Comparing instructor designed rubric aligned chatbot feedback and instructor feedback in higher education regarding student perceptions of clarity usefulness supportiveness and satisfaction

Comparing instructor designed rubric aligned chatbot feedback and instructor feedback in higher education regarding student perceptions of clarity usefulness supportiveness and satisfaction | 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 Comparing instructor designed rubric aligned chatbot feedback and instructor feedback in higher education regarding student perceptions of clarity usefulness supportiveness and satisfaction Iris Gertner Moryossef, Christina Emanuelli This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8888154/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 13 You are reading this latest preprint version Abstract As generative artificial intelligence increasingly shapes pedagogical practice, automated feedback in higher education requires careful evaluation that extends beyond technical performance to students’ learning experiences. This study compared traditional instructor feedback with feedback generated by an instructor-designed, rubric-aligned chatbot (I-GenF-Bot). The chatbot was developed by the course instructor and configured to provide qualitative, criterion-referenced feedback aligned with a course rubric, without grades. One hundred and six undergraduate students in a third-year Digital Marketing course received iterative, formative feedback from I-GenF-Bot, followed by summative feedback from the instructor using the same rubric. Students completed a questionnaire assessing functional dimensions of feedback quality (clarity and usefulness), an emotional–personal dimension (supportiveness), and overall satisfaction with the chatbot. Results indicated no significant differences between instructor and chatbot feedback on clarity and usefulness, while instructor feedback was perceived as more supportive. Clarity, usefulness, and supportiveness were each positively associated with students’ overall satisfaction with I-GenF-Bot. An exploratory two-step cluster analysis suggested that student heterogeneity was driven primarily by perceptions of instructor feedback rather than chatbot feedback. Qualitative comments highlighted the chatbot’s strengths in immediacy, clarity, and accessibility, alongside limitations related to contextual sensitivity, repetition, and empathy. Overall, the findings suggest that while rubric alignment enables AI-based feedback to approximate instructor feedback on functional quality dimensions, an ‘empathy gap’ remains a key constraint. A hybrid feedback model that combines scalable AI-supported guidance with instructor judgment and relational support therefore appears particularly promising for scalable feedback practices in higher education. Generative AI in education instructor-designed chatbot feedback quality student satisfaction higher education hybrid learning Figures Figure 1 Figure 2 Figure 3 Introduction High-quality feedback is central to effective learning in higher education because it supports student progress, self-regulation, and the development of disciplinary competence. Yet, feedback is also one of the most difficult instructional practices to deliver consistently at scale. In many courses, students need iterative guidance on drafts, structure, reasoning, and alignment with assessment criteria, while instructors face substantial constraints on time and capacity. This tension between the pedagogical value of feedback and the practical difficulty of providing it has become more salient with the expansion of enrolments and the growing emphasis on authentic, performance-based assessment. At a system level, the challenge of providing high-quality feedback is increasingly intertwined with issues of scalability, instructor workload, and equity in mass higher education. As student numbers grow and assessment tasks become more complex and authentic, traditional feedback models struggle to deliver timely and individualized guidance to all learners. This tension positions feedback not only as a pedagogical issue, but also as a structural challenge for the future of higher education. Within this context, generative AI has been positioned as a potential mechanism for supporting feedback processes. Technology-enhanced learning research indicates that GPT-based systems can automate routine instructional tasks, identify weaknesses efficiently, and generate targeted prompts, while still requiring careful instructional design and human oversight to avoid pedagogical misalignment (Terrón, 2024 ). Students report practical benefits of ChatGPT for writing support and exam preparation, alongside concerns that point to the need for guidance and accountability in educational use (Ibeh et al., 2025 ). These observations suggest that the educational value of generative-AI feedback is shaped not only by the capabilities of the model but also by how the system is implemented, constrained, and aligned with course goals. A key design distinction is the difference between generic, open-ended use of large language models and structured, course-specific implementations. Generic AI feedback may be perceived as inconsistent, vague, or misaligned with local expectations, whereas rubric-aligned systems can be designed to reflect the learning objectives and criteria that define quality within a course (Sánchez-Ramírez et al., 2022 ). In higher education, this distinction matters because feedback often serves as the bridge between assessment standards and student revision. When AI is integrated as part of the feedback process, it becomes important to evaluate whether students experience it as comparable to instructor feedback on core quality dimensions and whether it supports a positive learning experience. This study contributes to research on AI-mediated feedback in higher education in three ways. First, it demonstrates that an instructor-designed, rubric-aligned chatbot can approximate instructor feedback on functional dimensions of feedback quality, particularly clarity and usefulness. Second, it identifies supportiveness as the primary dimension in which instructor feedback retains a clear advantage, highlighting an ‘empathy gap’ that structured AI feedback does not fully bridge. Third, by comparing both feedback sources within the same course context, the study provides empirical support for hybrid feedback models that combine scalable AI-based guidance with instructor judgment and relational support. Literature review Feedback quality as a multidimensional construct Feedback is commonly conceptualized as more than a single informational message, with prior work emphasizing both functional and relational components. Three dimensions that recur across higher-education feedback research are clarity, usefulness, and supportiveness. Clarity refers to feedback that is understandable, transparent, and reasoned, enabling students to recognize strengths and diagnose weaknesses in their work (Boud & Molloy, 2013 ; Brookhart, 2017 ). Usefulness reflects the extent to which feedback provides actionable direction for improvement, supports reflection, and facilitates forward progress in subsequent revisions or tasks (Dawson et al., 2019 ; Hattie & Timperley, 2007 ). Supportiveness captures the extent to which feedback communicates care, respect, encouragement, and sensitivity to the learner, including how well the feedback feels tailored to the student’s needs and level (Lipnevich & Smith, 2008 ). This multidimensional framing is particularly relevant to AI-mediated feedback because automated systems may perform strongly on functional dimensions while remaining limited on emotional–personal aspects of feedback interaction. Recent scholarship reinforces the importance of supportive and relational qualities in learners’ feedback experiences and highlights their potential role in distinguishing human-delivered feedback from AI-mediated feedback (Henderson et al., 2025 ; Sáiz-Manzanares et al., 2023 ). A multidimensional framework therefore provides a rigorous basis for evaluating chatbot feedback as an educational practice rather than as a purely technical output.Validating this separation is critical for the proposed hybrid model: if the chatbot can be proven to match the instructor on 'functional' dimensions (clarity, usefulness), instructors can then justify reallocating their limited time almost exclusively to the 'supportive' dimension that AI cannot replicate. Why AI-mediated feedback is not only a technical question The effectiveness of AI-mediated feedback depends on learners’ perceptions, including whether students view the system as credible and whether they experience the interaction as socially meaningful. Human–Computer Interaction perspectives emphasize trust in AI as a key construct shaping user uptake. Trust in AI refers to a user’s confidence in a system’s reliability and competence, and it influences whether outputs are treated as legitimate guidance or dismissed as unhelpful or risky (Lee & See, 2004 ). In educational settings, the same feedback content may be acted upon or ignored depending on whether the system is perceived as expert, consistent, and aligned with course expectations. Social presence further shapes the learner’s experience with conversational systems. Social presence refers to the degree to which a user perceives an interaction partner as a real social entity (Short et al., 1976 ). In learning contexts, higher social presence may support a stronger sense of dialogue and reduce transactional distance, thereby increasing engagement with feedback and motivation to revise (Moore, 1993 ). In the context of this study, we investigate whether the chatbot's immediate availability creates a sense of 'closeness' (reduced transactional distance) through rapid response, or if the lack of human identity maintains a high psychological distance despite the technical efficiency These constructs imply that evaluating AI feedback requires attention to more than clarity and usefulness. Students’ responsiveness to chatbot feedback may depend on whether the interaction feels pedagogically grounded, trustworthy, and supportive. Accordingly, this study examines whether strictly aligning the chatbot with the course rubric helps overcome this 'trust barrier,' allowing students to perceive the AI as a 'competent expert' comparable to the instructor, rather than as a generic, unreliable tool. Evidence on AI-based feedback and the case for structured and hybrid designs The growing body of research on AI-mediated feedback indicates both promise and limitations. Different feedback types can serve complementary functions, with elaborate feedback supporting task-specific improvement and motivational feedback bolstering engagement and persistence (Feng et al., 2025 ). This theoretical distinction serves as the blueprint for our dual-feedback design: we posit that the rubric-aligned chatbot can effectively assume the 'elaborate feedback' function regarding task specifics, thereby freeing the instructor to focus primarily on the 'motivational' function that requires human connection. Systematic reviews show increasing adoption of AI-based feedback, alongside barriers related to pedagogical design, implementation quality, and learner experience (Nazaretsky et al., 2024 ; Liang et al., 2025 ; Luo et al., 2025 ). While empirical work reports high levels of satisfaction with chatbot support (Sáiz-Manzanares et al., 2023 ; Subaveerapandiyan, 2024 ), comparative studies suggest that generic LLM feedback may lack pedagogical nuance compared to structured designs (Li et al. 2024 ; Seßler et al., 2025 ). This study leverages this distinction by controlling the feedback content via a shared rubric. By keeping the criteria constant for both the AI and the instructor, we can isolate the effect of the 'delivery agent' on student perceptions, independent of the feedback's content validity. Consequently, these findings strengthen the case for hybrid approaches that combine the immediacy of automated guidance with human oversight, judgement, and pedagogical responsiveness (Nazaretsky et al., 2024 ; Topping et al., 2025 ; Verleger & Pembridge, 2018 ). Instructor-designed, rubric-aligned chatbots as a focused implementation approach Instructor-designed chatbots represent a targeted approach to AI-mediated feedback because they can be built around course-specific criteria and integrated into learning tasks in a controlled manner. Rather than relying on general prompts, a rubric-aligned chatbot can be configured to provide structured, criterion-referenced feedback that reflects the learning outcomes and performance standards taught in the course. This approach is intended to reduce the risk of generic advice and to increase perceived alignment between feedback and assessment expectations. Evidence suggests that students value accessibility and alignment with learning objectives in chatbot systems, while also reporting limitations related to adaptability, precision, contextual sensitivity, and empathy (Yoo et al., 2025 ). These findings imply that instructor-designed systems may improve functional alignment, yet may still face constraints on the emotional–personal dimensions of feedback. This makes rubric-aligned bots a particularly appropriate context for evaluating whether AI can approximate instructor feedback on clarity and usefulness while identifying where the supportive dimension remains distinctively human. Satisfaction as a holistic indicator of perceived value Satisfaction as a holistic indicator of perceived value In studies of educational technology, satisfaction is often treated as a holistic indicator because it integrates functional evaluations with affective and interactional experience. In the context of chatbot feedback, satisfaction may reflect whether students perceive the system as convenient, meeting expectations, and supportive of their learning needs, even when specific feedback attributes vary. Prior studies report high satisfaction with chatbot support across contexts, while also documenting differences across learner groups and conditions (Sáiz-Manzanares et al., 2023 ; Subaveerapandiyan, 2024 ; Vanichvasin, 2022 ). Work examining student use of ChatGPT also points