instruments in this sample. 5. Discussion 5.1 Summary and Interpretation of Primary Findings The primary aim of this study was to investigate whether AI-assisted piano instruction could mitigate student disengagement and support persistence in higher music education. The findings provide robust and consistent evidence that integration of adaptive AI tools into piano practice produces meaningful and statistically significant improvements in engagement, satisfaction, persistence intention, and basic psychological need satisfaction relative to traditional instruction. The magnitude of these effects—with Cohen’s d values ranging from 0.51 to 1.59 across the primary outcomes—is noteworthy, as educational interventions rarely produce effects of this size (Hattie, 2009 ). These results extend prior work on technology-enhanced learning in music education (Huang & Chen, 2023 ; Xie et al., 2022 ) by demonstrating that the benefits of AI-mediated feedback extend beyond the skill acquisition outcomes typically examined in music technology research to encompass motivational, affective, and retention-relevant variables that are directly implicated in student persistence. The large effect sizes observed for composite engagement (d = 1.11) and course satisfaction (d = 1.13) suggest that AI-assisted instruction substantially transformed the subjective quality of the practice experience for students in the experimental group. Rather than experiencing practice as a solitary, feedback-deficient activity punctuated by occasional instructor guidance, these students engaged with a responsive system that acknowledged their efforts, diagnosed their weaknesses, calibrated their challenges, and connected them to a community of fellow learners. This transformation of the practice experience appears to have had cascading effects on motivational processes, as reflected in the significant gains across all four engagement subscales and in the elevated persistence intentions reported by experimental-group students. From the perspective of self-determination theory, the pattern of SDT need satisfaction gains is theoretically coherent and practically illuminating. Competence exhibited the largest gain among the SDT subscales (d = 0.96), which aligns with the proposition that real-time performance feedback directly reinforces mastery experiences by providing students with clear, objective evidence of their progress and specific, actionable guidance for continued improvement (Bandura, 1997 ; Evans, 2015 ). When students can observe quantifiable improvements in pitch accuracy, rhythm precision, or dynamic control across practice sessions, their sense of musical competence is strengthened in ways that may not be possible through self-assessment alone. Autonomy showed a comparably large effect (d = 0.97), reflecting the platform’s capacity to empower students with choices about practice goals, exercise selection, and pacing—features that contrast with the relatively prescribed nature of traditional practice assignments. The comparatively smaller but still significant effect on relatedness (d = 0.72) suggests that while the social dashboard feature contributed to students’ sense of connection, the digital interaction afforded by the platform may not fully replicate the relational richness of in-person ensemble performances, shared practice room conversations, and face-to-face studio interactions that characterize the traditional conservatory experience. 5.2 Theoretical Contributions The integrated theoretical framework combining SDT, TAM, and Tinto’s student integration model proved productive for organizing and interpreting the multi-level mechanisms through which AI-assisted instruction influences retention-relevant outcomes. Each theoretical lens contributed a distinct and complementary perspective. SDT illuminated the motivational mechanisms by which technology-mediated feedback supports intrinsic engagement through the satisfaction of basic psychological needs. TAM helped explain why students adopted and sustained use of the AI platform—the prerequisite condition for motivational benefits to accrue—with high compliance rates (78.3%) and increasing interaction frequencies suggesting that students perceived the technology as both useful and usable. Tinto’s integration framework contextualized these individual-level gains within the broader institutional environment, framing the AI-assisted practice experience as a form of enhanced academic integration that strengthened students’ commitment to their degree programs. This tripartite framework addresses a noted limitation of prior research, which has tended to apply single theoretical perspectives to explain technology-enhanced learning outcomes (Kahu & Nelson, 2018 ). By bridging motivational psychology (SDT), technology adoption research (TAM), and institutional retention theory (Tinto), the present study offers a more holistic account of the mechanisms linking AI-assisted instruction to student persistence. Future research may productively extend this integrated framework by formally testing the mediating pathways hypothesized in the conceptual model—for example, examining whether SDT need satisfaction mediates the relationship between AI platform use and persistence intention, or whether technology acceptance moderates the strength of the SDT pathway. The cross-disciplinary nature of this framework also suggests its applicability beyond music education to other performance-intensive disciplines such as dance, theater, visual arts, and athletics, where practice-based learning, individual skill development, and motivational sustainability are similarly central concerns. 5.3 Practical Implications for Institutions and Educators The findings carry several actionable implications for institutional stakeholders, music department administrators, and piano instructors. First, the results suggest that music departments seeking to reduce attrition may benefit from integrating AI-assisted practice tools as a structured complement to traditional one-to-one instruction. The key phrase here is structured complement: the AI platform in this study did not replace instructor contact but rather enhanced the many hours of independent practice during which students traditionally receive no feedback. Institutions considering such adoption should ensure that AI tools are integrated into the curricular framework with clear usage guidelines, instructor oversight, and explicit connections to learning objectives—rather than being offered as optional add-ons that students must discover and adopt on their own. Second, the importance of the social dashboard feature in supporting relatedness satisfaction highlights the need for AI platforms in music education to include collaborative and community-building functionalities. Technology developers should recognize that music learning, despite its often-solitary practice modality, is fundamentally a social activity embedded in communities of practice (Wenger, 1998 ). Features that enable peer interaction, shared progress tracking, collaborative challenges, and mutual encouragement can help counteract the isolation that is a known driver of music student disengagement (Gaunt, 2010 ; Hatfield et al., 2022 ). Institutions might further amplify these benefits by organizing AI-platform-based practice communities, peer mentoring programs, or weekly sharing sessions where students discuss their AI-mediated practice experiences. Third, the non-significant dropout result, while directionally consistent with the hypothesis and reflecting a threefold reduction in relative risk, underscores that technology alone may be insufficient to eliminate dropout risk. This finding is consistent with Tinto’s ( 2017 ) emphasis that persistence requires not only individual engagement but also institutional commitment to creating supportive environments. Institutions that pair AI-assisted practice tools with comprehensive student support systems—including academic advising, mental health services, financial aid, performance anxiety workshops, and flexible assessment policies—are likely to achieve more substantial reductions in attrition than technology adoption alone can deliver. The AI platform may be most effective as one component within a multi-faceted retention strategy rather than as a standalone solution. Fourth, the finding that practice hours increased substantially in the experimental group has implications for practice pedagogy. Rather than simply telling students to practice more, which research has shown to be an ineffective motivational strategy (Jorgensen & Hallam, 2016 ), providing a feedback-rich practice environment appears to naturally increase practice engagement by making the activity more rewarding, more structured, and more clearly productive. This suggests a shift from quantity-focused practice norms toward quality-focused practice environments supported by intelligent feedback systems. 5.4 Limitations Several limitations of the present study warrant careful consideration and should inform the interpretation of findings as well as the design of future research. First, the quasi-experimental design with intact, pre-existing course sections limits the strength of causal inference. Although baseline equivalence was confirmed across all measured variables, the possibility of unmeasured confounding variables—such as differences in motivation to enroll in one section versus another, instructor preferences, or scheduling-related factors—cannot be entirely excluded. Future studies should employ randomized controlled designs where feasible, or utilize propensity score matching and instrumental variable approaches to strengthen causal claims. Second, the study was conducted at a single comprehensive university in central China, which limits the generalizability of findings to other cultural, institutional, and disciplinary contexts. Chinese music education traditions, student expectations, technology familiarity, and institutional structures may differ meaningfully from those in Western conservatories, liberal arts colleges, or other global contexts. Cross-cultural and multi-site replication studies are needed to determine whether the observed effects are robust across diverse educational settings. Third, the 16-week observation period, while encompassing a full academic semester, may not capture longer-term retention outcomes. Students may sustain engagement gains in the short term but revert to baseline patterns when the intervention is withdrawn or when novelty effects dissipate. Longitudinal studies tracking actual degree completion rates over two to four years would provide more definitive evidence of the intervention’s impact on real-world retention. Fourth, the binary operationalization of dropout risk, while grounded in observable behavioral indicators, may lack the sensitivity needed to detect gradual changes in attrition risk. Future research could benefit from continuous measures of disengagement, latent class growth modeling, or survival analysis approaches that model the timing and trajectory of withdrawal consideration. Fifth, the sample size of 97, while adequate for detecting the large effects observed on continuous outcomes, provided limited statistical power for the dropout risk analysis involving low-frequency binary events. Larger multi-site studies with sample sizes sufficient for subgroup analyses (e.g., by year of study, gender, or prior experience level) are needed to identify potential moderators of the intervention effect. Sixth, the study did not include qualitative data collection methods such as interviews, focus groups, or reflective journals that could illuminate how individual students experienced the AI-mediated practice environment and could reveal nuances in the mechanisms linking technology use to motivational outcomes that quantitative measures alone cannot capture. 5.5 Future Directions Several promising directions emerge for future research. First, mixed-methods designs that combine quantitative outcome measurement with qualitative exploration of student experiences would provide richer and more nuanced understandings of how AI-assisted instruction shapes motivational trajectories. Semi-structured interviews conducted at multiple time points during the intervention could illuminate critical incidents, turning points, and individual differences in how students respond to AI-mediated feedback. Second, research designs that formally test the mediating pathways specified in the conceptual model—using structural equation modeling or multilevel mediation analysis—would advance the field from demonstrating that AI-assisted instruction works to explaining how and why it works. Third, comparative studies examining different AI platform designs, feature sets, and implementation models would help identify the specific design elements that are most responsible for motivational gains, enabling evidence-based design recommendations for technology developers. Fourth, investigations of potential differential effects across student subgroups—such as novice versus advanced students, students with high versus low initial motivation, or students from different socioeconomic backgrounds—would support more targeted and equitable implementation strategies. 