Artificial intelligence in rehabilitation education: An assistive tool, not a substitute for scientific judgment

THE ASSIGNMENT AND APPROPRIATE AI USE

The professor assigned a comprehensive job-site ergonomic analysis, a foundational assessment task in musculoskeletal rehabilitation practice, to 70 undergraduate physical therapy students enrolled in the eighth level (PT8) of the Bachelor of Physical Therapy program. Students were provided with professional photographs taken directly at real occupational settings, depicting workers performing specific job tasks: A desk-based computer operator and a heating, ventilation, and air conditioning maintenance technician performing overhead tasks under physically demanding conditions.^[6,7]

The use of real job-site photographs was deliberate and pedagogically significant.^[3] It situated the assignment within authentic clinical reality rather than idealized textbook scenarios, requiring students to engage with the complexity, ambiguity, and contextual richness that characterize actual occupational biomechanics assessment. Students were explicitly encouraged to use AI as a legitimate assistive tool during the preliminary stages of the assignment.^[8-10] On this dimension, AI performed admirably. Students employed AI tools to generate frame-by-frame sequential illustrations of the occupational tasks depicted in the photographs, producing clear, systematic visual breakdowns of movement sequences that would have required hours of manual drawing to produce independently.^[9] AI also assisted students in identifying relevant ergonomics literature, organizing background information on occupational injury mechanisms, and structuring the preliminary sections of their reports. These applications genuinely improved efficiency and enabled students to devote more cognitive attention to the intellectually demanding components of the assignment.^[10]

WHERE CLINICAL REASONING SURPASSED AI

The difficulties emerged precisely where the assignment’ educational value was concentrated. When students attempted to use AI to superimpose biomechanical force analyses onto actual job-site photographs, including identifying external and internal forces acting on the musculoskeletal system, constructing force vectors, calculating moment arms and joint moments, and evaluating tissue loading at specific joint configurations, the limitations of AI became immediately and instructively apparent.^[3,6]

AI-generated force analyses, when applied to real-world photographic images of actual workers in actual occupational settings, consistently failed to account for the patient-specific anatomical variability, postural asymmetries, load distribution patterns, and environmental constraints visible in the photographs. The outputs appeared analytically credible on the surface but, on critical examination, were biomechanically inconsistent with the specific loading conditions depicted. Students who submitted AI-generated force analyses without independent verification demonstrated not the application of biomechanical knowledge but its circumvention.^[1-3,9]

The second domain of failure was equally revealing. Students experienced significant difficulty translating their biomechanical findings into clinically meaningful recommendations. Specifically, they struggled to develop evidence-based strategies to instruct workers on how to mitigate existing musculoskeletal symptoms, prevent the progression of ongoing complaints, modify task performance to reduce tissue loading, and support the retention of valued employees through targeted ergonomic interventions. This process of clinical reasoning, which connects biomechanical analysis with patient-centered intervention, represents a core competency in rehabilitation education.^[2,3,10] It requires the integration of biomechanical principles, clinical expertise, and evidence-based decision-making, all of which depend on human judgment and critical thinking rather than AI-generated outputs.^[1-3]

THE PEDAGOGICAL PRINCIPLE: A PERSPECTIVE FROM CLINICAL BIOMECHANICS PRACTICE

The prevention and health promotion assignment illustrates an important principle for rehabilitation education. AI is a valuable cognitive assistant for preparatory tasks such as literature exploration, information organization, and the generation of visual representations. However, it cannot replace activities that require the application of biomechanical principles, clinical reasoning, and professional judgment to authentic clinical problems.^[1-3]

The distinction between appropriate and inappropriate AI use reflects the difference between supporting learning and replacing it. Using AI to generate a frame-by-frame illustration of an occupational task facilitates learning.^[3,6] In contrast, relying on AI to produce a biomechanical force analysis or clinical recommendations without independent understanding undermines the development of scientific reasoning and professional competence. The educational issue, therefore, lies not in the technology itself, but in whether students critically engage with AI-generated outputs and transform them into evidence-based clinical knowledge.^[1-3]

Faculty members across health professions education bear a specific responsibility that this experience makes concrete: To design assessments that explicitly and structurally distinguish between AI-appropriate preliminary tasks and AI-inappropriate analytical and clinical tasks. Assignments that reward polished outputs without requiring transparent demonstration of reasoning will consistently fail to distinguish between students who have learned and students who have successfully delegated. The solution is not to prohibit AI but to design assessment tasks in which AI assistance alone is insufficient to demonstrate the clinical reasoning, scientific understanding, and professional competencies expected of rehabilitation students.^[1-4]

CONCLUSION

The students in the prevention and health promotion course gained a pedagogically invaluable experience by identifying the precise boundary between AI’s capabilities and the indispensable role of professional expertise. Through the intellectual challenge of a carefully designed assignment, they discovered that superimposing a biomechanically valid force analysis onto a photograph of a worker performing an authentic occupational task requires a depth of biomechanical knowledge, clinical reasoning, and contextual judgment that current AI systems cannot provide. Likewise, they learned that developing evidence-based recommendations to help symptomatic workers modify task performance, reduce tissue loading, manage ongoing musculoskeletal symptoms, and remain productive in the workplace depends on human expertise rather than automated responses.

These lessons extend far beyond a single classroom exercise. They cultivate the analytical thinking, scientific reasoning, and clinical judgment that define competent rehabilitation practice. Paradoxically, AI enhances learning not by providing answers but by exposing its limitations and encouraging deeper critical analysis. This insight captures the true educational value of AI. It should be embraced as an assistive tool that improves efficiency in literature searching, information synthesis, and academic organization while never replacing independent reasoning, evidence-based decision-making, or professional judgment. When integrated into thoughtfully designed curricula, AI serves not as a substitute for learning but as a catalyst for developing reflective, evidence-based rehabilitation professionals.

This educational experience was conducted at a single institution, which may limit the generalizability of the findings, and outcomes were based primarily on classroom experience and assignment performance rather than standardized objective measures. Future research should develop AI competency frameworks and objective assessment methods to support AI integration in rehabilitation education.