to tangible benefits alongside concerns that necessitate guidance and accountability (Ibeh et al., 2025 ). These findings justify treating satisfaction as a meaningful outcome that captures the overall experience of chatbot-mediated feedback beyond any single quality dimension. Gap and positioning of the present study Despite the rapid expansion of research on AI-mediated feedback, two gaps remain salient for higher education practice and theory. First, there is limited evidence that directly compares instructor feedback with instructor-designed, rubric-aligned chatbot feedback across both functional dimensions of feedback quality and emotional–personal dimensions that shape perceived care and support. Second, the relationship between perceived feedback quality and overall satisfaction with chatbot-mediated feedback remains under-specified, and learner heterogeneity in evaluating instructor versus chatbot feedback is rarely examined within the same course context. Addressing these gaps can clarify when a structured chatbot functions as a credible complement to instructor feedback and how different students experience AI-mediated feedback in ways that matter for adoption and learning design. The present study Building on this theoretical and empirical foundation, the present study compares students’ perceptions of two feedback sources within the same instructional setting. AI-based feedback was provided by I-GenF-Bot, a custom-built chatbot developed by the course instructor for a third-year undergraduate Digital Marketing course. Human feedback was provided by the course instructor. Both feedback processes were anchored in the same detailed rubric to ensure comparability of criteria, while differing in mode of delivery and interaction. The study was designed to examine students’ evaluations of feedback quality on clarity, usefulness, and supportiveness, and to determine how these perceived qualities relate to students’ overall satisfaction with chatbot-mediated feedback. In addition, the study explored whether distinct student profiles could be identified based on patterns of perceptions across instructor and chatbot feedback. By focusing on a rubric-aligned, instructor-designed system rather than generic AI use, the study aims to contribute to understanding how AI can be incorporated into higher education curricula in a manner consistent with pedagogical responsiveness (Huh et al., 2025 ). Research questions To examine these issues from an educational innovation and implementation perspective, the study addresses the following research questions: RQ1. Student-perceived differences between instructor feedback and I-GenF-Bot feedback across functional dimensions of feedback quality, clarity and usefulness, and across an emotional–personal dimension, supportiveness RQ2. Associations between perceived feedback quality dimensions and students’ overall satisfaction with the I-GenF-Bot RQ3. Emergence of distinct student profiles based on perceptions of instructor feedback versus I-GenF-Bot feedback, and characteristics differentiating those profiles Methods Research setting and participants This study was conducted in a third-year undergraduate Digital Marketing course in the School of Management at a higher education institution, during the 2025 academic year. The sample comprised 106 students, including 75 female students (70.8 percent) and 31 male students (29.2 percent). As part of the course requirements, students prepared and presented a professional marketing brief. To support learning and enhance the quality of students’ presentations, a dual-feedback approach was implemented in which students received summative feedback from the instructor and iterative formative feedback from I-GenF-Bot, a generative-AI feedback bot. It is important to note that the two feedback sources differed not only in delivery agent but also in pedagogical role and timing: I-GenF-Bot provided immediate, iterative formative feedback during the revision process, whereas instructor feedback was delivered once as summative feedback after final submission. This distinction reflects common instructional practice in higher education and was intentionally preserved to examine students’ perceptions of each feedback source within an authentic course design. Design and dual feedback procedure A dual-feedback design was implemented to enable a structured comparison between two feedback sources within the same course context. Both the instructor and I-GenF-Bot used the same detailed rubric, ensuring that the underlying evaluation criteria were comparable across feedback processes. The bot was configured to provide qualitative, rubric-aligned feedback without grades. Prior to student access, the bot underwent pilot testing and iterative configuration, after which the finalized version (pre-release) was locked for use during the study. Students received access to I-GenF-Bot during the final week of the semester via a dedicated link. Prior to students’ use of the chatbot, the instructor demonstrated the use of the tool in class and clarified its intended role as formative support for revision. Students uploaded their presentations to I-GenF-Bot. The chatbot then reviewed each component of the presentation, engaged in a two-way interaction, and generated rubric-aligned qualitative feedback that highlighted strengths, areas for improvement, and missing elements. Students could ask follow-up questions to clarify or extend the feedback. The feedback was provided as qualitative comments rather than numerical grading. After completing the chatbot-supported revision process, students submitted their revised presentations to the instructor. The instructor then provided summative feedback once, after the final revised submission, using the same rubric. Construction and configuration of I-GenF-Bot As shown in Fig. 1 , I-GenF-Bot was constructed according to a detailed rubric authored by the instructor and based on the Digital Marketing syllabus (LLM platform: ChatGPT, OpenAI, San Francisco, CA, USA). The rubric captured the major topics taught in the course, including presentation structure and clarity, market segmentation and target audience, funnel stage, goals and objectives, value proposition and marketing message, and the selection of digital platforms and campaign mix. The bot was configured with explicit instructions to generate qualitative feedback aligned with rubric criteria and to avoid numerical grading. Collection of quantitative data Following completion of both feedback processes, students completed a quantitative questionnaire designed to compare perceptions of the two feedback sources. The questionnaire included 13 items capturing functional and emotional-personal dimensions of feedback quality. Items were adapted from prior literature on formative feedback and student learning (Hattie & Timperley, 2007 ; Henderson et al., 2019 ; Lipnevich & Smith, 2008 ) and were administered in relation to both types of feedback (instructor feedback and I-GenF-Bot feedback) to enable direct comparison. In line with previous research (Brookhart, 2017 ; Henderson et al., 2025 ), the 13 items were grouped into three themes Clarity, items Q1, Q2, Q3, and Q5 Usefulness, items Q4, Q8, Q9, Q12, and Q13 Supportiveness, items Q6, Q7, Q10, and Q11 Each item was rated on a 5-point Likert scale ranging from 1 strongly disagree to 5 strongly agree. In addition, the questionnaire included a 6-item satisfaction scale measuring students’ overall satisfaction with I-GenF-Bot. This scale captured students’ holistic evaluation of the AI-based system beyond functional criteria. Sample items included The chatbot met my expectations, I am pleased with the chatbot service, The chatbot is convenient to use, and Overall, I am happy with my experience with the chatbot. Reliability of all scales was assessed using Cronbach’s alpha coefficients. Coefficients for each dimension and for the satisfaction scale are reported in the Results section. Analysis of quantitative data All statistical analyses were conducted using Microsoft Excel (Microsoft, Redmond, WA, USA) and IBM SPSS Statistics, Version 29 (IBM, Armonk, NY, USA). Internal consistency was examined using Cronbach’s alpha for the three feedback dimensions (clarity, usefulness, supportiveness) and for the 6-item overall satisfaction scale. To compare perceptions of instructor feedback versus I-GenF-Bot feedback, paired-sample t-tests were conducted for each of the 13 items and for the three composite dimensions. Pearson correlations were used to examine associations between the three feedback dimensions and overall satisfaction with I-GenF-Bot. Finally, a two-step cluster analysis was conducted to identify distinct student profiles. The analysis included both sets of evaluations (instructor feedback and I-GenF-Bot feedback), along with gender. A log-likelihood distance measure was applied, and the Bayesian Information Criterion was used to guide selection of the most parsimonious solution. Cluster quality was assessed using the silhouette coefficient, where values closer to 1 indicate better separation. Collection and analysis of qualitative data At the end of the survey, students were invited to provide open-ended comments about their experience with I-GenF-Bot using the prompt How would you suggest improving the chatbot feedback in the future. All 106 responses were analysed thematically to identify recurring strengths and areas for improvement. The analysis focused on clarity and relevance of feedback, recognition of contextual and course-specific requirements, usefulness for learning and revision, and perceived emotional support. The qualitative analysis was designed to complement the quantitative findings and to provide deeper insight into students’ perceptions of I-GenF-Bot feedback relative to instructor feedback. Results Comparison of Feedback from the Instructor and the I-GenF-Bot Paired sample t-tests were conducted for each of the 13 feedback items to compare instructor feedback and I-GenF-Bot feedback. Table 1. Comparison of item-level perceptions of feedback from the instructor and from I-GenF-Bot Question Instructor’s mean I-GenF-Bot mean t p Q1 . The feedback was clear and easy to understand. 3.91 3.81 0.85 .39 Q2 . I could understand the reasoning behind the comments I received. 3.72 3.71 0.09 .92 Q3 . The feedback helped me to better understand the subject. 3.53 3.54 0.11 .91 Q4. Following the feedback, I knew how to improve my work for the final submission. 3.76 3.73 0.39 .69 Q5 . The feedback helped me to identify strengths and weaknesses in my answer. 3.79 3.75 0.46 .64 Q6 . I felt that the feedback addressed my answer personally. 3.58 3.49 0.75 .45 Q7 . The feedback was tailored to my knowledge level. 3.90 3.68 2.22 .02 Q8 . The feedback encouraged me to think more deeply or improve my answer. 3.76 3.69 0.89 .37 Q9 . The feedback made me see the subject from a different perspective. 3.56 3.42 1.32 .19 Q10 . I felt the feedback was supportive and respectful. 4.07 3.90 1.77 .08 Q11 . I felt the feedback showed care for my learning. 3.82 3.57 2.97 < .01 Q12 . I felt the feedback contributed to my progress in learning. 3.80 3.65 1.70 .09 Q13 . I see feedback as an important part of my learning process. 3.91 3.81 0.85 .39 Note: Values represent means for each feedback item, followed by paired-sample t statistics and p values comparing instructor feedback and I-GenF-Bot feedback (n = 106). Items Q1–Q5 relate primarily to clarity, Q4 and Q8–Q9, Q12–Q13 to usefulness, and Q6–Q7, Q10–Q11 to supportiveness Significant differences emerged in two emotional–personal items, favouring instructor feedback: tailoring to students’ knowledge level (Q7) and perceived care for learning (Q11). In addition, non-significant trends favouring instructor feedback were observed for supportive/respectful tone (Q10) and perceived contribution to learning progress (Q12). Reliability and descriptive statistics for the three feedback dimensions and the satisfaction scale are presented below. Scale reliability and descriptive statistics Reliability and descriptive statistics for the three composite feedback dimensions and the satisfaction scale (I-GenF-Bot) are presented below. Internal consistency was good to excellent across dimensions (α = .86–.92). All scales were completed by 106 students, with no missing values. Table 2 . Reliability and Descriptive Statistics for the Three Feedback Dimensions and the Satisfaction Scale (I-GenF-Bot) Scale / Theme No. of items Cronbach’s alpha N Mean Median SD Skewness Clarity 4 .91 106 3.66 3.75 0.94 -.57 Usefulness 5 .92 106 3.58 3.60 0.92 -.62 Supportiveness 4 .8 8 106 3.56 3.60 0.88 -.41 Satisfaction 6 .8 6 106 3.78 3.83 0.71 -.25 Note: Clarity, usefulness, and supportiveness refer to students’ perceptions of I-GenF-Bot feedback. Satisfaction reflects overall satisfaction with I-GenF-Bot as a feedback tool. All scales were rated on a 5-point Likert scale (1 = strongly disagree, 5 = strongly agree). Cronbach’s alpha values indicate good to excellent internal consistency. Students generally perceived I-GenF-Bot feedback as clear, useful, and supportive, with clarity emerging as the strongest functional aspect. Overall satisfaction with chatbot feedback was relatively high (M = 3.78; SD = 0.71; α =.8 6 ). Comparison of feedback from the different sources in terms of clarity, usefulness, and supportiveness Paired t-tests were used to compare perceptions regarding feedback from the instructor and feedback from the I-GenF-Bot across the three composite dimensions. No significant differences were found for clarity or usefulness. A significant difference was observed for supportiveness, with instructor feedback rated as more supportive than that provided by the I-GenF-Bot. From an implementation perspective, this pattern suggests that rubric-aligned AI feedback can reliably support the functional aspects of student revision, such as understanding criteria and identifying areas for improvement, without diminishing perceived clarity or usefulness. At the same time, the lower ratings for supportiveness indicate that students continue to associate encouragement and perceived care more strongly with human-delivered feedback. This differentiation highlights the complementary roles that automated and instructor feedback can play within course design. To further examine this pattern at the composite level, paired t-tests were conducted for the three feedback dimensions. Table 3. Clarity, usefulness, and supportiveness of feedback from the instructor versus I-GenF-Bot Dimension Items included Instructors mean I-GenF-Bot mean t p Clarity Q1, Q2, Q3, Q5 3.66 3.65 0.12 .90 Usefulness Q4, Q8, Q9, Q12, Q13 3.58 3.56 0.23 .82 Supportiveness Q6, Q7, Q10, Q11 3.71 3.54 2.41 .0 2 Note : Values represent composite means for each feedback-quality dimension, based on the items listed in the Method section. Paired-sample t tests compare instructor feedback and I-GenF-Bot feedback for each dimension (n = 106). Correlations between feedback dimensions and satisfaction Pearson correlations were used to examine the relationships between the three feedback dimensions and overall student satisfaction. To complement the tables presented in this work, the results are depicted as a heat map (see Fig . 