6. Conclusion This study provides compelling evidence that AI-assisted piano instruction can meaningfully enhance engagement, satisfaction, and persistence intention among undergraduate music students while supporting the fulfillment of basic psychological needs for autonomy, competence, and relatedness. The large effect sizes observed across multiple outcome variables, the theoretically coherent pattern of SDT need satisfaction gains, and the significant increase in practice hours collectively paint a consistent picture of a technology-enhanced learning environment that transforms the quality of independent practice and strengthens students’ motivational connection to their musical studies. By integrating self-determination theory, the technology acceptance model, and Tinto’s student integration framework into a unified analytical lens, the investigation offers a multidimensional account of the mechanisms through which adaptive technology shapes the motivational and relational conditions that sustain student commitment in performance-intensive educational settings. The practical implications of these findings extend to music departments seeking evidence-based strategies for reducing student attrition, to technology developers designing AI-powered practice tools, and to higher education policymakers concerned with quality and equity in educational outcomes. While the non-significant dropout result cautions against viewing AI-assisted instruction as a sufficient solution to retention challenges, the magnitude and consistency of the motivational and engagement gains suggest that such tools represent a valuable component of comprehensive institutional retention strategies. Future research should pursue longitudinal, multi-site, and mixed-methods designs to extend these findings, test the proposed mediating mechanisms, and identify the conditions under which AI-assisted instruction is most effective for diverse student populations. As higher education systems worldwide grapple with the imperative to provide quality, inclusive, and sustainable education aligned with SDG 4, the intelligent integration of adaptive technology into instrumental music pedagogy offers a promising pathway for supporting student success and mitigating the persistent challenge of disengagement and dropout. Declarations Acknowledgements Funding Statement This research is funded by Han Dan University, under grant number Faculty of Music 2026-005. Conflict of Interest Statement The author declares that there are no conflicts of interest or competing interests regarding the publication of this article. The author has no relevant financial or non-financial interests to disclose. There are no personal, professional, or financial relationships that could potentially be construed as influencing the content or conclusions presented in this work. Ethics Statement The studies involving human participants were reviewed and approved by the Institutional Review Board (or Ethics Committee) of Han Dan University (IRB Protocol No.: Research Office of Handan University 202511120008). The study was conducted in accordance with the local legislation and institutional requirements, as well as the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. Prior to data collection, written informed consent to participate in this study was provided by all participants. Participants were fully informed regarding the study's objectives, procedures, and data handling processes. They were explicitly assured that their participation was entirely voluntary and that they could withdraw from the study at any time without any adverse consequences to their course grades, academic standing, or relationship with the instructors. To ensure confidentiality, all collected data, including pre- and post-test scores and AI platform usage logs, were strictly anonymized and aggregated prior to analysis. Appendix: Data Availability Statement The raw data supporting the conclusions of this article are available from the corresponding author upon reasonable request. Anonymized datasets, including participant-level pre- and post-test scores across all outcome measures, weekly practice logs with practice hours and AI interaction counts for each of the 16 weeks, and individual-level item responses for all survey scales, have been prepared in accordance with institutional data-sharing policies. Three data files are available: (a) Main_Data, containing demographic information and composite scale scores for all 97 participants; (b) Weekly_Practice_Logs, containing 1,552 weekly observations (97 participants × 16 weeks) of practice hours and AI feedback interaction counts; and (c) Scale_Items, containing 194 rows (97 participants × 2 time points) of individual item-level responses for all nine measurement constructs. All data were collected under IRB-approved protocols with written informed consent from all participants, and all identifying information has been removed to protect participant privacy. References Al-Emran, M., Mezhuyev, V., & Kamaludin, A. (2018). Technology acceptance model in M-learning context: A systematic review. Computers & Education, 125, 389–412. https://doi.org/10.1016/j.compedu.2018.06.008 Bandura, A. (1997). Self-efficacy: The exercise of control. W. H. Freeman. Bonneville-Roussy, A., Evans, P., Verner-Filion, J., Vallerand, R. J., & Bouffard, T. (2017). Motivation and coping with the stress of assessment: Gender differences in outcomes for university students. Contemporary Educational Psychology, 48, 28–42. https://doi.org/10.1016/j.cedpsych.2016.08.003 Brislin, R. W. (1970). Back-translation for cross-cultural research. Journal of Cross-Cultural Psychology, 1(3), 185–216. https://doi.org/10.1177/135910457000100301 Burt-Perkins, R., & Mills, J. (2019). The role of the first-year in undergraduate music students’ decisions to persist. British Journal of Music Education, 36(1), 33–45. https://doi.org/10.1017/S0265051718000256 Chen, B., Vansteenkiste, M., Beyers, W., Boone, L., Deci, E. L., Van der Kaap-Deeder, J., ... & Verstuyf, J. (2015). Basic psychological need satisfaction, need frustration, and need strength across four cultures. Motivation and Emotion, 39(2), 216–236. https://doi.org/10.1007/s11031-014-9450-1 Chen, X., Xie, H., Zou, D., & Hwang, G. J. (2022). Application and theory gaps during the rise of artificial intelligence in education. Computers and Education: Artificial Intelligence, 1, 100002. https://doi.org/10.1016/j.caeai.2020.100002 Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Lawrence Erlbaum Associates. Crawford, R. (2021). Lessons from the pandemic: What music education can learn from the COVID-19 crisis. International Journal of Music Education, 39(3), 277–293. https://doi.org/10.1177/02557614211024163 Creech, A., Papageorgi, I., Duffy, C., Morton, F., Haddon, E., Potter, J., ... & Welch, G. (2008). From music student to professional: The process of transition. British Journal of Music Education, 25(3), 315–331. https://doi.org/10.1017/S0265051708008127 Davidson, W. B., Beck, H. P., & Milligan, M. (2009). The college persistence questionnaire: Development and validation of an instrument that predicts student attrition. Journal of College Student Development, 50(4), 373–390. https://doi.org/10.1353/csd.0.0079 Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319–340. https://doi.org/10.2307/249008 Dye, K. (2022). Online piano pedagogy: Perspectives and practices during COVID-19. Music Education Research, 24(2), 176–189. https://doi.org/10.1080/14613808.2022.2032735 Evans, P. (2015). Self-determination theory: An approach to motivation in music education. Musicae Scientiae, 19(1), 65–83. https://doi.org/10.1177/1029864914568044 Gaunt, H. (2010). One-to-one tuition in a conservatoire: The perceptions of instrumental and vocal students. Psychology of Music, 38(2), 178–208. https://doi.org/10.1177/0305735609339467 Ginsborg, J., Kreutz, G., Thomas, M., & Williamon, A. (2009). Healthy behaviours in music and non-music performance students. Health Education, 109(3), 242–258. https://doi.org/10.1108/09654280910955575 Handelsman, M. M., Briggs, W. L., Sullivan, N., & Towler, A. (2005). A measure of college student course engagement. The Journal of Educational Research, 98(3), 184–192. https://doi.org/10.3200/JOER.98.3.184-192 Hatfield, J. L., Halvari, H., & Deci, E. L. (2022). Motivation, anxiety, and well-being among music academy students: The role of basic psychological need satisfaction. Frontiers in Psychology, 13, 891758. https://doi.org/10.3389/fpsyg.2022.891758 Hattie, J. (2009). Visible learning: A synthesis of over 800 meta-analyses relating to achievement. Routledge. Huang, A. Y., & Chen, N. S. (2023). AI in music education: A systematic review of intelligent tutoring systems and adaptive learning. Education and Information Technologies, 28(5), 5973–6002. https://doi.org/10.1007/s10639-022-11406-7 Hwang, G. J., Xie, H., Wah, B. W., & Gašević, D. (2020). Vision, challenges, roles and research issues of artificial intelligence in education. Computers and Education: Artificial Intelligence, 1, 100001. https://doi.org/10.1016/j.caeai.2020.100001 Jorgensen, H., & Hallam, S. (2016). Practicing. In S. Hallam, I. Cross, & M. Thaut (Eds.), The Oxford handbook of music psychology (2nd ed., pp. 449–462). Oxford University Press. Kahu, E. R., & Nelson, K. (2018). Student engagement in the educational interface: Understanding the mechanisms of student success. Higher Education Research & Development, 37(1), 58–71. https://doi.org/10.1080/07294360.2017.1344197 Kenny, D. T. (2011). The psychology of music performance anxiety. Oxford University Press. Miksza, P., Evans, P., & McPherson, G. E. (2022). Motivation to pursue music: A self-determination theory perspective. In G. E. McPherson (Ed.), The Oxford handbook of music performance (Vol. 2, pp. 567–591). Oxford University Press. Niemiec, C. P., & Ryan, R. M. (2009). Autonomy, competence, and relatedness in the classroom: Applying self-determination theory to educational practice. Theory and Research in Education, 7(2), 133–144. https://doi.org/10.1177/1477878509104318 Ryan, R. M., & Deci, E. L. (2020). Intrinsic and extrinsic motivation from a self-determination theory perspective: Definitions, theory, practices, and future directions. Contemporary Educational Psychology, 61, 101860. https://doi.org/10.1016/j.cedpsych.2020.101860 Scherer, R., Siddiq, F., & Tondeur, J. (2019). The technology acceptance model (TAM): A meta-analytic structural equation modeling approach to explaining teachers’ adoption of digital technology in education. Computers & Education, 128, 13–35. https://doi.org/10.1016/j.compedu.2018.09.009 Schreiner, L. A., & Juillerat, S. L. (1994). Student satisfaction inventory. Noel-Levitz. Tinto, V. (1993). Leaving college: Rethinking the causes and cures of student attrition (2nd ed.). University of Chicago Press. Tinto, V. (2017). Through the eyes of students. Journal of College Student Retention: Research, Theory & Practice, 19(3), 254–269. https://doi.org/10.1177/1521025115621917 Venkatesh, V., & Davis, F. D. (2000). A theoretical extension of the technology acceptance model: Four longitudinal field studies. Management Science, 46(2), 186–204. https://doi.org/10.1287/mnsc.46.2.186.11926 Wenger, E. (1998). Communities of practice: Learning, meaning, and identity. Cambridge University Press. Xie, H., Chu, H. C., Hwang, G. J., & Wang, C. C. (2022). Trends and development in technology-enhanced adaptive/personalized learning: A systematic review of journal publications from 2007 to 2021. Computers & Education, 140, 103599. https://doi.org/10.1016/j.compedu.2019.103599 Zhai, X., Chu, X., Chai, C. S., Jong, M. S., Istenic, A., Spector, M., ... & Li, Y. (2021). A review of artificial intelligence (AI) in education from 2010 to 2020. Complexity, 2021, 8812542. https://doi.org/10.1155/2021/8812542 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-9485161","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":627094245,"identity":"172c9c83-b6f3-47ed-9e69-5ff9c963dcd3","order_by":0,"name":"LIU CHANG","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7UlEQVRIiWNgGAWjYBACPgb+h49/VPyrl+dvPgDkS8gQ1MLGwMNszHDmQILhjGMJIC08xGhhk2ZsO5DAcCDHACRAhBb2s4eNC87cyWNsOPP51Y0aCx4G9sNHN+DVwpOX+HhGxbNidubebdY5x4AO40lLu4FXiwSDsQHPGWbGxoaz24xz2IBaJHjMCGkxk+BtY2ZsOJDzzDjnH1FaeMykedsOJwK1MD/ObSNGC09asuGMM2nGwEA2Y87tk+BhI+QXfvbDBx98qLCRA0bl48853+rkgCLH8GpB8xeIJFY5CDB/IEX1KBgFo2AUjBwAANqSSXWGMF7wAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0003-3162-2358","institution":"Han Dan University","correspondingAuthor":true,"prefix":"","firstName":"LIU","middleName":"","lastName":"CHANG","suffix":""}],"badges":[],"createdAt":"2026-04-21 13:40:38","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-9485161/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9485161/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107518511,"identity":"be056355-366c-457e-90c0-633d1310bf1c","added_by":"auto","created_at":"2026-04-22 08:47:51","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":921631,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eIntegrated theoretical framework combining SDT, TAM, and Tinto’s integration model.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-9485161/v1/bdca2d268b295e95f0f63872.png"},{"id":107705232,"identity":"bfd61ca6-1983-4eb7-abc3-924911a108dd","added_by":"auto","created_at":"2026-04-24 09:09:50","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":101409,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eConceptual model with hypothesized relationships among AI-assisted instruction, SDT mediators, and student outcomes.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage21.png","url":"https://assets-eu.researchsquare.com/files/rs-9485161/v1/7fd60a9dbbc2fa1fb340981f.png"},{"id":107518513,"identity":"5280c8b6-bbc3-41ca-8ac9-71816fc1a732","added_by":"auto","created_at":"2026-04-22 08:47:51","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":58534,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eQuasi-experimental research design overview showing the 16-week timeline.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-9485161/v1/016756968845015022218d10.png"},{"id":107705703,"identity":"2b1583de-ce99-4eb2-a9eb-6e826accdfad","added_by":"auto","created_at":"2026-04-24 09:14:41","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":284711,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003eComparative outcomes of AI-assisted versus traditional piano instruction across four dimensions. (A) Pre-to-post mean gain scores by group; (B) Post-test SDT need satisfaction radar profiles; (C) Pre-to-post trajectories for engagement, persistence, and satisfaction; (D) Dropout risk distribution.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-9485161/v1/236bcba9529a4749586d4e83.png"},{"id":107710347,"identity":"d294ef60-d6ac-45c0-8efd-0d899cb525c8","added_by":"auto","created_at":"2026-04-24 09:40:24","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1463422,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9485161/v1/5af0bf06-e577-4274-a005-00d5b266b91f.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eMitigating Disengagement in Higher Music Education:The Role of AI-Assisted Piano Instruction in Sustaining Student Persistence\u003c/p\u003e","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eHigher education institutions worldwide face persistent and multifaceted challenges in retaining students through degree completion. Despite decades of research on student attrition, dropout rates remain stubbornly high across many disciplines, with consequences that extend beyond individual students to affect institutional resources, program viability, and national human capital development (Tinto, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). The phenomenon of student dropout is particularly complex because it emerges from the intersection of individual characteristics, academic experiences, institutional environments, and broader socioeconomic conditions, making it resistant to simple interventions or one-size-fits-all solutions. Within this landscape, music programs confront distinctive retention pressures that have received comparatively limited empirical attention, despite the demanding nature of conservatory and university-level musical training.