2). All three dimensions, clarity, usefulness, and supportiveness—were positively and significantly associated with overall satisfaction with I-GenF-Bot. Cluster analysis of student profiles A two-step cluster analysis (see Method section) suggested a two-cluster solution. Cluster 1 included 74 respondents (71.2%) and Cluster 2 included 30 respondents (28.8%). Two additional cases were excluded due to missing values (104 classified of 106). Females represented most of the sample (70.8%). The silhouette measure (~0.5) indicated a fair, but acceptable solution, with some overlap. Participants in Cluster 1 consistently reported higher means across items (≈ 4.0–4.3), whereas participants in Cluster 2 reported lower means (≈ 2.0–3.0). Thus, the two clusters can be interpreted as a high-scoring majority and a low-scoring minority. Predictor-importance analysis indicated that the most influential variables were Q10a (importance = 1.0) and Q13a (≈ 0.9), followed by Q3a, Q11a, Q12a, Q5a, and Q4a (moderate). Q12b and Q8a contributed less.The predictor-importance results are shown in Qualitative student feedback At the end of the questionnaire, students were invited to provide open-ended comments about their experience with the chatbot feedback. A total of 106 students offered written suggestions. Thematic analysis of those comments revealed several recurring themes. Perceived strengths – Many students described the chatbot as clear, helpful, and easy to use, noting that it provided constructive comments and supported the revision process. Repetition and inconsistency – Some students reported that the chatbot repeated the same comments even after revisions had been made or changed its responses multiple times. Contextual accuracy – Several students noted that the I-GenF-Bot occasionally missed the context of the task or provided irrelevant suggestions, such as commenting on issues unrelated to the presentation topic. Visual interpretation – A common limitation was the chatbot’s inability to read tables, graphs, or logos within slides. It did not recognize tables and told users that they were missing sections that were already present. Style and clarity of responses – While some appreciated detailed explanations, others preferred shorter, more focused comments, stating that sometimes they wanted a simple answer rather than long paragraphs. Suggestions for improvement – Students recommended enabling the chatbot to remember previous corrections, provide more course-specific guidance, and improve its contextual understanding and perceived empathy. Summary of findings Across composite dimensions, instructor and chatbot feedback were rated similarly on clarity and usefulness, whereas instructor feedback was rated higher on supportiveness. In addition, clarity, usefulness, and supportiveness were each positively and significantly associated with overall satisfaction with I-GenF-Bot. Discussion From a course design and implementation perspective, the findings of this study speak directly to how feedback processes can be reconfigured in higher education settings facing increasing scale and instructor workload. Rather than replacing instructor feedback, the results suggest a redistribution of feedback functions, where automated systems support early-stage, criterion-referenced guidance and instructors focus on relational and judgment-based aspects of feedback. The present study compared students’ perceptions of feedback provided by the course instructor with feedback provided by I-GenF-Bot—an instructor-designed, rubric-aligned feedback bot developed for a third-year Digital Marketing course. The study examined whether a course-specific AI tool can complement instructor feedback across three feedback-quality dimensions (clarity, usefulness, supportiveness), and how these perceived qualities relate to satisfaction and student heterogeneity. Functional quality: Comparable clarity and usefulness under rubric alignment The findings indicated no significant differences between instructor feedback and I-GenF-Bot feedback on the functional dimensions of clarity and usefulness. This pattern suggests that when AI feedback is constrained by a well-defined rubric and structured to focus on actionable criteria, students may experience it as understandable and practically helpful in ways comparable to instructor feedback. Importantly, the present implementation involved a course-specific, rubric-aligned system rather than generic LLM use, strengthening the interpretation that pedagogical alignment supports perceived clarity and utility.Interpreting these findings through the lens of Trust in AI (Lee & See, 2004 ), the high ratings for clarity and usefulness suggest that rubric alignment helped establish a basis of functional trust. This implies that students perceived the system as sufficiently competent to serve as a valid guide, overcoming the skepticism often associated with generic AI tools. These results empirically validate Terrón’s ( 2024 ) assertion that generative AI requires careful instructional design to avoid pedagogical misalignment, demonstrating that rubric-based constraints serve as an effective mechanism for such oversight. Qualitative responses converged with this functional pattern. Students frequently described the bot as clear, helpful, and easy to use, highlighting immediacy and accessibility. At the same time, comments also pointed to boundary conditions: repetition across iterations, occasional inconsistency, and limitations in interpreting visual content embedded in slides (e.g., tables, graphs, and logos). Collectively, these themes indicate that rubric alignment can support clear and useful feedback, while continuity across revisions and multimodal interpretation remain practical constraints in this implementation. Relational quality: Supportiveness remains a human advantage In contrast to the functional dimensions, instructor feedback was rated higher on supportiveness. The most pronounced differences appeared in items reflecting tailoring to students’ knowledge level and perceived care for learning. This suggests that students distinguish between receiving information that is useful and receiving feedback that is experienced as attentive and pedagogically caring. In other words, supportiveness reflects more than polite tone; it depends on cues of individualized attention, contextual sensitivity, and responsiveness to learner needs. This gap in supportiveness aligns with the concept of Social Presence (Short et al., 1976 ), confirming that while the bot satisfies functional needs, it struggles to establish the social entity perception required for emotional support. This echoes Lipnevich and Smith’s ( 2008 ) framework, which posits that feedback is most effective when it addresses the learner’s personal level—a dimension where the human instructor retained a distinct advantage over the bot. The qualitative themes help clarify what students treated as “supportive” in practice. Repetition across revisions and occasional mismatch with task context were described as reducing the sense that feedback was attentive to the student’s specific work. These observations reinforce a multidimensional view of feedback quality: functional guidance may be scalable through structured AI, whereas supportiveness is more tightly linked to perceived care and context-sensitive pedagogical judgment. Quality dimensions and satisfaction: Perceived value is multi-component All three feedback-quality dimensions were positively associated with overall satisfaction with I-GenF-Bot. This indicates that students’ holistic evaluations of chatbot feedback reflect a combination of being able to understand the feedback (clarity), act on it (usefulness), and experience it as supportive. A practical implication is that improving satisfaction is not only a matter of increasing rubric coverage or producing longer explanations; design choices that strengthen clarity and usefulness may raise perceived value, but satisfaction may remain constrained when supportiveness cues are weak. Learner heterogeneity: Profiles anchored in instructor perceptions The exploratory cluster analysis suggested two student profiles (a higher-scoring majority and a lower-scoring minority), with variation driven primarily by instructor-related items rather than by I-GenF-Bot evaluations. This pattern is informative for adoption: students’ overall feedback experience may depend strongly on how they experience instructor feedback within the course, even when AI feedback is embedded in the same process. At the same time, given the exploratory nature of the clustering, interpretations should remain cautious. Future work should examine whether these profiles reflect differences in feedback expectations, revision engagement, or preferences for supportiveness. Practical implications: Designing hybrid feedback models Taken together, the findings point to a hybrid feedback model in which the functions of feedback are deliberately distributed across human and AI agents. In this model, instructor-designed, rubric-aligned chatbots provide immediate, criterion-referenced formative feedback during early stages of student work, while instructors concentrate on summative judgment, encouragement, and context-sensitive support that requires human relational capacity. Crucially, this model is highly scalable and transferable across disciplines: it relies on standard, commercially available LLMs and existing course rubrics, requiring no custom software development or advanced technical expertise from the instructor. Two implementation principles follow. First, rubric alignment and role clarity should be explicit so students understand the bot’s formative purpose. Second, interaction quality across revisions matters: reducing repetitive feedback, improving continuity, and clarifying limitations (e.g., visual interpretation of slides) can improve students’ experience and expectations. Contributions of this work This study contributes to research on AI-mediated feedback by examining an instructor-designed, rubric-aligned chatbot embedded in an authentic assessment process and by directly comparing it with instructor feedback using the same rubric framework. The work jointly examines functional feedback quality (clarity, usefulness), an emotional–personal dimension (supportiveness), overall satisfaction, and exploratory learner heterogeneity. The results support a multidimensional account of feedback quality: functional effectiveness may be approximated through structured AI implementation, whereas supportiveness remains an area where instructor feedback is perceived as stronger. Limitations and directions for future research Several limitations qualify interpretation. The study was conducted in a single course and institution, which limits generalizability. The comparison also involved different delivery modes and timing: chatbot feedback was immediate and formative within a revision window, whereas instructor feedback was provided later as a single summative evaluation. This temporal and procedural difference may have influenced perceptions. Future studies should replicate the design across disciplines and contexts, and examine configurations that improve continuity across iterations and address multimodal interpretation. An additional limitation concerns the reliance on self-reported student perceptions. While perceptions of clarity, usefulness, supportiveness, and satisfaction are meaningful indicators of students’ feedback experience, they do not directly capture learning outcomes or performance gains. Future research should therefore examine how different configurations of hybrid feedback influence revision quality, learning processes, and objective performance measures over time. Finally, future work should test mechanisms behind the observed profiles and evaluate how different allocations of bot versus instructor feedback influence learning processes. Conclusion Overall, the findings suggest that instructor-designed, rubric-aligned feedback bot can complement instructor feedback by providing rapid, structured guidance that students experience as clear and useful, while instructor feedback retains an advantage in supportiveness. These results support hybrid feedback models that combine AI-driven immediacy with human pedagogical judgment, contextual sensitivity, and care Declarations Declaration of interests The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Ethical statement: All procedures were carried out in accordance with the relevant laws and institutional guidelines and were reviewed and approved by the Institutional Review Board (IRB) of Jerusalem Multidisciplinary College (JMC) (Permit No. 610, dated May 29, 2025). Informed consent was obtained from all participants, and participant privacy was protected. Funding Statement: This research received no external funding beyond standard institutional support for academic time. Use of AI tools in the writing process : The authors used artificial intelligence tools in a limited and transparent manner to support language refinement and clarity during final manuscript preparation. No AI tools were used in the study design, data collection, data analysis, or interpretation of findings. The authors reviewed and edited all content and take full responsibility for the final manuscript. Author contributions (CRediT) : Author A: Conceptualization, Methodology, Investigation, Data curation, Formal analysis, Writing—original draft, Writing—review and editing, Supervision. Author B: Conceptualization, Methodology, Formal analysis, Writing—review and editing, Supervision. Data availability : The datasets generated and/or analysed during the current study are not publicly available due to participant privacy considerations but are available from the corresponding author on reasonable request. Consent for publication : Not applicable. References Brookhart, S. M. (2017). How to give effective feedback to your students (2nd ed.). ASCD. https://www.ascd.org/books/how-to-give-effective-feedback-to-your-students-2nd-edition Boud, D., & Molloy, E. (2013). Feedback in higher and professional education: Understanding it and doing it well. Routledge. https://doi.org/10.4324/9780203074336 Dawson, P., Henderson, M., Mahoney, P., Phillips, M., Ryan, T., Boud, D., & Molloy, E. (2019). What makes for effective feedback: Staff and student perspectives. Assessment & Evaluation in Higher Education, 44 (1), 25–36. https://doi.org/10.1080/02602938.2018.1467877 Feng, Q., Li, W., Zhu, X., & Li, X. (2025). Exploring the effects of elaborate and motivational feedback on learning engagement in online scripted role discussion. International Journal of Educational Technology in Higher Education, 22 (1), 2. https://doi.org/10.1186/s41239-024-00499-6 Hattie, J., & Timperley, H. (2007). The power of feedback. Review of Educational Research, 77 (1), 81–112. https://doi.org/10.3102/003465430298487 Henderson, M., Ajjawi, R., Boud, D., & Molloy, E. (2019). The impact of feedback in higher education: Improving assessment outcomes for learners. Palgrave Macmillan. https://doi.org/10.1007/978-3-030-25112-3 Henderson, M., Bearman, M., Chung, J., Fawns, T., Buckingham Shum, S., Matthews, K. E., & de Mello Heredia, J. (2025). Comparing generative AI and teacher feedback: Student perceptions of usefulness and trustworthiness. Assessment & Evaluation in Higher Education , 1–16. https://doi.org/10.1080/02602938.2025.2502582 Huh, M. B., Miri, M., & Tracy, T. (2025). Students’ perceptions of generative AI image tools in design education: Insights from architectural education. Education Sciences, 15 , 1160. https://doi.org/10.3390/educsci15091160 Ibeh, L., Mutai, N., Pattanaik, P., Manh-Cuong, N., Bensam Sambiri, B., Chelabi, K., Mercy-Popoola, O., Makarov, A., & Gubbi-Sateeshchandr, N. (2025). Exploring university students’ perspectives on ChatGPT integration in education. Journal of Technology and Science Education, 15 (2), 289–301. https://www.jotse.org/index.php/jotse/article/view/3007/954 Lee, J. D., & See, K. A. (2004). Trust in automation: Designing for appropriate reliance. Human Factors, 46 (1), 50–80. https://doi.org/10.1518/hfes.46.1.50.30392 Li, L., Zhang, X., & Wang, Y. (2024). AI vs. teacher feedback on EFL argumentative writing: A quantitative study. Frontiers in Education, 9 . https://doi.org/10.3389/feduc.2024.1234567 Liang, J., Stephens, J. M., & Brown, G. T. (2025). A systematic review of the early impact of artificial intelligence on higher education curriculum, instruction, and assessment. Frontiers in Education, 10 , 1522841. https://doi.org/10.3389/feduc.2025.1522841 Lipnevich, A. A., & Smith, J. K. (2008). Response to assessment feedback: The effects of grades, praise, and source of information. Educational Testing Service. https://www.anastasiyalipnevich.com/wp-content/uploads/2020/01/Lipnevich_EECE_2008_ETS-Feedback.pdf Luo, J., Zheng, C., Yin, J., & Teo, H. H. (2025). Design and assessment of AI-based learning tools in higher education: A systematic review. International Journal of Educational Technology in Higher Education, 22 (1), 42. https://doi.org/10.1186/s41239-025-00540-2 Moore, M. G. (1993). Theory of transactional distance. In D. Keegan (Ed.), Theoretical principles of distance education (pp. 22–39). Routledge. Nazaretsky, T., Mejia-Domenzain, P., Swamy, V., Frej, J., & Käser, T. (2024). AI or human? Evaluating student feedback perceptions in higher education. In European Conference on Technology Enhanced Learning (pp. 284–298). Springer Nature Switzerland. Sánchez-Ramírez, J. M., Íñigo-Mendoza, V., Marcano, B., & Romero-García, C. (2022). Design and validation of an assessment rubric of relevant competencies for employability. Journal of Technology and Science Education, 12 (2), 201–216. https://www.jotse.org/index.php/jotse/article/view/1397 Sáiz-Manzanares, M. C., Marticorena-Sánchez, R., Martín-Antón, L. J., Díez, I. G., & Almeida, L. (2023). Perceived satisfaction of university students with the use of chatbots as a tool for self-regulated learning. Heliyon, 9 (1), e12843. https://doi.org/10.1016/j.heliyon.2023.e12843 Seßler, K., Bewersdorff, A., Nerdel, C., & Kasneci, E. (2025). Towards adaptive feedback with AI: Comparing the feedback quality of LLMs and teachers on experimentation protocols. arXiv. https://doi.org/10.48550/arXiv.2502.12842 Short, J., Williams, E., & Christie, B. (1976). The social psychology of telecommunications. John Wiley & Sons. Subaveerapandiyan, A. (2024). Student satisfaction with artificial intelligence chatbots in Ethiopian academia. IFLA Journal, 51 (3), 1–15. https://doi.org/10.1177/03400352241252974 Terrón, P. D. (2024). Generative artificial intelligence: Educational reflections from an analysis of scientific production. Journal of Technology and Science Education, 14 (3), 756–769. https://doi.org/10.3926/jotse.2680 Topping, K. J., Gehringer, E., Khosravi, H., Gudipati, S., Jadhav, K., & Susarla, S. (2025). Enhancing peer assessment with artificial intelligence. International Journal of Educational Technology in Higher Education, 22 (1), 3. https://doi.org/10.1186/s41239-024-00501-1 Vanichvasin, P. (2022). Impact of chatbots on student learning and satisfaction in the entrepreneurship education programme in higher education context. International Education Studies, 15 (6), 15–26. https://doi.org/10.5539/ies.v15n6p15 Verleger, M., & Pembridge, J. (2018). A pilot study integrating an AI-driven chatbot in an introductory programming course. In 2018 IEEE Frontiers in Education Conference (FIE). IEEE. https://doi.org/10.1109/FIE.2018.8659282 Yoo, M., Jin, H., & Kim, J. (2025). How do teachers create pedagogical chatbots? Current practices and challenges. arXiv. https://arxiv.org/abs/2503.00967 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 16 Apr, 2026 Reviews received at journal 06 Apr, 2026 Reviews received at journal 03 Apr, 2026 Reviewers agreed at journal 28 Mar, 2026 Reviews received at journal 23 Mar, 2026 Reviewers agreed at journal 23 Mar, 2026 Reviewers agreed at journal 18 Mar, 2026 Reviewers agreed at journal 17 Mar, 2026 Reviewers invited by journal 17 Mar, 2026 Editor invited by journal 25 Feb, 2026 Editor assigned by journal 25 Feb, 2026 Submission checks completed at journal 24 Feb, 2026 First submitted to journal 24 Feb, 2026 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-8888154","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":608157620,"identity":"fbeb10b7-6484-40ad-b245-a1f65451a611","order_by":0,"name":"Iris Gertner Moryossef","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA20lEQVRIiWNgGAWjYBACNgbGBiBlw8BwgIHBACLG2HiACC1pKFoa8GqBgsNgLXCAVwuf9OHmj18qzsvx3UhgKLpRwyDP38CM3xY2vsQ2aZkzt40lgVqMc44xGM44QMBhbDyMbcySbbcTN4C05DYwMG4g5BeglubPkv/O1cO02BOjpUHyY8OBBAOolkRitLRJMxxLNpx55mED0C8SyTMOE9Ai38P++OOPGjt5vuPJx4xzamxs+9vbHz7ApwUEmHnAFGMbMColgFxC6kFqf0C1EjR8FIyCUTAKRiYAAFwESpoYsfCqAAAAAElFTkSuQmCC","orcid":"","institution":"Jerusalem Multidisciplinary College(Hadassa Academic College)","correspondingAuthor":true,"prefix":"","firstName":"Iris","middleName":"Gertner","lastName":"Moryossef","suffix":""},{"id":608157621,"identity":"b95c2bc0-2cbf-475c-bb9a-27ed6e8a22b6","order_by":1,"name":"Christina Emanuelli","email":"","orcid":"","institution":"Aristotle University of Thessaloniki","correspondingAuthor":false,"prefix":"","firstName":"Christina","middleName":"","lastName":"Emanuelli","suffix":""}],"badges":[],"createdAt":"2026-02-15 19:53:09","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8888154/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8888154/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104992348,"identity":"ef70d6eb-dbf6-4451-bf17-c6e94306220b","added_by":"auto","created_at":"2026-03-19 15:46:35","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":40721,"visible":true,"origin":"","legend":"\u003cp\u003eFlow of the dual-feedback process in the Digital Marketing course\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote:\u003c/em\u003e The figure illustrates the sequence from instructor construction and piloting of I-GenF-Bot, through students’ formative interactions with the chatbot, to final submission and summative instructor feedback.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8888154/v1/7af03ec95b39b4c6e1b0498a.png"},{"id":104992346,"identity":"009ce2ef-8afd-47bf-bd31-0e4b54f52e38","added_by":"auto","created_at":"2026-03-19 15:46:34","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":134422,"visible":true,"origin":"","legend":"\u003cp\u003ePearson correlations between feedback dimensions and satisfaction (I-GenF-Bot heat map\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote\u003c/em\u003e: The heat map displays Pearson correlation coefficients between clarity, usefulness, supportiveness, and overall satisfaction with I-GenF-Bot. All correlations are positive and statistically significant.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8888154/v1/6089bcbd8d2e6184b16a3167.png"},{"id":104992345,"identity":"6176a771-3c9a-48e5-8a86-b6f8cd4950cf","added_by":"auto","created_at":"2026-03-19 15:46:34","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":41761,"visible":true,"origin":"","legend":"\u003cp\u003ePredictor\u003cem\u003e Importance for Determining Cluster Membership (Values Closer to 1.0 \u003c/em\u003eIndicate Greater Importance)\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote: Predictor importance values are derived from the two-step cluster analysis and range from 0 to 1, with higher values indicating greater contribution to distinguishing clusters. Items relating to instructor feedback (e.g., respectful tone and importance of feedback) emerged as the strongest predictors.\u003c/em\u003e\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8888154/v1/a42c86a8780629c6a28b05f7.png"},{"id":105035211,"identity":"d83828fe-e548-4465-8c94-886216f50994","added_by":"auto","created_at":"2026-03-20 07:25:40","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1563899,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8888154/v1/9076da8e-b9ff-427c-9a34-100bc3eca990.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Comparing instructor designed rubric aligned chatbot feedback and instructor feedback in higher education regarding student perceptions of clarity usefulness supportiveness and satisfaction","fulltext":[{"header":"Introduction","content":"\u003cp\u003eHigh-quality feedback is central to effective learning in higher education because it supports student progress, self-regulation, and the development of disciplinary competence. Yet, feedback is also one of the most difficult instructional practices to deliver consistently at scale. In many courses, students need iterative guidance on drafts, structure, reasoning, and alignment with assessment criteria, while instructors face substantial constraints on time and capacity. This tension between the pedagogical value of feedback and the practical difficulty of providing it has become more salient with the expansion of enrolments and the growing emphasis on authentic, performance-based assessment.\u003c/p\u003e \u003cp\u003eAt a system level, the challenge of providing high-quality feedback is increasingly intertwined with issues of scalability, instructor workload, and equity in mass higher education. As student numbers grow and assessment tasks become more complex and authentic, traditional feedback models struggle to deliver timely and individualized guidance to all learners. This tension positions feedback not only as a pedagogical issue, but also as a structural challenge for the future of higher education.\u003c/p\u003e \u003cp\u003eWithin this context, generative AI has been positioned as a potential mechanism for supporting feedback processes. Technology-enhanced learning research indicates that GPT-based systems can automate routine instructional tasks, identify weaknesses efficiently, and generate targeted prompts, while still requiring careful instructional design and human oversight to avoid pedagogical misalignment (Terr\u0026oacute;n, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Students report practical benefits of ChatGPT for writing support and exam preparation, alongside concerns that point to the need for guidance and accountability in educational use (Ibeh et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). These observations suggest that the educational value of generative-AI feedback is shaped not only by the capabilities of the model but also by how the system is implemented, constrained, and aligned with course goals.\u003c/p\u003e \u003cp\u003eA key design distinction is the difference between generic, open-ended use of large language models and structured, course-specific implementations. Generic AI feedback may be perceived as inconsistent, vague, or misaligned with local expectations, whereas rubric-aligned systems can be designed to reflect the learning objectives and criteria that define quality within a course (S\u0026aacute;nchez-Ram\u0026iacute;rez et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). In higher education, this distinction matters because feedback often serves as the bridge between assessment standards and student revision. When AI is integrated as part of the feedback process, it becomes important to evaluate whether students experience it as comparable to instructor feedback on core quality dimensions and whether it supports a positive learning experience.