\u003c/p\u003e \u003cp\u003eMusic education at the tertiary level differs fundamentally from most academic disciplines in its reliance on extended, solitary practice as the primary mechanism of skill development. Piano students, in particular, typically spend between 15 and 25 hours per week in independent practice, a largely unsupervised activity that requires sustained concentration, self-regulation, and intrinsic motivation (Jorgensen \u0026amp; Hallam, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). During these practice sessions, students must diagnose their own technical and interpretive weaknesses, devise strategies for improvement, and maintain the emotional resilience to persist through plateaus and setbacks\u0026mdash;all without the immediate guidance of an instructor. This pedagogical structure creates fertile conditions for disengagement, as students who lack adequate feedback mechanisms, clear progress indicators, or a sense of connectedness to their musical community may gradually withdraw their effort, reduce their practice time, and ultimately question their commitment to the degree program (Creech et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2008\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eEmpirical evidence from European conservatories and North American music programs suggests that dropout rates in music higher education range from 15% to 30%, with attrition concentrated in the first two years of study (Burt-Perkins \u0026amp; Mills, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Factors contributing to music student dropout include performance anxiety, burnout from excessive practice demands, perceived mismatch between expectations and reality, financial pressures associated with instrument maintenance and lesson costs, competitive studio cultures that undermine belonging, and insufficient institutional support for students experiencing motivational crises (Bonneville-Roussy et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2017\u003c/span\u003e; Hatfield et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Importantly, these factors often interact in complex, nonlinear ways: a student experiencing performance anxiety may reduce practice, leading to diminished competence perceptions, which further erodes motivation and ultimately triggers withdrawal consideration. Understanding and interrupting these cascading disengagement trajectories requires interventions that address multiple dimensions of the student experience simultaneously.\u003c/p\u003e \u003cp\u003eThe rapid integration of artificial intelligence into educational settings has prompted renewed interest in whether technology-mediated instruction can alleviate patterns of disengagement that traditional pedagogical approaches alone have struggled to resolve (Chen et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Hwang et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). AI-driven tools in music education now offer capabilities that were unimaginable a decade ago: real-time acoustic analysis of pitch accuracy, rhythm precision, and dynamic expression; machine learning algorithms that generate personalized practice recommendations based on individual performance trajectories; adaptive difficulty calibration that ensures students are consistently challenged without being overwhelmed; and social features that connect learners to peer communities for collaborative feedback and mutual encouragement (Huang \u0026amp; Chen, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Xie et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Unlike conventional self-assessment methods such as metronome-based practice or self-recording, AI-powered platforms provide immediate, detailed, and objective diagnostics that approximate elements of expert mentorship during the many hours when students practice independently.\u003c/p\u003e \u003cp\u003eDespite growing enthusiasm for AI applications in education, the empirical base linking AI-assisted instrumental instruction to student persistence outcomes remains limited, particularly within higher education contexts. The majority of existing studies on AI in music education focus on skill acquisition outcomes\u0026mdash;such as improved pitch accuracy or sight-reading speed\u0026mdash;rather than motivational, affective, or retention-related variables (Zhai et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Furthermore, most technology-enhanced learning research has been conducted with K\u0026ndash;12 populations or in general academic subjects such as mathematics and language learning, leaving a significant gap in understanding how AI integration functions within the specialized pedagogical environment of university-level music training. The one-to-one lesson format, the centrality of practice as a learning modality, and the high degree of autonomy required of music students create a context that may respond to AI-assisted instruction in ways that differ meaningfully from classroom-based disciplines (Gaunt, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2010\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eA further gap in the literature concerns the theoretical mechanisms through which technology-enhanced instruction influences persistence. Relatively few investigations have situated AI-assisted music instruction within established motivational and retention frameworks capable of explaining how technological support translates into sustained engagement and reduced attrition (Kahu \u0026amp; Nelson, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Without such theoretical grounding, empirical findings risk remaining descriptive rather than explanatory, limiting their utility for designing principled interventions. The present study addresses this gap by integrating three complementary theoretical perspectives: self-determination theory (SDT; Ryan \u0026amp; Deci, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2020\u003c/span\u003e), which explains how environments that support autonomy, competence, and relatedness foster intrinsic motivation; the technology acceptance model (TAM; Venkatesh \u0026amp; Davis, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2000\u003c/span\u003e), which identifies the cognitive conditions under which students adopt and sustain use of technological tools; and Tinto\u0026rsquo;s student integration model (Tinto, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e1993\u003c/span\u003e, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2017\u003c/span\u003e), which contextualizes how academic and social integration within institutional environments promotes persistence to degree completion.\u003c/p\u003e \u003cp\u003eThis study therefore investigates the effects of a 16-week AI-assisted piano instruction intervention on undergraduate music students\u0026rsquo; engagement, persistence intention, satisfaction, and dropout risk, while examining whether AI-mediated instruction enhances the fulfillment of basic psychological needs as theorized by SDT. The central research questions are: (RQ1) Does AI-assisted piano instruction produce significantly higher post-intervention scores on engagement, persistence intention, and satisfaction compared with traditional instruction? (RQ2) Does AI-assisted instruction enhance students\u0026rsquo; perceived autonomy, competence, and relatedness? (RQ3) Is dropout risk lower among students receiving AI-assisted instruction? These questions carry practical significance for institutions seeking evidence-based strategies to support music student retention, and for the broader conversation around equitable, technology-enhanced quality education aligned with Sustainable Development Goal 4.\u003c/p\u003e"},{"header":"2. Literature Review and Theoretical Framework","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Student Disengagement and Dropout in Music Higher Education\u003c/h2\u003e \u003cp\u003eResearch on student attrition in music programs has identified a distinctive combination of academic, psychosocial, and institutional factors that set music education apart from other disciplines. Performance-based assessment, where students\u0026rsquo; competence is publicly evaluated in recitals and jury examinations, creates high-stakes evaluative environments that can produce debilitating anxiety and undermine the intrinsic enjoyment that initially drew students to music (Kenny, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). The intense practice demands of conservatory-style training, often exceeding 20 hours per week of solitary instrumental work, can lead to physical injuries such as repetitive strain and focal dystonia, as well as psychological burnout characterized by emotional exhaustion, depersonalization, and reduced personal accomplishment (Ginsborg et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). Competitive studio cultures, in which students are ranked against their peers for performance opportunities and teacher attention, can erode the sense of belonging and relatedness that is critical for sustained engagement (Bonneville-Roussy et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eLongitudinal data from European conservatories and university music departments indicate that between 15% and 30% of enrolled music students leave their programs before completion, with the highest attrition rates occurring during the transition from the first to second year of study (Burt-Perkins \u0026amp; Mills, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Students who withdraw often report a combination of diminished motivation, feelings of isolation during practice, inadequate or infrequent feedback from instructors, financial strain, and a growing disconnect between their initial musical aspirations and the realities of professional training (Gaunt, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Hatfield et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Notably, these factors frequently operate in mutually reinforcing cycles: insufficient feedback leads to stagnating skill development, which diminishes competence perceptions, which reduces practice motivation, which further impairs progress\u0026mdash;a downward spiral that, left unaddressed, culminates in withdrawal. The challenge for music educators and institutions is to identify leverage points within these cycles where targeted interventions can interrupt disengagement trajectories and restore positive motivational dynamics.\u003c/p\u003e \u003cp\u003eRecent shifts toward hybrid and online learning, accelerated by the global pandemic, have introduced additional challenges for music students. Virtual instruction often fails to capture the nuances of acoustic performance, creating frustrations for both students and instructors who rely on real-time sonic feedback during lessons (Dye, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Digital fatigue, reduced peer interaction, and the absence of shared performance spaces have exacerbated feelings of isolation and disconnection among music students (Crawford, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). These developments underscore the urgency of exploring how technology can be deployed not merely as a substitute for in-person instruction but as a complement that addresses the specific motivational and relational needs of music learners in contemporary higher education environments.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Self-Determination Theory and Music Learning\u003c/h2\u003e \u003cp\u003eSelf-determination theory (SDT) provides a comprehensive framework for understanding human motivation as shaped by the satisfaction of three innate and universal psychological needs: autonomy, competence, and relatedness (Ryan \u0026amp; Deci, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2020\u003c/span\u003e). Autonomy refers to the experience of volitional action and self-endorsement\u0026mdash;the sense that one\u0026rsquo;s behavior originates from the self rather than from external pressures. Competence involves the experience of effectance and mastery\u0026mdash;the perception that one is capable of achieving desired outcomes and growing in skill. Relatedness encompasses the feeling of connection, belonging, and care within social environments\u0026mdash;the sense that one matters to others and is part of a meaningful community. According to SDT, environments that support the satisfaction of these three needs foster the internalization of motivation, promote deeper engagement with learning activities, enhance well-being, and increase the likelihood of sustained participation over time (Niemiec \u0026amp; Ryan, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2009\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eWithin music education, SDT-informed research has demonstrated that studio environments promoting learner choice in repertoire selection, constructive and process-oriented feedback, and opportunities for peer collaboration are associated with higher practice motivation, greater intrinsic interest in musical activities, and reduced attrition (Evans, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Miksza et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Conversely, controlling teaching styles that emphasize compliance, performance outcomes over learning processes, and normative comparison among students tend to undermine need satisfaction and promote external forms of motivation that are associated with anxiety, disengagement, and dropout (Bonneville-Roussy et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). The theoretical prediction of SDT\u0026mdash;that need-supportive environments promote persistence while need-thwarting environments promote withdrawal\u0026mdash;maps directly onto the observed dynamics of music student retention.\u003c/p\u003e \u003cp\u003eAI-assisted instruction has the potential to enhance need satisfaction through several mechanisms. Autonomy may be supported when AI platforms allow learners to set their own practice goals, choose the order and pace of exercises, and receive individualized recommendations rather than standardized assignments. Competence is enhanced when real-time feedback provides precise, actionable information about performance quality, enabling students to identify specific areas for improvement and to perceive tangible progress across practice sessions. Relatedness can be supported through social features such as peer dashboards, collaborative challenges, shared recordings, and community forums that maintain a sense of connection even during solitary practice (Xie et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The present study tests whether these theorized mechanisms are empirically supported in the context of university-level piano instruction.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Technology Acceptance and AI in Education\u003c/h2\u003e \u003cp\u003eThe technology acceptance model (TAM), originally proposed by Davis (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e1989\u003c/span\u003e) and subsequently extended by Venkatesh and Davis (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2000\u003c/span\u003e), identifies perceived usefulness and perceived ease of use as the primary cognitive determinants of technology adoption behavior. Perceived usefulness refers to the degree to which a person believes that using a particular technology would enhance their performance, while perceived ease of use refers to the degree to which a person believes that using the technology would be free of effort. When students perceive an AI-assisted practice tool as both useful for improving their musical skills and easy to integrate into their existing practice routines, they are more likely to adopt the technology, use it consistently, and derive motivational and learning benefits from it (Hwang et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn higher education contexts, TAM has been extensively applied to explain students\u0026rsquo; acceptance of learning management systems, intelligent tutoring platforms, and AI-based educational tools across diverse disciplinary settings (Al-Emran et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Scherer et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). A meta-analytic review by Scherer et al. (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) found that perceived usefulness consistently exhibited the strongest relationship with technology acceptance among educational stakeholders, while perceived ease of use operated both directly and indirectly through its influence on perceived usefulness. For AI-assisted music instruction, these findings suggest that the design of the technological interface matters: platforms that are intuitive, responsive, and clearly connected to musical learning goals are more likely to be adopted and sustained by students, thereby maximizing the motivational benefits of AI-mediated feedback.\u003c/p\u003e \u003cp\u003eTAM complements SDT by addressing a prerequisite condition for need satisfaction: if students do not accept and consistently use the AI tool, the potential benefits for autonomy, competence, and relatedness cannot be realized. In this sense, technology acceptance serves as a gateway mechanism that must be satisfied before the motivational dynamics predicted by SDT can operate. The integration of TAM with SDT in the present study provides a more complete account of the pathway from technology introduction to motivational outcomes to persistence behavior.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4 Tinto\u0026rsquo;s Integration Model and Institutional Persistence\u003c/h2\u003e \u003cp\u003eTinto\u0026rsquo;s model of student integration (Tinto, \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e1993\u003c/span\u003e, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) has been among the most influential frameworks in higher education retention research for over four decades. The model posits that students persist in their educational programs to the extent that they achieve sufficient academic and social integration within their institutional environment. Academic integration encompasses intellectual development, satisfactory academic performance, and meaningful interactions with academic systems and personnel, including faculty, advisors, and curricula. Social integration involves the development of interpersonal connections, peer relationships, and a sense of belonging within the broader campus community. When students experience high levels of both academic and social integration, they develop stronger institutional commitment and goal commitment, which in turn increase the likelihood of persistence to degree completion.\u003c/p\u003e \u003cp\u003eIn music departments, the primary sites of academic and social integration include the one-to-one lesson with a studio teacher, ensemble rehearsals and performances, peer interactions in practice rooms and common areas, and participation in masterclasses and departmental events (Gaunt, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). The studio teacher relationship is often the single most important factor in music students\u0026rsquo; institutional integration, serving simultaneously as a source of technical instruction, artistic mentorship, emotional support, and professional socialization (Creech et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). However, this relationship is typically limited to one or two contact hours per week, leaving students to navigate the majority of their practice time without direct guidance or social connection.\u003c/p\u003e \u003cp\u003eAI-assisted tools may contribute to academic integration by structuring independent practice in ways that deepen skill acquisition, by providing data-rich performance summaries that enrich teacher\u0026ndash;student dialogue during lessons, and by helping students set and track meaningful learning goals that connect practice to broader curricular objectives. They may also support social integration when collaborative features enable students to share recordings, comment on peers\u0026rsquo; progress, participate in group challenges, and maintain ongoing communication with fellow learners within the platform. In Tinto\u0026rsquo;s (\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) updated framework, which emphasizes the importance of self-efficacy, belonging, and perceived relevance of the curriculum, AI-assisted instruction may function as a tool that enhances all three of these conditions by making practice more productive, more connected, and more clearly aligned with students\u0026rsquo; evolving musical goals.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.5 Hypotheses\u003c/h2\u003e \u003cp\u003eBuilding on the integrated theoretical framework depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, the following hypotheses were formulated to guide the empirical investigation. These hypotheses reflect the theorized mechanisms through which AI-assisted piano instruction influences motivational processes and retention-relevant outcomes:\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eHypothesis 1\u003c/strong\u003e \u003cp\u003e \u003cb\u003e(H1a\u0026ndash;H1c)\u003c/b\u003e: Students receiving AI-assisted piano instruction will report significantly higher post-intervention levels of perceived autonomy (H1a), competence (H1b), and relatedness (H1c) compared with students receiving traditional instruction, after controlling for pre-test scores.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eHypothesis 2\u003c/strong\u003e \u003cp\u003e \u003cb\u003e(H2a\u0026ndash;H2c)\u003c/b\u003e: AI-assisted instruction will produce significantly greater improvements in composite engagement (H2a), persistence intention (H2b), and course satisfaction (H2c) relative to traditional instruction, after controlling for pre-test scores.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eHypothesis 3\u003c/strong\u003e \u003cp\u003e \u003cb\u003e(H3)\u003c/b\u003e: The proportion of students classified as at risk of dropout will be significantly lower in the AI-assisted group than in the control group at the end of the 16-week intervention period.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"3. Methods","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Research Design and Context\u003c/h2\u003e \u003cp\u003eA quasi-experimental pre-test\u0026ndash;post-test design with a non-equivalent control group was employed to evaluate the effects of AI-assisted piano instruction on student engagement, persistence, satisfaction, and dropout risk. This design was selected because random assignment at the individual level was not feasible within the existing course structure of the participating institution, where students were pre-enrolled in designated sections of the piano performance course based on scheduling availability and prior year of study (see Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Two sections of an undergraduate piano performance course offered at a comprehensive public university in central China during the Fall 2024 semester were designated as the experimental and control conditions. The study spanned the full 16-week semester, with baseline data collected during Week 0 (prior to the start of the intervention) and post-test data collected during Week 16 (following the completion of the intervention period in Week 14, with a two-week buffer for post-testing). Ethical approval was obtained from the university\u0026rsquo;s institutional review board (IRB Protocol No. [to be inserted]), and all participants provided written informed consent prior to data collection. Students were informed that participation was voluntary and that their decision to participate or withdraw would have no impact on their course grades or academic standing.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Participants\u003c/h2\u003e \u003cp\u003eThe sample comprised 97 undergraduate students enrolled in piano performance courses at the participating university. The experimental group (n\u0026thinsp;=\u0026thinsp;49) received AI-assisted piano instruction supplemented with a commercial adaptive practice platform, while the control group (n\u0026thinsp;=\u0026thinsp;48) received traditional piano instruction following the same curricular objectives and syllabus without AI supplementation. An a priori power analysis conducted using G*Power (version 3.1) indicated that a sample size of 90 participants (45 per group) would provide 80% power to detect a medium-to-large effect size (d\u0026thinsp;=\u0026thinsp;0.60) at an alpha level of .05 using independent-samples t-tests. The obtained sample of 97 exceeded this threshold, ensuring adequate statistical power for the primary analyses.\u003c/p\u003e \u003cp\u003eParticipants ranged in age from 18 to 23 years (M\u0026thinsp;=\u0026thinsp;19.67, SD\u0026thinsp;=\u0026thinsp;1.36). The sample included 43 male students (44.3%) and 54 female students (55.7%), a gender distribution consistent with enrollment patterns in Chinese music programs. Participants represented all four undergraduate years: first-year (n\u0026thinsp;=\u0026thinsp;29, 29.9%), second-year (n\u0026thinsp;=\u0026thinsp;34, 35.1%), third-year (n\u0026thinsp;=\u0026thinsp;21, 21.6%), and fourth-year (n\u0026thinsp;=\u0026thinsp;13, 13.4%). Prior piano experience ranged from 0.5 to 8.0 years (M\u0026thinsp;=\u0026thinsp;4.33, SD\u0026thinsp;=\u0026thinsp;2.37), reflecting the diversity typical of university music programs that admit students with varying levels of pre-collegiate training. Cumulative grade point averages ranged from 2.00 to 4.00 (M\u0026thinsp;=\u0026thinsp;3.21, SD\u0026thinsp;=\u0026thinsp;0.43). Baseline comparisons using independent-samples t-tests confirmed that the two groups did not differ significantly on any demographic variable or pre-test measure (all p values \u0026gt; .05), supporting the assumption of baseline equivalence necessary for valid quasi-experimental inference.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Intervention\u003c/h2\u003e \u003cp\u003eBoth the experimental and control groups received two 50-minute individual piano lessons per week from the same pool of four qualified piano instructors, all of whom held master\u0026rsquo;s degrees or higher in piano performance and had a minimum of five years of university-level teaching experience. Instructors were randomly assigned to teach both experimental and control students to minimize instructor effects. The instructional syllabus, shared across both conditions, covered technical exercises including scales, arpeggios, and \u0026eacute;tudes; repertoire development spanning Baroque, Classical, Romantic, and contemporary periods; sight-reading proficiency; and expressive performance skills including dynamics, phrasing, and pedaling technique. Lesson content, sequencing, and assessment criteria were standardized through a jointly developed syllabus document and biweekly instructor meetings to ensure fidelity of implementation across both conditions.\u003c/p\u003e \u003cp\u003eThe experimental group additionally used an AI-powered piano practice application during their independent practice sessions throughout the 14-week intervention period (Weeks 1\u0026ndash;14). The application, a commercially available platform widely used in Chinese music education institutions, incorporated the following core functionalities: (a) real-time pitch and rhythm analysis through acoustic signal processing that detected deviations from the score and provided immediate visual feedback on a tablet display; (b) technique-specific diagnostics generated by machine learning algorithms trained on a large corpus of expert performance data, offering targeted suggestions for finger positioning, hand shape, wrist relaxation, and articulation; (c) adaptive practice scheduling that automatically adjusted exercise difficulty, tempo, and duration based on the learner\u0026rsquo;s performance trajectory across sessions, ensuring an appropriate balance between challenge and mastery; and (d) a social dashboard feature that enabled students to share practice clips, view anonymized peer progress summaries, post encouraging comments, and participate in weekly practice challenges organized by the research team.\u003c/p\u003e \u003cp\u003eStudents in the experimental group were instructed to use the AI platform during at least 60% of their weekly independent practice hours and to log all practice sessions through the application\u0026rsquo;s built-in tracking system. Compliance was monitored through weekly review of platform usage data by the research team, and students who fell below the 60% threshold for two consecutive weeks received a brief motivational reminder. The average compliance rate across the 14-week intervention period was 78.3% (SD\u0026thinsp;=\u0026thinsp;11.6%), indicating that most students substantially exceeded the minimum usage requirement. Control-group students practiced independently using conventional methods, including metronomes, self-recording on personal devices, written instructor notes, and commercially available score-following applications that did not incorporate AI-based adaptive feedback. Control-group students were not prohibited from using any technology other than the specific AI platform employed in the experimental condition.