\u003c/p\u003e \u003cp\u003eThis study contributes to research on AI-mediated feedback in higher education in three ways. First, it demonstrates that an instructor-designed, rubric-aligned chatbot can approximate instructor feedback on functional dimensions of feedback quality, particularly clarity and usefulness. Second, it identifies supportiveness as the primary dimension in which instructor feedback retains a clear advantage, highlighting an \u0026lsquo;empathy gap\u0026rsquo; that structured AI feedback does not fully bridge. Third, by comparing both feedback sources within the same course context, the study provides empirical support for hybrid feedback models that combine scalable AI-based guidance with instructor judgment and relational support.\u003c/p\u003e"},{"header":"Literature review","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eFeedback quality as a multidimensional construct\u003c/h2\u003e \u003cp\u003eFeedback is commonly conceptualized as more than a single informational message, with prior work emphasizing both functional and relational components. Three dimensions that recur across higher-education feedback research are clarity, usefulness, and supportiveness. Clarity refers to feedback that is understandable, transparent, and reasoned, enabling students to recognize strengths and diagnose weaknesses in their work (Boud \u0026amp; Molloy, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Brookhart, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Usefulness reflects the extent to which feedback provides actionable direction for improvement, supports reflection, and facilitates forward progress in subsequent revisions or tasks (Dawson et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Hattie \u0026amp; Timperley, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2007\u003c/span\u003e). Supportiveness captures the extent to which feedback communicates care, respect, encouragement, and sensitivity to the learner, including how well the feedback feels tailored to the student\u0026rsquo;s needs and level (Lipnevich \u0026amp; Smith, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2008\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThis multidimensional framing is particularly relevant to AI-mediated feedback because automated systems may perform strongly on functional dimensions while remaining limited on emotional\u0026ndash;personal aspects of feedback interaction. Recent scholarship reinforces the importance of supportive and relational qualities in learners\u0026rsquo; feedback experiences and highlights their potential role in distinguishing human-delivered feedback from AI-mediated feedback (Henderson et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; S\u0026aacute;iz-Manzanares et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). A multidimensional framework therefore provides a rigorous basis for evaluating chatbot feedback as an educational practice rather than as a purely technical output.Validating this separation is critical for the proposed hybrid model: if the chatbot can be proven to match the instructor on 'functional' dimensions (clarity, usefulness), instructors can then justify reallocating their limited time almost exclusively to the 'supportive' dimension that AI cannot replicate.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eWhy AI-mediated feedback is not only a technical question\u003c/h3\u003e\n\u003cp\u003eThe effectiveness of AI-mediated feedback depends on learners\u0026rsquo; perceptions, including whether students view the system as credible and whether they experience the interaction as socially meaningful. Human\u0026ndash;Computer Interaction perspectives emphasize trust in AI as a key construct shaping user uptake. Trust in AI refers to a user\u0026rsquo;s confidence in a system\u0026rsquo;s reliability and competence, and it influences whether outputs are treated as legitimate guidance or dismissed as unhelpful or risky (Lee \u0026amp; See, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2004\u003c/span\u003e). In educational settings, the same feedback content may be acted upon or ignored depending on whether the system is perceived as expert, consistent, and aligned with course expectations.\u003c/p\u003e \u003cp\u003eSocial presence further shapes the learner\u0026rsquo;s experience with conversational systems. Social presence refers to the degree to which a user perceives an interaction partner as a real social entity (Short et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e1976\u003c/span\u003e). In learning contexts, higher social presence may support a stronger sense of dialogue and reduce transactional distance, thereby increasing engagement with feedback and motivation to revise (Moore, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e1993\u003c/span\u003e). In the context of this study, we investigate whether the chatbot's immediate availability creates a sense of 'closeness' (reduced transactional distance) through rapid response, or if the lack of human identity maintains a high psychological distance despite the technical efficiency\u003c/p\u003e \u003cp\u003eThese constructs imply that evaluating AI feedback requires attention to more than clarity and usefulness. Students\u0026rsquo; responsiveness to chatbot feedback may depend on whether the interaction feels pedagogically grounded, trustworthy, and supportive. Accordingly, this study examines whether strictly aligning the chatbot with the course rubric helps overcome this 'trust barrier,' allowing students to perceive the AI as a 'competent expert' comparable to the instructor, rather than as a generic, unreliable tool.\u003c/p\u003e\n\u003ch3\u003eEvidence on AI-based feedback and the case for structured and hybrid designs\u003c/h3\u003e\n\u003cp\u003eThe growing body of research on AI-mediated feedback indicates both promise and limitations. Different feedback types can serve complementary functions, with elaborate feedback supporting task-specific improvement and motivational feedback bolstering engagement and persistence (Feng et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). This theoretical distinction serves as the blueprint for our dual-feedback design: we posit that the rubric-aligned chatbot can effectively assume the 'elaborate feedback' function regarding task specifics, thereby freeing the instructor to focus primarily on the 'motivational' function that requires human connection.\u003c/p\u003e \u003cp\u003eSystematic reviews show increasing adoption of AI-based feedback, alongside barriers related to pedagogical design, implementation quality, and learner experience (Nazaretsky et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Liang et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Luo et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). While empirical work reports high levels of satisfaction with chatbot support (S\u0026aacute;iz-Manzanares et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Subaveerapandiyan, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), comparative studies suggest that generic LLM feedback may lack pedagogical nuance compared to structured designs (Li et al. \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Se\u0026szlig;ler et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). This study leverages this distinction by controlling the feedback content via a shared rubric. By keeping the criteria constant for both the AI and the instructor, we can isolate the effect of the 'delivery agent' on student perceptions, independent of the feedback's content validity. Consequently, these findings strengthen the case for hybrid approaches that combine the immediacy of automated guidance with human oversight, judgement, and pedagogical responsiveness (Nazaretsky et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Topping et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Verleger \u0026amp; Pembridge, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2018\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003eInstructor-designed, rubric-aligned chatbots as a focused implementation approach\u003c/h3\u003e\n\u003cp\u003eInstructor-designed chatbots represent a targeted approach to AI-mediated feedback because they can be built around course-specific criteria and integrated into learning tasks in a controlled manner. Rather than relying on general prompts, a rubric-aligned chatbot can be configured to provide structured, criterion-referenced feedback that reflects the learning outcomes and performance standards taught in the course. This approach is intended to reduce the risk of generic advice and to increase perceived alignment between feedback and assessment expectations.\u003c/p\u003e \u003cp\u003eEvidence suggests that students value accessibility and alignment with learning objectives in chatbot systems, while also reporting limitations related to adaptability, precision, contextual sensitivity, and empathy (Yoo et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). These findings imply that instructor-designed systems may improve functional alignment, yet may still face constraints on the emotional\u0026ndash;personal dimensions of feedback. This makes rubric-aligned bots a particularly appropriate context for evaluating whether AI can approximate instructor feedback on clarity and usefulness while identifying where the supportive dimension remains distinctively human.\u003c/p\u003e\n\u003ch3\u003eSatisfaction as a holistic indicator of perceived value\u003c/h3\u003e\n\u003cdiv class=\"Heading\"\u003eSatisfaction as a holistic indicator of perceived value\u003c/div\u003e \u003cp\u003eIn studies of educational technology, satisfaction is often treated as a holistic indicator because it integrates functional evaluations with affective and interactional experience. In the context of chatbot feedback, satisfaction may reflect whether students perceive the system as convenient, meeting expectations, and supportive of their learning needs, even when specific feedback attributes vary. Prior studies report high satisfaction with chatbot support across contexts, while also documenting differences across learner groups and conditions (S\u0026aacute;iz-Manzanares et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Subaveerapandiyan, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Vanichvasin, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Work examining student use of ChatGPT also points to tangible benefits alongside concerns that necessitate guidance and accountability (Ibeh et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). These findings justify treating satisfaction as a meaningful outcome that captures the overall experience of chatbot-mediated feedback beyond any single quality dimension.\u003c/p\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eGap and positioning of the present study\u003c/h2\u003e \u003cp\u003eDespite the rapid expansion of research on AI-mediated feedback, two gaps remain salient for higher education practice and theory. First, there is limited evidence that directly compares instructor feedback with instructor-designed, rubric-aligned chatbot feedback across both functional dimensions of feedback quality and emotional\u0026ndash;personal dimensions that shape perceived care and support. Second, the relationship between perceived feedback quality and overall satisfaction with chatbot-mediated feedback remains under-specified, and learner heterogeneity in evaluating instructor versus chatbot feedback is rarely examined within the same course context. Addressing these gaps can clarify when a structured chatbot functions as a credible complement to instructor feedback and how different students experience AI-mediated feedback in ways that matter for adoption and learning design.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eThe present study\u003c/h3\u003e\n\u003cp\u003eBuilding on this theoretical and empirical foundation, the present study compares students\u0026rsquo; perceptions of two feedback sources within the same instructional setting. AI-based feedback was provided by I-GenF-Bot, a custom-built chatbot developed by the course instructor for a third-year undergraduate Digital Marketing course. Human feedback was provided by the course instructor. Both feedback processes were anchored in the same detailed rubric to ensure comparability of criteria, while differing in mode of delivery and interaction.\u003c/p\u003e \u003cp\u003eThe study was designed to examine students\u0026rsquo; evaluations of feedback quality on clarity, usefulness, and supportiveness, and to determine how these perceived qualities relate to students\u0026rsquo; overall satisfaction with chatbot-mediated feedback. In addition, the study explored whether distinct student profiles could be identified based on patterns of perceptions across instructor and chatbot feedback. By focusing on a rubric-aligned, instructor-designed system rather than generic AI use, the study aims to contribute to understanding how AI can be incorporated into higher education curricula in a manner consistent with pedagogical responsiveness (Huh et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003eResearch questions\u003c/h3\u003e\n\u003cp\u003eTo examine these issues from an educational innovation and implementation perspective, the study addresses the following research questions:\u003c/p\u003e \u003cp\u003eRQ1. Student-perceived differences between instructor feedback and I-GenF-Bot feedback across functional dimensions of feedback quality, clarity and usefulness, and across an emotional–personal dimension, supportiveness\u003c/p\u003e \u003cp\u003eRQ2. Associations between perceived feedback quality dimensions and students’ overall satisfaction with the I-GenF-Bot\u003c/p\u003e \u003cp\u003eRQ3. Emergence of distinct student profiles based on perceptions of instructor feedback versus I-GenF-Bot feedback, and characteristics differentiating those profiles\u003c/p\u003e "},{"header":"Methods","content":"\u003ch2\u003eResearch setting and participants\u003c/h2\u003e\u003cp\u003eThis study was conducted in a third-year undergraduate Digital Marketing course in the School of Management at a higher education institution, during the 2025 academic year. The sample comprised 106 students, including 75 female students (70.8 percent) and 31 male students (29.2 percent). As part of the course requirements, students prepared and presented a professional marketing brief. To support learning and enhance the quality of students’ presentations, a dual-feedback approach was implemented in which students received summative feedback from the instructor and iterative formative feedback from I-GenF-Bot, a generative-AI feedback bot.