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Instruments\u003c/h2\u003e \u003cp\u003eAll instruments employed five-point Likert response scales (1\u0026thinsp;=\u0026thinsp;strongly disagree to 5\u0026thinsp;=\u0026thinsp;strongly agree) unless otherwise noted. Instruments were originally developed in English and translated into Mandarin Chinese following standard back-translation procedures (Brislin, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e1970\u003c/span\u003e). The translated instruments were piloted with a separate sample of 30 music students (not included in the study sample) to verify clarity, cultural appropriateness, and psychometric adequacy. Minor wording adjustments were made based on pilot feedback. Instruments were administered online via a secure survey platform at both the pre-test (Week 0) and post-test (Week 16) time points, with completion typically requiring 15\u0026ndash;20 minutes.\u003c/p\u003e \u003cp\u003e \u003cb\u003eStudent Engagement.\u003c/b\u003e Engagement was measured using an adapted version of the Student Course Engagement Questionnaire (SCEQ; Handelsman et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2005\u003c/span\u003e), modified for music education contexts through consultation with three music education researchers and two piano instructors. The adapted instrument comprised four subscales: intrinsic motivation (6 items; sample item: I find the practice activities genuinely interesting; pre-test α\u0026thinsp;=\u0026thinsp;.90), behavioral engagement (6 items; sample item: I consistently put effort into my piano practice sessions; α\u0026thinsp;=\u0026thinsp;.89), emotional engagement (6 items; sample item: I feel enthusiastic when practicing piano; α\u0026thinsp;=\u0026thinsp;.92), and cognitive engagement (5 items; sample item: I try to connect new techniques with what I already know; α\u0026thinsp;=\u0026thinsp;.85). A composite engagement score was computed as the unweighted mean of the four subscale scores.\u003c/p\u003e \u003cp\u003e \u003cb\u003eBasic Psychological Need Satisfaction.\u003c/b\u003e Perceived autonomy (5 items; sample item: I feel I have choices in how I approach my piano practice; α\u0026thinsp;=\u0026thinsp;.81), competence (6 items; sample item: I feel capable of mastering challenging passages; α\u0026thinsp;=\u0026thinsp;.87), and relatedness (5 items; sample item: I feel connected to other students in the piano program; α\u0026thinsp;=\u0026thinsp;.83) were assessed using the Basic Psychological Need Satisfaction Scale adapted for educational technology contexts (Chen et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). A composite SDT score was computed as the mean of the three subscale scores.\u003c/p\u003e \u003cp\u003e \u003cb\u003ePersistence Intention.\u003c/b\u003e Persistence was measured using a 6-item scale adapted from the College Persistence Questionnaire (Davidson et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2009\u003c/span\u003e), modified to reflect the specific context of music degree programs. Items assessed students\u0026rsquo; intention to continue their music studies through degree completion, their commitment to the program in the face of difficulties, and their confidence in finishing the degree. A sample item reads: I am committed to completing my music degree even when coursework feels overwhelming. Internal consistency was excellent (α\u0026thinsp;=\u0026thinsp;.91).\u003c/p\u003e \u003cp\u003e \u003cb\u003eCourse Satisfaction.\u003c/b\u003e Satisfaction was measured using an 8-item instrument adapted from the Student Satisfaction Inventory (Schreiner \u0026amp; Juillerat, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e1994\u003c/span\u003e), tailored to reflect music course contexts. Items addressed satisfaction with instructional quality, feedback adequacy, practice resources, course organization, and overall learning experience. Internal consistency was good (α\u0026thinsp;=\u0026thinsp;.88).\u003c/p\u003e \u003cp\u003e \u003cb\u003eDropout Risk.\u003c/b\u003e Dropout risk was operationalized as a binary variable (0\u0026thinsp;=\u0026thinsp;not at risk, 1\u0026thinsp;=\u0026thinsp;at risk) indicating whether a student exhibited two or more of the following behavioral indicators by the end of the semester: (a) cumulative attendance below 75% of scheduled lessons; (b) failure to submit two or more required practice logs during the intervention period; (c) documented expression of intention to withdraw, transfer, or take a leave of absence as recorded in academic advising records or instructor communications. This multi-indicator approach was adopted to increase the ecological validity of the dropout risk measure beyond single-item self-report scales, which may be subject to social desirability bias.\u003c/p\u003e \u003cp\u003e \u003cb\u003eWeekly Practice Logs.\u003c/b\u003e Students in both groups maintained weekly practice logs documenting total practice hours and practice activities for each of the 16 weeks. Experimental-group students\u0026rsquo; practice hours were automatically recorded by the AI platform, while control-group students submitted self-reported logs through the course management system. Additionally, the number of AI feedback interactions per practice session was automatically recorded for experimental-group students, providing a measure of technology engagement intensity.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e3.5 Data Analysis\u003c/h2\u003e \u003cp\u003eData were analyzed using IBM SPSS Statistics (Version 28) and R (Version 4.3.2). The analytical strategy proceeded in three stages. First, independent-samples t-tests and chi-square tests were conducted to verify baseline equivalence between the experimental and control groups on all demographic variables and pre-test outcome measures. Second, the primary hypotheses (H1\u0026ndash;H3) were tested using independent-samples t-tests to compare post-test means between groups, with Cohen\u0026rsquo;s d calculated as the standardized mean difference using the pooled standard deviation. Effect sizes were interpreted according to conventional benchmarks: small (d\u0026thinsp;=\u0026thinsp;0.20), medium (d\u0026thinsp;=\u0026thinsp;0.50), and large (d\u0026thinsp;=\u0026thinsp;0.80; Cohen, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e1988\u003c/span\u003e). Ninety-five percent confidence intervals for all effect sizes were computed using the noncentral t-distribution method. The chi-square test of independence was used to compare the proportion of students classified as at risk of dropout between groups. Third, supplementary analyses examined weekly practice hours and AI interaction patterns across the 16-week period to provide additional context for interpreting the primary findings. All tests used a two-tailed significance threshold of α\u0026thinsp;=\u0026thinsp;.05. Given the family-wise error rate concerns associated with multiple comparisons across outcome variables, a Bonferroni-corrected threshold of p \u0026lt; .007 (for seven primary comparisons) was also applied as a sensitivity check, and all significant results except one remained significant under this more conservative criterion.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Results","content":"\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Baseline Equivalence\u003c/h2\u003e \u003cp\u003eIndependent-samples t-tests confirmed that the experimental and control groups did not differ significantly at the pre-test time point on any measured variable. Results indicated no significant group differences for age (t(95)\u0026thinsp;=\u0026thinsp;0.32, p = .747, d\u0026thinsp;=\u0026thinsp;0.07), cumulative GPA (t(95)\u0026thinsp;=\u0026thinsp;1.75, p = .084, d\u0026thinsp;=\u0026thinsp;0.35), prior piano experience in years (t(95)\u0026thinsp;=\u0026thinsp;\u0026minus;\u0026thinsp;0.35, p = .729, d\u0026thinsp;=\u0026thinsp;0.07), composite engagement (t(95)\u0026thinsp;=\u0026thinsp;0.61, p = .541, d\u0026thinsp;=\u0026thinsp;0.12), SDT composite (t(95)\u0026thinsp;=\u0026thinsp;1.34, p = .183, d\u0026thinsp;=\u0026thinsp;0.27), persistence intention (t(95)\u0026thinsp;=\u0026thinsp;0.86, p = .394, d\u0026thinsp;=\u0026thinsp;0.17), and course satisfaction (t(95)\u0026thinsp;=\u0026thinsp;1.06, p = .290, d\u0026thinsp;=\u0026thinsp;0.22). A chi-square test confirmed no significant difference in gender distribution between groups (χ\u0026sup2;(1)\u0026thinsp;=\u0026thinsp;0.65, p = .421). These results provide strong evidence that any observed post-test differences can be attributed to the intervention rather than to pre-existing group differences.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Descriptive Statistics\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e presents the pre- and post-test means and standard deviations for both groups across all primary outcome variables. Visual inspection of the descriptive statistics reveals a consistent pattern: both groups had comparable scores at baseline, but the experimental group showed substantially larger gains from pre-test to post-test across all engagement subscales, SDT need satisfaction measures, persistence intention, and course satisfaction. The control group showed modest improvements or relative stability across the same period, suggesting that typical instructional exposure produced limited motivational gains in the absence of AI-assisted supplementation.\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\u003ePre- and Post-Test Descriptive Statistics by Group\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExperimental (n\u0026thinsp;=\u0026thinsp;49)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eControl (n\u0026thinsp;=\u0026thinsp;48)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePre M\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePre SD\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePost M\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003ePost SD\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003ePre M\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003ePre SD\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003ePost M\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003ePost SD\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEngagement (comp.)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.37\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e3.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.39\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eIntrinsic Motiv.\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.61\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.72\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e3.01\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.68\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eBehavioral Eng.\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.63\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.83\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e2.93\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.62\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eEmotional Eng.\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.66\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e3.02\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.71\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e3.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.70\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eCognitive Eng.\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2.80\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.64\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.28\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.63\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.79\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e2.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.63\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSDT Composite\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.03\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.37\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.87\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.38\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e3.02\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.38\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eAutonomy\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.04\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.52\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.89\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e2.98\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.63\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eCompetence\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.58\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.47\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.74\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.67\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e2.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.66\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cem\u003eRelatedness\u003c/em\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.24\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.52\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e3.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e3.09\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.57\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePersistence\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.63\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.73\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.68\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e3.08\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.75\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e3.08\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.80\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSatisfaction\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2.99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.50\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e2.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e0.60\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e \u003cp\u003e2.94\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c9\"\u003e \u003cp\u003e0.65\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"9\"\u003e\u003cem\u003eNote.\u003c/em\u003e Comp. = composite. All scores on a 1\u0026ndash;5 Likert scale. Pre\u0026thinsp;=\u0026thinsp;pre-test (Week 0); Post\u0026thinsp;=\u0026thinsp;post-test (Week 16).