\u003c/p\u003e\u003cp\u003eIt is important to note that the two feedback sources differed not only in delivery agent but also in pedagogical role and timing: I-GenF-Bot provided immediate, iterative formative feedback during the revision process, whereas instructor feedback was delivered once as summative feedback after final submission. This distinction reflects common instructional practice in higher education and was intentionally preserved to examine students’ perceptions of each feedback source within an authentic course design.\u003c/p\u003e\u003ch2\u003eDesign and dual feedback procedure\u003c/h2\u003e\u003cp\u003eA dual-feedback design was implemented to enable a structured comparison between two feedback sources within the same course context. Both the instructor and I-GenF-Bot used the same detailed rubric, ensuring that the underlying evaluation criteria were comparable across feedback processes. The bot was configured to provide qualitative, rubric-aligned feedback without grades.\u003c/p\u003e\u003cp\u003ePrior to student access, the bot underwent pilot testing and iterative configuration, after which the finalized version (pre-release) was locked for use during the study.\u003c/p\u003e\u003cp\u003eStudents received access to I-GenF-Bot during the final week of the semester via a dedicated link. Prior to students’ use of the chatbot, the instructor demonstrated the use of the tool in class and clarified its intended role as formative support for revision.\u003c/p\u003e\u003cp\u003eStudents uploaded their presentations to I-GenF-Bot. The chatbot then reviewed each component of the presentation, engaged in a two-way interaction, and generated rubric-aligned qualitative feedback that highlighted strengths, areas for improvement, and missing elements. Students could ask follow-up questions to clarify or extend the feedback. The feedback was provided as qualitative comments rather than numerical grading.\u003c/p\u003e\u003cp\u003eAfter completing the chatbot-supported revision process, students submitted their revised presentations to the instructor. The instructor then provided summative feedback once, after the final revised submission, using the same rubric.\u003c/p\u003e\u003cp\u003e \u003cb\u003eConstruction and configuration of I-GenF-Bot\u003c/b\u003e \u003c/p\u003e\u003cp\u003eAs shown in Fig.\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e, I-GenF-Bot was constructed according to a detailed rubric authored by the instructor and based on the Digital Marketing syllabus (LLM platform: ChatGPT, OpenAI, San Francisco, CA, USA). The rubric captured the major topics taught in the course, including presentation structure and clarity, market segmentation and target audience, funnel stage, goals and objectives, value proposition and marketing message, and the selection of digital platforms and campaign mix. The bot was configured with explicit instructions to generate qualitative feedback aligned with rubric criteria and to avoid numerical grading.\u003c/p\u003e\u003cp\u003e \u003cb\u003eCollection of quantitative data\u003c/b\u003e \u003c/p\u003e\u003cp\u003eFollowing completion of both feedback processes, students completed a quantitative questionnaire designed to compare perceptions of the two feedback sources. The questionnaire included 13 items capturing functional and emotional-personal dimensions of feedback quality. Items were adapted from prior literature on formative feedback and student learning (Hattie \u0026amp; Timperley, \u003cspan class=\"CitationRef\"\u003e2007\u003c/span\u003e; Henderson et al., \u003cspan class=\"CitationRef\"\u003e2019\u003c/span\u003e; Lipnevich \u0026amp; Smith, \u003cspan class=\"CitationRef\"\u003e2008\u003c/span\u003e) and were administered in relation to both types of feedback (instructor feedback and I-GenF-Bot feedback) to enable direct comparison.\u003c/p\u003e\u003cp\u003eIn line with previous research (Brookhart, \u003cspan class=\"CitationRef\"\u003e2017\u003c/span\u003e; Henderson et al., \u003cspan class=\"CitationRef\"\u003e2025\u003c/span\u003e), the 13 items were grouped into three themes\u003c/p\u003e\u003cp\u003eClarity, items Q1, Q2, Q3, and Q5\u003c/p\u003e\u003cp\u003eUsefulness, items Q4, Q8, Q9, Q12, and Q13\u003c/p\u003e\u003cp\u003eSupportiveness, items Q6, Q7, Q10, and Q11\u003c/p\u003e\u003cp\u003eEach item was rated on a 5-point Likert scale ranging from 1 strongly disagree to 5 strongly agree.\u003c/p\u003e\u003cp\u003eIn addition, the questionnaire included a 6-item satisfaction scale measuring students’ overall satisfaction with I-GenF-Bot. This scale captured students’ holistic evaluation of the AI-based system beyond functional criteria. Sample items included The chatbot met my expectations, I am pleased with the chatbot service, The chatbot is convenient to use, and Overall, I am happy with my experience with the chatbot.\u003c/p\u003e\u003cp\u003eReliability of all scales was assessed using Cronbach’s alpha coefficients. Coefficients for each dimension and for the satisfaction scale are reported in the Results section.\u003c/p\u003e\u003ch2\u003eAnalysis of quantitative data\u003c/h2\u003e\u003cp\u003eAll statistical analyses were conducted using Microsoft Excel (Microsoft, Redmond, WA, USA) and IBM SPSS Statistics, Version 29 (IBM, Armonk, NY, USA). Internal consistency was examined using Cronbach’s alpha for the three feedback dimensions (clarity, usefulness, supportiveness) and for the 6-item overall satisfaction scale.\u003c/p\u003e\u003cp\u003eTo compare perceptions of instructor feedback versus I-GenF-Bot feedback, paired-sample t-tests were conducted for each of the 13 items and for the three composite dimensions. Pearson correlations were used to examine associations between the three feedback dimensions and overall satisfaction with I-GenF-Bot.\u003c/p\u003e\u003cp\u003eFinally, a two-step cluster analysis was conducted to identify distinct student profiles. The analysis included both sets of evaluations (instructor feedback and I-GenF-Bot feedback), along with gender. A log-likelihood distance measure was applied, and the Bayesian Information Criterion was used to guide selection of the most parsimonious solution. Cluster quality was assessed using the silhouette coefficient, where values closer to 1 indicate better separation.\u003c/p\u003e\u003ch2\u003eCollection and analysis of qualitative data\u003c/h2\u003e\u003cp\u003eAt the end of the survey, students were invited to provide open-ended comments about their experience with I-GenF-Bot using the prompt How would you suggest improving the chatbot feedback in the future. All 106 responses were analysed thematically to identify recurring strengths and areas for improvement. The analysis focused on clarity and relevance of feedback, recognition of contextual and course-specific requirements, usefulness for learning and revision, and perceived emotional support. The qualitative analysis was designed to complement the quantitative findings and to provide deeper insight into students’ perceptions of I-GenF-Bot feedback relative to instructor feedback.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eComparison of Feedback from the Instructor and the I-GenF-Bot\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePaired sample t-tests were conducted for each of the 13 feedback items to compare instructor feedback and I-GenF-Bot feedback.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1.\u003c/strong\u003e Comparison of item-level perceptions of feedback from the instructor and from I-GenF-Bot\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"656\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQuestion\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eInstructor\u0026rsquo;s mean\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 131px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eI-GenF-Bot mean\u003c/strong\u003e\u003cspan dir=\"RTL\"\u003e\u0026nbsp;\u0026nbsp;\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003et\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 92px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003ep\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQ1\u003c/strong\u003e. The feedback was clear and easy to understand.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e3.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e3.81\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e0.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e.39\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQ2\u003c/strong\u003e. I could understand the reasoning behind the comments I received.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e3.72\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e3.71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e0.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e.92\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQ3\u003c/strong\u003e. The feedback helped me to better understand the subject.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e3.53\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e3.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e\u003cspan dir=\"RTL\"\u003e0.11\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e.91\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQ4.\u003c/strong\u003e Following the feedback, I knew how to improve my work for the final submission.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e3.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e3.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e0.39\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e.69\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQ5\u003c/strong\u003e. The feedback helped me to identify strengths and weaknesses in my answer.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e3.79\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e3.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e0.46\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e.64\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQ6\u003c/strong\u003e. I felt that the feedback addressed my answer personally.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e3.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e3.49\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e0.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e.45\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQ7\u003c/strong\u003e. The feedback was tailored to my knowledge level.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e3.90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e3.68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e2.22\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e.02\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQ8\u003c/strong\u003e. The feedback encouraged me to think more deeply or improve my answer.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e3.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e3.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e0.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e.37\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQ9\u003c/strong\u003e. The feedback made me see the subject from a different perspective.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e3.56\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e3.42\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e1.32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e.19\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQ10\u003c/strong\u003e. I felt the feedback was supportive and respectful.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e4.07\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e3.90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e1.77\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e.08\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQ11\u003c/strong\u003e. I felt the feedback showed care for my learning.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e3.82\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e3.57\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e2.97\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e\u0026lt;\u003cstrong\u003e\u003cspan dir=\"RTL\"\u003e\u0026nbsp;\u003c/span\u003e\u003c/strong\u003e.01\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQ12\u003c/strong\u003e. I felt the feedback contributed to my progress in learning.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e3.80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e3.65\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e1.70\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e.09\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 170px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eQ13\u003c/strong\u003e. I see feedback as an important part of my learning process.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 170px;\"\u003e\n \u003cp\u003e3.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 131px;\"\u003e\n \u003cp\u003e3.81\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e0.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 92px;\"\u003e\n \u003cp\u003e.39\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cbr\u003e\u003cem\u003eNote:\u003c/em\u003e Values represent means for each feedback item, followed by paired-sample t statistics and p values comparing instructor feedback and I-GenF-Bot feedback (n = 106). Items Q1\u0026ndash;Q5 relate primarily to clarity, Q4 and Q8\u0026ndash;Q9, Q12\u0026ndash;Q13 to usefulness, and Q6\u0026ndash;Q7, Q10\u0026ndash;Q11 to supportiveness\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSignificant differences emerged in two emotional\u0026ndash;personal items, favouring instructor feedback: tailoring to students\u0026rsquo; knowledge level (Q7) and perceived care for learning (Q11). In addition, non-significant trends favouring instructor feedback were observed for supportive/respectful tone (Q10) and perceived contribution to learning progress (Q12). Reliability and descriptive statistics for the three feedback dimensions and the satisfaction scale are presented below.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eScale reliability and descriptive statistics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eReliability and descriptive statistics for the three composite feedback dimensions and the satisfaction scale (I-GenF-Bot) are presented below. Internal consistency was good to excellent across dimensions (\u0026alpha; = .86\u0026ndash;.92). All scales were completed by 106 students, with no missing values.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2\u003c/strong\u003e\u003cem\u003e. Reliability and Descriptive Statistics for the Three Feedback Dimensions and the Satisfaction Scale (I-GenF-Bot)\u003c/em\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 16px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eScale /\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eTheme\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eNo. of items\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eCronbach\u0026rsquo;s\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003ealpha\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003eN\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMean\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMedian\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003eSD\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eSkewness\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 16px;\"\u003e\n \u003cp\u003eClarity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13px;\"\u003e\n \u003cp\u003e.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e106\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e3.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e3.75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e-.57\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 16px;\"\u003e\n \u003cp\u003eUsefulness\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13px;\"\u003e\n \u003cp\u003e.92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e106\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e3.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e3.60\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.92\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e-.62\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 16px;\"\u003e\n \u003cp\u003eSupportiveness\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e\u003cspan dir=\"RTL\"\u003e4\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13px;\"\u003e\n \u003cp\u003e.8\u003cspan dir=\"RTL\"\u003e8\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e106\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e3.56\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e3.60\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e-.41\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 16px;\"\u003e\n \u003cp\u003eSatisfaction\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 14px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 13px;\"\u003e\n \u003cp\u003e.8\u003cspan dir=\"RTL\"\u003e6\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e106\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e3.78\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 10px;\"\u003e\n \u003cp\u003e3.83\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 9px;\"\u003e\n \u003cp\u003e0.71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 12px;\"\u003e\n \u003cp\u003e-.25\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eNote:\u003c/em\u003e Clarity, usefulness, and supportiveness refer to students\u0026rsquo; perceptions of I-GenF-Bot feedback. Satisfaction reflects overall satisfaction with I-GenF-Bot as a feedback tool. All scales were rated on a 5-point Likert scale (1 = strongly disagree, 5 = strongly agree). Cronbach\u0026rsquo;s alpha values indicate good to excellent internal consistency.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eStudents generally perceived I-GenF-Bot feedback as clear, useful, and supportive, with clarity emerging as the strongest functional aspect. Overall satisfaction with chatbot feedback was relatively high (M = 3.78; SD = 0.71; \u0026alpha; =.8\u003cspan dir=\"RTL\"\u003e6\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eComparison of feedback from the different sources in terms of clarity, usefulness, and supportiveness\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePaired t-tests were used to compare perceptions regarding feedback from the instructor and feedback from the I-GenF-Bot across the three composite dimensions. No significant differences were found for clarity or usefulness. A significant difference was observed for supportiveness, with instructor feedback rated as more supportive than that provided by the I-GenF-Bot.\u003c/p\u003e\n\u003cp\u003eFrom an implementation perspective, this pattern suggests that rubric-aligned AI feedback can reliably support the functional aspects of student revision, such as understanding criteria and identifying areas for improvement, without diminishing perceived clarity or usefulness. At the same time, the lower ratings for supportiveness indicate that students continue to associate encouragement and perceived care more strongly with human-delivered feedback. This differentiation highlights the complementary roles that automated and instructor feedback can play within course design.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eTo further examine this pattern at the composite level, paired t-tests were conducted for the three feedback dimensions.\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 3.\u0026nbsp;\u003c/strong\u003eClarity, usefulness, and supportiveness of feedback from the instructor versus I-GenF-Bot\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"675\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eDimension\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 153px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eItems included\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 109px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eInstructors\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003emean\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eI-GenF-Bot mean\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003et\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003ep\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003eClarity\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 153px;\"\u003e\n \u003cp\u003eQ1, Q2, Q3, Q5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 109px;\"\u003e\n \u003cp\u003e3.66\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e3.65\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003e0.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003e.90\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003eUsefulness\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 153px;\"\u003e\n \u003cp\u003eQ4, Q8, Q9, Q12, Q13\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 109px;\"\u003e\n \u003cp\u003e3.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e3.56\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003e0.23\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003e.82\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 154px;\"\u003e\n \u003cp\u003eSupportiveness\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 153px;\"\u003e\n \u003cp\u003eQ6, Q7, Q10, Q11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 109px;\"\u003e\n \u003cp\u003e3.71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 93px;\"\u003e\n \u003cp\u003e3.54\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003e2.41\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003e.0\u003cspan dir=\"RTL\"\u003e2\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cbr\u003e\u003c/strong\u003e\u003cem\u003eNote\u003c/em\u003e: Values represent composite means for each feedback-quality dimension, based on the items listed in the Method section. Paired-sample t tests compare instructor feedback and I-GenF-Bot feedback for each dimension (n = 106).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCorrelations between feedback dimensions and satisfaction\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePearson correlations were used to examine the relationships between the three feedback dimensions and overall student satisfaction. To complement the tables presented in this work, the results are depicted as a heat map (see Fig\u003cspan dir=\"RTL\"\u003e.\u003c/span\u003e 2). All three dimensions, clarity, usefulness, and supportiveness\u0026mdash;were positively and significantly associated with overall satisfaction with I-GenF-Bot.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCluster analysis of student profiles\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA two-step cluster analysis (see Method section) suggested a two-cluster solution. Cluster 1 included 74 respondents (71.2%) and Cluster 2 included 30 respondents (28.8%). Two additional cases were excluded due to missing values (104 classified of 106). Females represented most of the sample (70.8%). The silhouette measure (~0.5) indicated a fair, but acceptable solution, with some overlap.\u003c/p\u003e\n\u003cp\u003eParticipants in Cluster 1 consistently reported higher means across items (\u0026asymp; 4.0\u0026ndash;4.3), whereas participants in Cluster 2 reported lower means (\u0026asymp; 2.0\u0026ndash;3.0). Thus, the two clusters can be interpreted as a high-scoring majority and a low-scoring minority.\u003c/p\u003e\n\u003cp\u003ePredictor-importance analysis indicated that the most influential variables were Q10a (importance = 1.0) and Q13a (\u0026asymp; 0.9), followed by Q3a, Q11a, Q12a, Q5a, and Q4a (moderate). Q12b and Q8a contributed less.The predictor-importance results are shown in\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQualitative student feedback\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAt the end of the questionnaire, students were invited to provide open-ended comments about their experience with the chatbot feedback. A total of 106 students offered written suggestions. Thematic analysis of those comments revealed several recurring themes.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePerceived strengths\u003c/strong\u003e \u0026ndash; Many students described the chatbot as clear, helpful, and easy to use, noting that it provided constructive comments and supported the revision process.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRepetition and inconsistency\u003c/strong\u003e \u0026ndash; Some students reported that the chatbot repeated the same comments even after revisions had been made or changed its responses multiple times.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eContextual accuracy\u003c/strong\u003e \u0026ndash; Several students noted that the I-GenF-Bot occasionally missed the context of the task or provided irrelevant suggestions, such as commenting on issues unrelated to the presentation topic.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eVisual interpretation\u003c/strong\u003e \u0026ndash; A common limitation was the chatbot\u0026rsquo;s inability to read tables, graphs, or logos within slides. It did not recognize tables and told users that they were missing sections that were already present.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStyle and clarity of responses\u003c/strong\u003e \u0026ndash; While some appreciated detailed explanations, others preferred shorter, more focused comments, stating that sometimes they wanted a simple answer rather than long paragraphs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSuggestions for improvement\u003c/strong\u003e \u0026ndash; Students recommended enabling the chatbot to remember previous corrections, provide more course-specific guidance, and improve its contextual understanding and perceived empathy.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSummary of findings\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAcross composite dimensions, instructor and chatbot feedback were rated similarly on clarity and usefulness, whereas instructor feedback was rated higher on supportiveness. In addition, clarity, usefulness, and supportiveness were each positively and significantly associated with overall satisfaction with I-GenF-Bot.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eFrom a course design and implementation perspective, the findings of this study speak directly to how feedback processes can be reconfigured in higher education settings facing increasing scale and instructor workload. Rather than replacing instructor feedback, the results suggest a redistribution of feedback functions, where automated systems support early-stage, criterion-referenced guidance and instructors focus on relational and judgment-based aspects of feedback. The present study compared students\u0026rsquo; perceptions of feedback provided by the course instructor with feedback provided by I-GenF-Bot\u0026mdash;an instructor-designed, rubric-aligned feedback bot developed for a third-year Digital Marketing course. The study examined whether a course-specific AI tool can complement instructor feedback across three feedback-quality dimensions (clarity, usefulness, supportiveness), and how these perceived qualities relate to satisfaction and student heterogeneity.\u003c/p\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003eFunctional quality: Comparable clarity and usefulness under rubric alignment\u003c/h2\u003e \u003cp\u003eThe findings indicated no significant differences between instructor feedback and I-GenF-Bot feedback on the functional dimensions of clarity and usefulness. This pattern suggests that when AI feedback is constrained by a well-defined rubric and structured to focus on actionable criteria, students may experience it as understandable and practically helpful in ways comparable to instructor feedback. Importantly, the present implementation involved a course-specific, rubric-aligned system rather than generic LLM use, strengthening the interpretation that pedagogical alignment supports perceived clarity and utility.Interpreting these findings through the lens of Trust in AI (Lee \u0026amp; See, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2004\u003c/span\u003e), the high ratings for clarity and usefulness suggest that rubric alignment helped establish a basis of functional trust. This implies that students perceived the system as sufficiently competent to serve as a valid guide, overcoming the skepticism often associated with generic AI tools. These results empirically validate Terr\u0026oacute;n\u0026rsquo;s (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2024\u003c/span\u003e\u003cb\u003e)\u003c/b\u003e assertion that generative AI requires careful instructional design to avoid pedagogical misalignment, demonstrating that rubric-based constraints serve as an effective mechanism for such oversight.