\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Post-Test Group Comparisons and Effect Sizes\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e summarizes the results of post-test comparisons between the experimental and control groups, including t-statistics, significance values, effect sizes (Cohen\u0026rsquo;s d), and 95% confidence intervals for the effect sizes. All primary outcome variables showed statistically significant differences favoring the experimental group. With the exception of cognitive engagement (p = .014), all results remained significant after applying Bonferroni correction (adjusted α\u0026thinsp;=\u0026thinsp;.007).\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\u003ePost-Test Group Comparisons with Effect Sizes\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExp M(SD)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCtrl M(SD)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003et(95)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003ep\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003ed\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003e95% CI\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEngagement\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.45(.42)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.00(.39)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5.46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e[0.68,1.53]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIntrinsic Mot.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.54(.61)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.01(.68)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e[0.41,1.23]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBehavioral Eng.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.35(.60)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2.93(.62)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.39\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.69\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e[0.28,1.09]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEmotional Eng.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.66(.60)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.11(.70)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4.00\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.85\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e[0.43,1.26]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCognitive Eng.\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.28(.63)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2.96(.63)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.51\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.014\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.51\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e[0.10,0.91]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSDT Composite\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.62(.37)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.02(.38)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e7.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.59\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e[1.13,2.04]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAutonomy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.54(.52)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2.98(.63)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4.77\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.97\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e[0.55,1.38]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCompetence\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.47(.60)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2.85(.66)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4.71\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e[0.54,1.37]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRelatedness\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.55(.56)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.09(.57)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.72\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e[0.31,1.13]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePersistence\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.73(.68)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e3.08(.80)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e4.32\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e[0.46,1.29]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSatisfaction\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.65(.62)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2.94(.65)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5.55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e \u003cp\u003e1.13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e \u003cp\u003e[0.70,1.55]\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"7\"\u003e\u003cem\u003eNote.\u003c/em\u003e Exp\u0026thinsp;=\u0026thinsp;Experimental; Ctrl\u0026thinsp;=\u0026thinsp;Control. CI\u0026thinsp;=\u0026thinsp;confidence interval for Cohen\u0026rsquo;s d. All Bonferroni-corrected results (adjusted α\u0026thinsp;=\u0026thinsp;.007) remained significant except cognitive engagement (p = .014).\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e4.4 Hypothesis Testing\u003c/h2\u003e \u003cp\u003eHypotheses H1a through H1c predicted that AI-assisted instruction would enhance the satisfaction of basic psychological needs as conceptualized by SDT. All three hypotheses were fully supported by the data. Students in the experimental group reported significantly higher post-test autonomy (t(95)\u0026thinsp;=\u0026thinsp;4.77, p \u0026lt; .001, d\u0026thinsp;=\u0026thinsp;0.97, 95% CI [0.55, 1.38]) than their counterparts in the control group, indicating that the AI platform\u0026rsquo;s self-directed goal-setting and personalized recommendation features meaningfully enhanced students\u0026rsquo; sense of volitional engagement with practice. Competence showed similarly large gains (t(95)\u0026thinsp;=\u0026thinsp;4.71, p \u0026lt; .001, d\u0026thinsp;=\u0026thinsp;0.96, 95% CI [0.54, 1.37]), consistent with the theoretical prediction that real-time performance feedback strengthens mastery experiences and reinforces self-efficacy. Relatedness exhibited a medium-to-large effect (t(95)\u0026thinsp;=\u0026thinsp;3.56, p \u0026lt; .001, d\u0026thinsp;=\u0026thinsp;0.72, 95% CI [0.31, 1.13]), suggesting that the social dashboard feature contributed to students\u0026rsquo; sense of connectedness, though to a lesser extent than the competence and autonomy enhancements.\u003c/p\u003e \u003cp\u003eHypotheses H2a through H2c predicted that AI-assisted instruction would produce significantly greater improvements in engagement, persistence, and satisfaction. All three hypotheses were strongly supported. The experimental group demonstrated significantly higher post-test composite engagement (t(95)\u0026thinsp;=\u0026thinsp;5.46, p \u0026lt; .001, d\u0026thinsp;=\u0026thinsp;1.11, 95% CI [0.68, 1.53]), with large effects observed for emotional engagement (d\u0026thinsp;=\u0026thinsp;0.85) and intrinsic motivation (d\u0026thinsp;=\u0026thinsp;0.82), and medium effects for behavioral engagement (d\u0026thinsp;=\u0026thinsp;0.69) and cognitive engagement (d\u0026thinsp;=\u0026thinsp;0.51). Persistence intention showed a large effect (t(95)\u0026thinsp;=\u0026thinsp;4.32, p \u0026lt; .001, d\u0026thinsp;=\u0026thinsp;0.88, 95% CI [0.46, 1.29]), indicating that AI-assisted students expressed substantially stronger intentions to complete their music degree programs. Course satisfaction exhibited the largest effect among the primary outcomes (t(95)\u0026thinsp;=\u0026thinsp;5.55, p \u0026lt; .001, d\u0026thinsp;=\u0026thinsp;1.13, 95% CI [0.70, 1.55]), suggesting that the AI-enhanced practice experience fundamentally improved students\u0026rsquo; evaluative assessments of their educational experience.\u003c/p\u003e \u003cp\u003eHypothesis H3 predicted that dropout risk would be significantly lower in the AI-assisted group. This hypothesis was partially supported in terms of direction but did not reach conventional statistical significance. The proportion of students classified as at risk was 6.1% in the experimental group (3 out of 49 students) compared with 18.8% in the control group (9 out of 48 students)\u0026mdash;a difference of 12.7 percentage points representing a threefold reduction in relative risk. However, the chi-square test did not reach the conventional significance threshold (χ\u0026sup2;(1)\u0026thinsp;=\u0026thinsp;2.50, p = .114, Cramer\u0026rsquo;s V = .16). A post hoc power analysis indicated that detecting a significant difference in proportions of this magnitude (6% vs. 19%) at α\u0026thinsp;=\u0026thinsp;.05 with 80% power would require approximately 140 participants per group, suggesting that the non-significant result likely reflects insufficient statistical power for detecting differences in low-frequency binary outcomes rather than a true null effect.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e4.5 Supplementary Analyses: Practice Behavior and AI Engagement\u003c/h2\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents a multi-panel visualization of the comparative outcomes. Panel A displays horizontal bar charts of mean gain scores (post-test minus pre-test) for each outcome variable, making the differential improvement between groups immediately visible. Satisfaction showed the largest absolute gain in the AI-assisted group (+\u0026thinsp;0.66 points) compared with negligible change in the control group (+\u0026thinsp;0.06), while emotional engagement demonstrated the second-largest differential (+\u0026thinsp;0.56 vs. +0.09). Notably, persistence in the control group registered zero net change over the semester, underscoring the stagnation of motivation under conventional instruction. Panel B uses a radar chart to compare post-test SDT need satisfaction profiles, illustrating that the AI-assisted group achieved meaningfully higher scores across all three needs, with the most pronounced divergence on competence (3.47 vs. 2.85) and autonomy (3.54 vs. 2.98). Panel C tracks the pre-to-post trajectories for three key outcomes, revealing that the experimental group\u0026rsquo;s scores diverged sharply upward while the control group\u0026rsquo;s lines remained nearly flat. This divergence is especially striking for satisfaction, where the experimental trajectory rose from 2.99 to 3.65 while the control trajectory barely moved from 2.88 to 2.94. Panel D presents the dropout risk distribution using stacked bars, showing that only 3 students (6.1%) in the AI-assisted group were classified as at risk compared with 9 students (18.8%) in the control group, though this difference did not reach statistical significance (χ\u0026sup2;(1)\u0026thinsp;=\u0026thinsp;2.50, p = .114).\u003c/p\u003e \u003cp\u003eAnalysis of weekly practice logs revealed that students in the experimental group practiced a mean of 6.82 hours per week (SD\u0026thinsp;=\u0026thinsp;1.94) across the 16-week observation period, compared with 5.23 hours per week (SD\u0026thinsp;=\u0026thinsp;2.01) in the control group. This difference was statistically significant (t(95)\u0026thinsp;=\u0026thinsp;3.91, p \u0026lt; .001, d\u0026thinsp;=\u0026thinsp;0.80), representing a large effect that indicates the AI-assisted platform was associated with meaningfully increased practice engagement over time. Importantly, longitudinal examination of weekly practice hours revealed diverging trajectories: experimental-group students showed a gradual increase in practice hours from an average of 5.9 hours in Week 1 to 7.8 hours in Week 16, while control-group students showed relative stability with a slight declining trend from 5.5 hours in Week 1 to 5.0 hours in Week 16.\u003c/p\u003e \u003cp\u003eStudents in the experimental group generated an average of 9.4 AI feedback interactions per practice session, with interaction frequency increasing from a mean of 8.2 interactions in Week 1 to 12.6 interactions in Week 16. This upward trend suggests that students became increasingly engaged with the AI feedback system over time rather than experiencing novelty effects that dissipated after initial exposure. The correlation between cumulative AI interaction count and post-test engagement composite was positive and moderate (r = .38, p = .007), providing preliminary evidence that more intensive use of the AI platform was associated with greater motivational gains.\u003c/p\u003e \u003cp\u003eInternal consistency reliability was assessed for all scales at both time points and was found to be satisfactory to excellent across all measures. Pre-test Cronbach\u0026rsquo;s alpha values ranged from .81 (autonomy) to .92 (emotional engagement), and post-test values ranged from .83 (relatedness) to .93 (emotional engagement). These values exceed the conventional threshold of .70 recommended for research purposes and support the psychometric adequacy of the measurement instruments in this sample.\u003c/p\u003e \u003c/div\u003e"},{"header":"5. Discussion","content":"\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e5.1 Summary and Interpretation of Primary Findings\u003c/h2\u003e \u003cp\u003eThe primary aim of this study was to investigate whether AI-assisted piano instruction could mitigate student disengagement and support persistence in higher music education. The findings provide robust and consistent evidence that integration of adaptive AI tools into piano practice produces meaningful and statistically significant improvements in engagement, satisfaction, persistence intention, and basic psychological need satisfaction relative to traditional instruction. The magnitude of these effects\u0026mdash;with Cohen\u0026rsquo;s d values ranging from 0.51 to 1.59 across the primary outcomes\u0026mdash;is noteworthy, as educational interventions rarely produce effects of this size (Hattie, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2009\u003c/span\u003e). These results extend prior work on technology-enhanced learning in music education (Huang \u0026amp; Chen, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Xie et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2022\u003c/span\u003e) by demonstrating that the benefits of AI-mediated feedback extend beyond the skill acquisition outcomes typically examined in music technology research to encompass motivational, affective, and retention-relevant variables that are directly implicated in student persistence.\u003c/p\u003e \u003cp\u003eThe large effect sizes observed for composite engagement (d\u0026thinsp;=\u0026thinsp;1.11) and course satisfaction (d\u0026thinsp;=\u0026thinsp;1.13) suggest that AI-assisted instruction substantially transformed the subjective quality of the practice experience for students in the experimental group. Rather than experiencing practice as a solitary, feedback-deficient activity punctuated by occasional instructor guidance, these students engaged with a responsive system that acknowledged their efforts, diagnosed their weaknesses, calibrated their challenges, and connected them to a community of fellow learners. This transformation of the practice experience appears to have had cascading effects on motivational processes, as reflected in the significant gains across all four engagement subscales and in the elevated persistence intentions reported by experimental-group students.\u003c/p\u003e \u003cp\u003eFrom the perspective of self-determination theory, the pattern of SDT need satisfaction gains is theoretically coherent and practically illuminating. Competence exhibited the largest gain among the SDT subscales (d\u0026thinsp;=\u0026thinsp;0.96), which aligns with the proposition that real-time performance feedback directly reinforces mastery experiences by providing students with clear, objective evidence of their progress and specific, actionable guidance for continued improvement (Bandura, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e1997\u003c/span\u003e; Evans, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). When students can observe quantifiable improvements in pitch accuracy, rhythm precision, or dynamic control across practice sessions, their sense of musical competence is strengthened in ways that may not be possible through self-assessment alone. Autonomy showed a comparably large effect (d\u0026thinsp;=\u0026thinsp;0.97), reflecting the platform\u0026rsquo;s capacity to empower students with choices about practice goals, exercise selection, and pacing\u0026mdash;features that contrast with the relatively prescribed nature of traditional practice assignments. The comparatively smaller but still significant effect on relatedness (d\u0026thinsp;=\u0026thinsp;0.72) suggests that while the social dashboard feature contributed to students\u0026rsquo; sense of connection, the digital interaction afforded by the platform may not fully replicate the relational richness of in-person ensemble performances, shared practice room conversations, and face-to-face studio interactions that characterize the traditional conservatory experience.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003e5.2 Theoretical Contributions\u003c/h2\u003e \u003cp\u003eThe integrated theoretical framework combining SDT, TAM, and Tinto\u0026rsquo;s student integration model proved productive for organizing and interpreting the multi-level mechanisms through which AI-assisted instruction influences retention-relevant outcomes. Each theoretical lens contributed a distinct and complementary perspective. SDT illuminated the motivational mechanisms by which technology-mediated feedback supports intrinsic engagement through the satisfaction of basic psychological needs. TAM helped explain why students adopted and sustained use of the AI platform\u0026mdash;the prerequisite condition for motivational benefits to accrue\u0026mdash;with high compliance rates (78.3%) and increasing interaction frequencies suggesting that students perceived the technology as both useful and usable. Tinto\u0026rsquo;s integration framework contextualized these individual-level gains within the broader institutional environment, framing the AI-assisted practice experience as a form of enhanced academic integration that strengthened students\u0026rsquo; commitment to their degree programs.\u003c/p\u003e \u003cp\u003eThis tripartite framework addresses a noted limitation of prior research, which has tended to apply single theoretical perspectives to explain technology-enhanced learning outcomes (Kahu \u0026amp; Nelson, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). By bridging motivational psychology (SDT), technology adoption research (TAM), and institutional retention theory (Tinto), the present study offers a more holistic account of the mechanisms linking AI-assisted instruction to student persistence. Future research may productively extend this integrated framework by formally testing the mediating pathways hypothesized in the conceptual model\u0026mdash;for example, examining whether SDT need satisfaction mediates the relationship between AI platform use and persistence intention, or whether technology acceptance moderates the strength of the SDT pathway. The cross-disciplinary nature of this framework also suggests its applicability beyond music education to other performance-intensive disciplines such as dance, theater, visual arts, and athletics, where practice-based learning, individual skill development, and motivational sustainability are similarly central concerns.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec23\" class=\"Section2\"\u003e \u003ch2\u003e5.3 Practical Implications for Institutions and Educators\u003c/h2\u003e \u003cp\u003eThe findings carry several actionable implications for institutional stakeholders, music department administrators, and piano instructors. First, the results suggest that music departments seeking to reduce attrition may benefit from integrating AI-assisted practice tools as a structured complement to traditional one-to-one instruction. The key phrase here is structured complement: the AI platform in this study did not replace instructor contact but rather enhanced the many hours of independent practice during which students traditionally receive no feedback. Institutions considering such adoption should ensure that AI tools are integrated into the curricular framework with clear usage guidelines, instructor oversight, and explicit connections to learning objectives\u0026mdash;rather than being offered as optional add-ons that students must discover and adopt on their own.\u003c/p\u003e \u003cp\u003eSecond, the importance of the social dashboard feature in supporting relatedness satisfaction highlights the need for AI platforms in music education to include collaborative and community-building functionalities. Technology developers should recognize that music learning, despite its often-solitary practice modality, is fundamentally a social activity embedded in communities of practice (Wenger, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e1998\u003c/span\u003e). Features that enable peer interaction, shared progress tracking, collaborative challenges, and mutual encouragement can help counteract the isolation that is a known driver of music student disengagement (Gaunt, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2010\u003c/span\u003e; Hatfield et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Institutions might further amplify these benefits by organizing AI-platform-based practice communities, peer mentoring programs, or weekly sharing sessions where students discuss their AI-mediated practice experiences.\u003c/p\u003e \u003cp\u003eThird, the non-significant dropout result, while directionally consistent with the hypothesis and reflecting a threefold reduction in relative risk, underscores that technology alone may be insufficient to eliminate dropout risk. This finding is consistent with Tinto\u0026rsquo;s (\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2017\u003c/span\u003e) emphasis that persistence requires not only individual engagement but also institutional commitment to creating supportive environments. Institutions that pair AI-assisted practice tools with comprehensive student support systems\u0026mdash;including academic advising, mental health services, financial aid, performance anxiety workshops, and flexible assessment policies\u0026mdash;are likely to achieve more substantial reductions in attrition than technology adoption alone can deliver. The AI platform may be most effective as one component within a multi-faceted retention strategy rather than as a standalone solution.\u003c/p\u003e \u003cp\u003eFourth, the finding that practice hours increased substantially in the experimental group has implications for practice pedagogy. Rather than simply telling students to practice more, which research has shown to be an ineffective motivational strategy (Jorgensen \u0026amp; Hallam, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), providing a feedback-rich practice environment appears to naturally increase practice engagement by making the activity more rewarding, more structured, and more clearly productive. This suggests a shift from quantity-focused practice norms toward quality-focused practice environments supported by intelligent feedback systems.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003e5.4 Limitations\u003c/h2\u003e \u003cp\u003eSeveral limitations of the present study warrant careful consideration and should inform the interpretation of findings as well as the design of future research. First, the quasi-experimental design with intact, pre-existing course sections limits the strength of causal inference. Although baseline equivalence was confirmed across all measured variables, the possibility of unmeasured confounding variables\u0026mdash;such as differences in motivation to enroll in one section versus another, instructor preferences, or scheduling-related factors\u0026mdash;cannot be entirely excluded. Future studies should employ randomized controlled designs where feasible, or utilize propensity score matching and instrumental variable approaches to strengthen causal claims.\u003c/p\u003e \u003cp\u003eSecond, the study was conducted at a single comprehensive university in central China, which limits the generalizability of findings to other cultural, institutional, and disciplinary contexts. Chinese music education traditions, student expectations, technology familiarity, and institutional structures may differ meaningfully from those in Western conservatories, liberal arts colleges, or other global contexts. Cross-cultural and multi-site replication studies are needed to determine whether the observed effects are robust across diverse educational settings.\u003c/p\u003e \u003cp\u003eThird, the 16-week observation period, while encompassing a full academic semester, may not capture longer-term retention outcomes. Students may sustain engagement gains in the short term but revert to baseline patterns when the intervention is withdrawn or when novelty effects dissipate. Longitudinal studies tracking actual degree completion rates over two to four years would provide more definitive evidence of the intervention\u0026rsquo;s impact on real-world retention. Fourth, the binary operationalization of dropout risk, while grounded in observable behavioral indicators, may lack the sensitivity needed to detect gradual changes in attrition risk. Future research could benefit from continuous measures of disengagement, latent class growth modeling, or survival analysis approaches that model the timing and trajectory of withdrawal consideration.\u003c/p\u003e \u003cp\u003eFifth, the sample size of 97, while adequate for detecting the large effects observed on continuous outcomes, provided limited statistical power for the dropout risk analysis involving low-frequency binary events. Larger multi-site studies with sample sizes sufficient for subgroup analyses (e.g., by year of study, gender, or prior experience level) are needed to identify potential moderators of the intervention effect. Sixth, the study did not include qualitative data collection methods such as interviews, focus groups, or reflective journals that could illuminate how individual students experienced the AI-mediated practice environment and could reveal nuances in the mechanisms linking technology use to motivational outcomes that quantitative measures alone cannot capture.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec25\" class=\"Section2\"\u003e \u003ch2\u003e5.5 Future Directions\u003c/h2\u003e \u003cp\u003eSeveral promising directions emerge for future research. First, mixed-methods designs that combine quantitative outcome measurement with qualitative exploration of student experiences would provide richer and more nuanced understandings of how AI-assisted instruction shapes motivational trajectories. Semi-structured interviews conducted at multiple time points during the intervention could illuminate critical incidents, turning points, and individual differences in how students respond to AI-mediated feedback. Second, research designs that formally test the mediating pathways specified in the conceptual model\u0026mdash;using structural equation modeling or multilevel mediation analysis\u0026mdash;would advance the field from demonstrating that AI-assisted instruction works to explaining how and why it works. Third, comparative studies examining different AI platform designs, feature sets, and implementation models would help identify the specific design elements that are most responsible for motivational gains, enabling evidence-based design recommendations for technology developers. Fourth, investigations of potential differential effects across student subgroups\u0026mdash;such as novice versus advanced students, students with high versus low initial motivation, or students from different socioeconomic backgrounds\u0026mdash;would support more targeted and equitable implementation strategies.\u003c/p\u003e \u003c/div\u003e"},{"header":"6. Conclusion","content":"\u003cp\u003eThis study provides compelling evidence that AI-assisted piano instruction can meaningfully enhance engagement, satisfaction, and persistence intention among undergraduate music students while supporting the fulfillment of basic psychological needs for autonomy, competence, and relatedness. The large effect sizes observed across multiple outcome variables, the theoretically coherent pattern of SDT need satisfaction gains, and the significant increase in practice hours collectively paint a consistent picture of a technology-enhanced learning environment that transforms the quality of independent practice and strengthens students\u0026rsquo; motivational connection to their musical studies. By integrating self-determination theory, the technology acceptance model, and Tinto\u0026rsquo;s student integration framework into a unified analytical lens, the investigation offers a multidimensional account of the mechanisms through which adaptive technology shapes the motivational and relational conditions that sustain student commitment in performance-intensive educational settings.\u003c/p\u003e \u003cp\u003eThe practical implications of these findings extend to music departments seeking evidence-based strategies for reducing student attrition, to technology developers designing AI-powered practice tools, and to higher education policymakers concerned with quality and equity in educational outcomes. While the non-significant dropout result cautions against viewing AI-assisted instruction as a sufficient solution to retention challenges, the magnitude and consistency of the motivational and engagement gains suggest that such tools represent a valuable component of comprehensive institutional retention strategies. Future research should pursue longitudinal, multi-site, and mixed-methods designs to extend these findings, test the proposed mediating mechanisms, and identify the conditions under which AI-assisted instruction is most effective for diverse student populations. As higher education systems worldwide grapple with the imperative to provide quality, inclusive, and sustainable education aligned with SDG 4, the intelligent integration of adaptive technology into instrumental music pedagogy offers a promising pathway for supporting student success and mitigating the persistent challenge of disengagement and dropout.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research is funded by Han Dan University, under grant number Faculty of Music 2026-005.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of Interest Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe author declares that there are no conflicts of interest or competing interests regarding the publication of this article. The author has no relevant financial or non-financial interests to disclose. There are no personal, professional, or financial relationships that could potentially be construed as influencing the content or conclusions presented in this work.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe studies involving human participants were reviewed and approved by the Institutional Review Board (or Ethics Committee) of Han Dan University (IRB Protocol No.: Research Office of Handan University 202511120008).\u003c/p\u003e\n\u003cp\u003eThe study was conducted in accordance with the local legislation and institutional requirements, as well as the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments. Prior to data collection, written informed consent to participate in this study was provided by all participants. Participants were fully informed regarding the study\u0026apos;s objectives, procedures, and data handling processes. They were explicitly assured that their participation was entirely voluntary and that they could withdraw from the study at any time without any adverse consequences to their course grades, academic standing, or relationship with the instructors. To ensure confidentiality, all collected data, including pre- and post-test scores and AI platform usage logs, were strictly anonymized and aggregated prior to analysis.\u003c/p\u003e\u003cp\u003e\u003cstrong\u003eAppendix: Data Availability Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe raw data supporting the conclusions of this article are available from the corresponding author upon reasonable request. Anonymized datasets, including participant-level pre- and post-test scores across all outcome measures, weekly practice logs with practice hours and AI interaction counts for each of the 16 weeks, and individual-level item responses for all survey scales, have been prepared in accordance with institutional data-sharing policies. Three data files are available: (a) Main_Data, containing demographic information and composite scale scores for all 97 participants; (b) Weekly_Practice_Logs, containing 1,552 weekly observations (97 participants \u0026times; 16 weeks) of practice hours and AI feedback interaction counts; and (c) Scale_Items, containing 194 rows (97 participants \u0026times; 2 time points) of individual item-level responses for all nine measurement constructs. All data were collected under IRB-approved protocols with written informed consent from all participants, and all identifying information has been removed to protect participant privacy.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eAl-Emran, M., Mezhuyev, V., \u0026amp; Kamaludin, A. (2018). Technology acceptance model in M-learning context: A systematic review. Computers \u0026amp; Education, 125, 389\u0026ndash;412. https://doi.org/10.1016/j.compedu.2018.06.008\u003c/li\u003e\n \u003cli\u003eBandura, A. (1997). Self-efficacy: The exercise of control. W. H. Freeman.\u003c/li\u003e\n \u003cli\u003eBonneville-Roussy, A., Evans, P., Verner-Filion, J., Vallerand, R. J., \u0026amp; Bouffard, T. (2017). Motivation and coping with the stress of assessment: Gender differences in outcomes for university students. Contemporary Educational Psychology, 48, 28\u0026ndash;42. https://doi.org/10.1016/j.cedpsych.2016.08.003\u003c/li\u003e\n \u003cli\u003eBrislin, R. W. (1970). Back-translation for cross-cultural research. Journal of Cross-Cultural Psychology, 1(3), 185\u0026ndash;216. https://doi.org/10.1177/135910457000100301\u003c/li\u003e\n \u003cli\u003eBurt-Perkins, R., \u0026amp; Mills, J. (2019). The role of the first-year in undergraduate music students\u0026rsquo; decisions to persist. British Journal of Music Education, 36(1), 33\u0026ndash;45. https://doi.org/10.1017/S0265051718000256\u003c/li\u003e\n \u003cli\u003eChen, B., Vansteenkiste, M., Beyers, W., Boone, L., Deci, E. L., Van der Kaap-Deeder, J., ... \u0026amp; Verstuyf, J. (2015). Basic psychological need satisfaction, need frustration, and need strength across four cultures. Motivation and Emotion, 39(2), 216\u0026ndash;236. https://doi.org/10.1007/s11031-014-9450-1\u003c/li\u003e\n \u003cli\u003eChen, X., Xie, H., Zou, D., \u0026amp; Hwang, G. J. (2022). Application and theory gaps during the rise of artificial intelligence in education. Computers and Education: Artificial Intelligence, 1, 100002. https://doi.org/10.1016/j.caeai.2020.100002\u003c/li\u003e\n \u003cli\u003eCohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Lawrence Erlbaum Associates.\u003c/li\u003e\n \u003cli\u003eCrawford, R. (2021). Lessons from the pandemic: What music education can learn from the COVID-19 crisis. International Journal of Music Education, 39(3), 277\u0026ndash;293. https://doi.org/10.1177/02557614211024163\u003c/li\u003e\n \u003cli\u003eCreech, A., Papageorgi, I., Duffy, C., Morton, F., Haddon, E., Potter, J., ... \u0026amp; Welch, G. (2008). From music student to professional: The process of transition. British Journal of Music Education, 25(3), 315\u0026ndash;331. https://doi.org/10.1017/S0265051708008127\u003c/li\u003e\n \u003cli\u003eDavidson, W. B., Beck, H. P., \u0026amp; Milligan, M. (2009). The college persistence questionnaire: Development and validation of an instrument that predicts student attrition. Journal of College Student Development, 50(4), 373\u0026ndash;390. https://doi.org/10.1353/csd.0.0079\u003c/li\u003e\n \u003cli\u003eDavis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3), 319\u0026ndash;340. https://doi.org/10.2307/249008\u003c/li\u003e\n \u003cli\u003eDye, K. (2022). Online piano pedagogy: Perspectives and practices during COVID-19. Music Education Research, 24(2), 176\u0026ndash;189. https://doi.org/10.1080/14613808.2022.2032735\u003c/li\u003e\n \u003cli\u003eEvans, P. (2015). Self-determination theory: An approach to motivation in music education. Musicae Scientiae, 19(1), 65\u0026ndash;83. https://doi.org/10.1177/1029864914568044\u003c/li\u003e\n \u003cli\u003eGaunt, H. (2010). One-to-one tuition in a conservatoire: The perceptions of instrumental and vocal students. Psychology of Music, 38(2), 178\u0026ndash;208. https://doi.org/10.1177/0305735609339467\u003c/li\u003e\n \u003cli\u003eGinsborg, J., Kreutz, G., Thomas, M., \u0026amp; Williamon, A. (2009). Healthy behaviours in music and non-music performance students. Health Education, 109(3), 242\u0026ndash;258. https://doi.org/10.1108/09654280910955575\u003c/li\u003e\n \u003cli\u003eHandelsman, M. M., Briggs, W. L., Sullivan, N., \u0026amp; Towler, A. (2005). A measure of college student course engagement. The Journal of Educational Research, 98(3), 184\u0026ndash;192. https://doi.org/10.3200/JOER.98.3.184-192\u003c/li\u003e\n \u003cli\u003eHatfield, J. L., Halvari, H., \u0026amp; Deci, E. L. (2022). Motivation, anxiety, and well-being among music academy students: The role of basic psychological need satisfaction. Frontiers in Psychology, 13, 891758. https://doi.org/10.3389/fpsyg.2022.891758\u003c/li\u003e\n \u003cli\u003eHattie, J. (2009). Visible learning: A synthesis of over 800 meta-analyses relating to achievement. Routledge.\u003c/li\u003e\n \u003cli\u003eHuang, A. Y., \u0026amp; Chen, N. S. (2023). AI in music education: A systematic review of intelligent tutoring systems and adaptive learning. Education and Information Technologies, 28(5), 5973\u0026ndash;6002. https://doi.org/10.1007/s10639-022-11406-7\u003c/li\u003e\n \u003cli\u003eHwang, G. J., Xie, H., Wah, B. W., \u0026amp; Ga\u0026scaron;ević, D. (2020). Vision, challenges, roles and research issues of artificial intelligence in education. Computers and Education: Artificial Intelligence, 1, 100001. https://doi.org/10.1016/j.caeai.2020.100001\u003c/li\u003e\n \u003cli\u003eJorgensen, H., \u0026amp; Hallam, S. (2016). Practicing. In S. Hallam, I. Cross, \u0026amp; M. Thaut (Eds.), The Oxford handbook of music psychology (2nd ed., pp. 449\u0026ndash;462). Oxford University Press.\u003c/li\u003e\n \u003cli\u003eKahu, E. R., \u0026amp; Nelson, K. (2018). Student engagement in the educational interface: Understanding the mechanisms of student success. Higher Education Research \u0026amp; Development, 37(1), 58\u0026ndash;71. https://doi.org/10.1080/07294360.2017.1344197\u003c/li\u003e\n \u003cli\u003eKenny, D. T. (2011). The psychology of music performance anxiety. Oxford University Press.\u003c/li\u003e\n \u003cli\u003eMiksza, P., Evans, P., \u0026amp; McPherson, G. E. (2022). Motivation to pursue music: A self-determination theory perspective. In G. E. McPherson (Ed.), The Oxford handbook of music performance (Vol. 2, pp. 567\u0026ndash;591). Oxford University Press.\u003c/li\u003e\n \u003cli\u003eNiemiec, C. P., \u0026amp; Ryan, R. M. (2009). Autonomy, competence, and relatedness in the classroom: Applying self-determination theory to educational practice. Theory and Research in Education, 7(2), 133\u0026ndash;144. https://doi.org/10.1177/1477878509104318\u003c/li\u003e\n \u003cli\u003eRyan, R. M., \u0026amp; Deci, E. L. (2020). Intrinsic and extrinsic motivation from a self-determination theory perspective: Definitions, theory, practices, and future directions. Contemporary Educational Psychology, 61, 101860. https://doi.org/10.1016/j.cedpsych.2020.101860\u003c/li\u003e\n \u003cli\u003eScherer, R., Siddiq, F., \u0026amp; Tondeur, J. (2019). The technology acceptance model (TAM): A meta-analytic structural equation modeling approach to explaining teachers\u0026rsquo; adoption of digital technology in education. Computers \u0026amp; Education, 128, 13\u0026ndash;35. https://doi.org/10.1016/j.compedu.2018.09.009\u003c/li\u003e\n \u003cli\u003eSchreiner, L. A., \u0026amp; Juillerat, S. L. (1994). Student satisfaction inventory. Noel-Levitz.\u003c/li\u003e\n \u003cli\u003eTinto, V. (1993). Leaving college: Rethinking the causes and cures of student attrition (2nd ed.). University of Chicago Press.\u003c/li\u003e\n \u003cli\u003eTinto, V. (2017). Through the eyes of students. Journal of College Student Retention: Research, Theory \u0026amp; Practice, 19(3), 254\u0026ndash;269. https://doi.org/10.1177/1521025115621917\u003c/li\u003e\n \u003cli\u003eVenkatesh, V., \u0026amp; Davis, F. D. (2000). A theoretical extension of the technology acceptance model: Four longitudinal field studies. Management Science, 46(2), 186\u0026ndash;204. https://doi.org/10.1287/mnsc.46.2.186.11926\u003c/li\u003e\n \u003cli\u003eWenger, E. (1998). Communities of practice: Learning, meaning, and identity. Cambridge University Press.\u003c/li\u003e\n \u003cli\u003eXie, H., Chu, H. C., Hwang, G. J., \u0026amp; Wang, C. C. (2022). Trends and development in technology-enhanced adaptive/personalized learning: A systematic review of journal publications from 2007 to 2021. Computers \u0026amp; Education, 140, 103599. https://doi.org/10.1016/j.compedu.2019.103599\u003c/li\u003e\n \u003cli\u003eZhai, X., Chu, X., Chai, C. S., Jong, M. S., Istenic, A., Spector, M., ... \u0026amp; Li, Y. (2021). A review of artificial intelligence (AI) in education from 2010 to 2020. Complexity, 2021, 8812542. https://doi.org/10.1155/2021/8812542\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Han Dan Univeristy","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":"AI-assisted instruction, piano pedagogy, student persistence, engagement, self-determination theory, higher music education, dropout, technology-enhanced learning, adaptive feedback, SDG 4","lastPublishedDoi":"10.21203/rs.3.rs-9485161/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9485161/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study investigated whether AI-assisted piano instruction enhances engagement, persistence, and satisfaction among 97 undergraduate music students at a Chinese university. Grounded in self-determination theory, the technology acceptance model, and Tinto's integration framework, a 16-week quasi-experimental design compared an AI-assisted group (n\u0026thinsp;=\u0026thinsp;49) with a traditional instruction group (n\u0026thinsp;=\u0026thinsp;48). Results revealed significantly higher post-test scores in the AI-assisted group for engagement (d\u0026thinsp;=\u0026thinsp;1.11), persistence intention (d\u0026thinsp;=\u0026thinsp;0.88), and satisfaction (d\u0026thinsp;=\u0026thinsp;1.13), all at p \u0026lt; .001. Autonomy (d\u0026thinsp;=\u0026thinsp;0.97), competence (d\u0026thinsp;=\u0026thinsp;0.96), and relatedness (d\u0026thinsp;=\u0026thinsp;0.72) also improved significantly. Dropout risk was directionally lower in the AI-assisted group (6.1% vs. 18.8%) but non-significant (p = .114). Weekly practice hours were significantly higher in the experimental group (M\u0026thinsp;=\u0026thinsp;6.82 vs. 5.23, d\u0026thinsp;=\u0026thinsp;0.80). These findings suggest AI-assisted piano pedagogy can mitigate disengagement by fulfilling psychological needs, supporting retention strategies aligned with SDG 4.\u003c/p\u003e","manuscriptTitle":"Mitigating Disengagement in Higher Music Education:The Role of AI-Assisted Piano Instruction in Sustaining Student Persistence","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-22 08:47:47","doi":"10.21203/rs.3.rs-9485161/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":"d493c778-d615-4344-963a-03f6c130c4dc","owner":[],"postedDate":"April 22nd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":66742486,"name":"Educational Psychology"}],"tags":[],"updatedAt":"2026-04-22T08:47:47+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-22 08:47:47","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9485161","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9485161","identity":"rs-9485161","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}