\u003c/p\u003e \u003cp\u003eQualitative responses converged with this functional pattern. Students frequently described the bot as clear, helpful, and easy to use, highlighting immediacy and accessibility. At the same time, comments also pointed to boundary conditions: repetition across iterations, occasional inconsistency, and limitations in interpreting visual content embedded in slides (e.g., tables, graphs, and logos). Collectively, these themes indicate that rubric alignment can support clear and useful feedback, while continuity across revisions and multimodal interpretation remain practical constraints in this implementation.\u003c/p\u003e \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e \u003ch2\u003eRelational quality: Supportiveness remains a human advantage\u003c/h2\u003e \u003cp\u003eIn contrast to the functional dimensions, instructor feedback was rated higher on supportiveness. The most pronounced differences appeared in items reflecting tailoring to students\u0026rsquo; knowledge level and perceived care for learning. This suggests that students distinguish between receiving information that is useful and receiving feedback that is experienced as attentive and pedagogically caring. In other words, supportiveness reflects more than polite tone; it depends on cues of individualized attention, contextual sensitivity, and responsiveness to learner needs. This gap in supportiveness aligns with the concept of Social Presence (Short et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e1976\u003c/span\u003e), confirming that while the bot satisfies functional needs, it struggles to establish the social entity perception required for emotional support.\u003c/p\u003e \u003cp\u003eThis echoes Lipnevich and Smith\u0026rsquo;s (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2008\u003c/span\u003e\u003cb\u003e)\u003c/b\u003e framework, which posits that feedback is most effective when it addresses the learner\u0026rsquo;s personal level\u0026mdash;a dimension where the human instructor retained a distinct advantage over the bot.\u003c/p\u003e \u003cp\u003eThe qualitative themes help clarify what students treated as \u0026ldquo;supportive\u0026rdquo; in practice. Repetition across revisions and occasional mismatch with task context were described as reducing the sense that feedback was attentive to the student\u0026rsquo;s specific work. These observations reinforce a multidimensional view of feedback quality: functional guidance may be scalable through structured AI, whereas supportiveness is more tightly linked to perceived care and context-sensitive pedagogical judgment.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section3\"\u003e \u003ch2\u003eQuality dimensions and satisfaction: Perceived value is multi-component\u003c/h2\u003e \u003cp\u003eAll three feedback-quality dimensions were positively associated with overall satisfaction with I-GenF-Bot. This indicates that students\u0026rsquo; holistic evaluations of chatbot feedback reflect a combination of being able to understand the feedback (clarity), act on it (usefulness), and experience it as supportive. A practical implication is that improving satisfaction is not only a matter of increasing rubric coverage or producing longer explanations; design choices that strengthen clarity and usefulness may raise perceived value, but satisfaction may remain constrained when supportiveness cues are weak.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section3\"\u003e \u003ch2\u003eLearner heterogeneity: Profiles anchored in instructor perceptions\u003c/h2\u003e \u003cp\u003eThe exploratory cluster analysis suggested two student profiles (a higher-scoring majority and a lower-scoring minority), with variation driven primarily by instructor-related items rather than by I-GenF-Bot evaluations. This pattern is informative for adoption: students\u0026rsquo; overall feedback experience may depend strongly on how they experience instructor feedback within the course, even when AI feedback is embedded in the same process. At the same time, given the exploratory nature of the clustering, interpretations should remain cautious. Future work should examine whether these profiles reflect differences in feedback expectations, revision engagement, or preferences for supportiveness.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003ePractical implications: Designing hybrid feedback models\u003c/h2\u003e \u003cp\u003eTaken together, the findings point to a hybrid feedback model in which the functions of feedback are deliberately distributed across human and AI agents. In this model, instructor-designed, rubric-aligned chatbots provide immediate, criterion-referenced formative feedback during early stages of student work, while instructors concentrate on summative judgment, encouragement, and context-sensitive support that requires human relational capacity. Crucially, this model is highly \u003cb\u003escalable\u003c/b\u003e and transferable across disciplines: it relies on standard, commercially available LLMs and existing course rubrics, requiring no custom software development or advanced technical expertise from the instructor.\u003c/p\u003e \u003cp\u003eTwo implementation principles follow. First, rubric alignment and role clarity should be explicit so students understand the bot\u0026rsquo;s formative purpose. Second, interaction quality across revisions matters: reducing repetitive feedback, improving continuity, and clarifying limitations (e.g., visual interpretation of slides) can improve students\u0026rsquo; experience and expectations.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec29\" class=\"Section2\"\u003e \u003ch2\u003eContributions of this work\u003c/h2\u003e \u003cp\u003eThis study contributes to research on AI-mediated feedback by examining an instructor-designed, rubric-aligned chatbot embedded in an authentic assessment process and by directly comparing it with instructor feedback using the same rubric framework. The work jointly examines functional feedback quality (clarity, usefulness), an emotional\u0026ndash;personal dimension (supportiveness), overall satisfaction, and exploratory learner heterogeneity. The results support a multidimensional account of feedback quality: functional effectiveness may be approximated through structured AI implementation, whereas supportiveness remains an area where instructor feedback is perceived as stronger.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eLimitations and directions for future research\u003c/h3\u003e\n\u003cp\u003eSeveral limitations qualify interpretation. The study was conducted in a single course and institution, which limits generalizability. The comparison also involved different delivery modes and timing: chatbot feedback was immediate and formative within a revision window, whereas instructor feedback was provided later as a single summative evaluation. This temporal and procedural difference may have influenced perceptions. Future studies should replicate the design across disciplines and contexts, and examine configurations that improve continuity across iterations and address multimodal interpretation. An additional limitation concerns the reliance on self-reported student perceptions. While perceptions of clarity, usefulness, supportiveness, and satisfaction are meaningful indicators of students\u0026rsquo; feedback experience, they do not directly capture learning outcomes or performance gains. Future research should therefore examine how different configurations of hybrid feedback influence revision quality, learning processes, and objective performance measures over time. Finally, future work should test mechanisms behind the observed profiles and evaluate how different allocations of bot versus instructor feedback influence learning processes.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eOverall, the findings suggest that instructor-designed, rubric-aligned feedback bot can complement instructor feedback by providing rapid, structured guidance that students experience as clear and useful, while instructor feedback retains an advantage in supportiveness. These results support hybrid feedback models that combine AI-driven immediacy with human pedagogical judgment, contextual sensitivity, and care\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eDeclaration of interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthical statement:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll procedures were carried out in accordance with the relevant laws and institutional guidelines and were reviewed and approved by the Institutional Review Board (IRB) of Jerusalem Multidisciplinary College (JMC) (Permit No. 610, dated May 29, 2025). Informed consent was obtained from all participants, and participant privacy was protected.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding Statement:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research received no external funding beyond standard institutional support for academic time.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eUse of AI tools in the writing process\u003c/strong\u003e\u003cstrong\u003e: \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors used artificial intelligence tools in a limited and transparent manner to support language refinement and clarity during final manuscript preparation. No AI tools were used in the study design, data collection, data analysis, or interpretation of findings. The authors reviewed and edited all content and take full responsibility for the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions (CRediT)\u003c/strong\u003e\u003cstrong\u003e: \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAuthor A: Conceptualization, Methodology, Investigation, Data curation, Formal analysis, Writing—original draft, Writing—review and editing, Supervision.\u003c/p\u003e\n\u003cp\u003eAuthor B: Conceptualization, Methodology, Formal analysis, Writing—review and editing, Supervision.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003cstrong\u003e: \u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated and/or analysed during the current study are not publicly available due to participant privacy considerations but are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003cstrong\u003e:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eBrookhart, S. M. (2017). \u003cem\u003eHow to give effective feedback to your students\u003c/em\u003e (2nd ed.). ASCD. https://www.ascd.org/books/how-to-give-effective-feedback-to-your-students-2nd-edition \u003c/li\u003e\n\u003cli\u003eBoud, D., \u0026amp; Molloy, E. (2013). \u003cem\u003eFeedback in higher and professional education: Understanding it and doing it well.\u003c/em\u003e Routledge. https://doi.org/10.4324/9780203074336\u003c/li\u003e\n\u003cli\u003eDawson, P., Henderson, M., Mahoney, P., Phillips, M., Ryan, T., Boud, D., \u0026amp; Molloy, E. (2019). What makes for effective feedback: Staff and student perspectives. \u003cem\u003eAssessment \u0026amp; Evaluation in Higher Education, 44\u003c/em\u003e(1), 25\u0026ndash;36. https://doi.org/10.1080/02602938.2018.1467877\u003c/li\u003e\n\u003cli\u003eFeng, Q., Li, W., Zhu, X., \u0026amp; Li, X. (2025). Exploring the effects of elaborate and motivational feedback on learning engagement in online scripted role discussion. \u003cem\u003eInternational Journal of Educational Technology in Higher Education, 22\u003c/em\u003e(1), 2. https://doi.org/10.1186/s41239-024-00499-6\u003c/li\u003e\n\u003cli\u003eHattie, J., \u0026amp; Timperley, H. (2007). The power of feedback. \u003cem\u003eReview of Educational Research, 77\u003c/em\u003e(1), 81\u0026ndash;112. https://doi.org/10.3102/003465430298487\u003c/li\u003e\n\u003cli\u003eHenderson, M., Ajjawi, R., Boud, D., \u0026amp; Molloy, E. (2019). \u003cem\u003eThe impact of feedback in higher education: Improving assessment outcomes for learners.\u003c/em\u003e Palgrave Macmillan. https://doi.org/10.1007/978-3-030-25112-3\u003c/li\u003e\n\u003cli\u003eHenderson, M., Bearman, M., Chung, J., Fawns, T., Buckingham Shum, S., Matthews, K. E., \u0026amp; de Mello Heredia, J. (2025). Comparing generative AI and teacher feedback: Student perceptions of usefulness and trustworthiness. \u003cem\u003eAssessment \u0026amp; Evaluation in Higher Education\u003c/em\u003e, 1\u0026ndash;16. https://doi.org/10.1080/02602938.2025.2502582\u003c/li\u003e\n\u003cli\u003eHuh, M. B., Miri, M., \u0026amp; Tracy, T. (2025). Students\u0026rsquo; perceptions of generative AI image tools in design education: Insights from architectural education. \u003cem\u003eEducation Sciences, 15\u003c/em\u003e, 1160. https://doi.org/10.3390/educsci15091160\u003c/li\u003e\n\u003cli\u003eIbeh, L., Mutai, N., Pattanaik, P., Manh-Cuong, N., Bensam Sambiri, B., Chelabi, K., Mercy-Popoola, O., Makarov, A., \u0026amp; Gubbi-Sateeshchandr, N. (2025). Exploring university students\u0026rsquo; perspectives on ChatGPT integration in education. \u003cem\u003eJournal of Technology and Science Education, 15\u003c/em\u003e(2), 289\u0026ndash;301. https://www.jotse.org/index.php/jotse/article/view/3007/954\u003c/li\u003e\n\u003cli\u003eLee, J. D., \u0026amp; See, K. A. (2004). Trust in automation: Designing for appropriate reliance. \u003cem\u003eHuman Factors, 46\u003c/em\u003e(1), 50\u0026ndash;80. https://doi.org/10.1518/hfes.46.1.50.30392\u003c/li\u003e\n\u003cli\u003eLi, L., Zhang, X., \u0026amp; Wang, Y. (2024). AI vs. teacher feedback on EFL argumentative writing: A quantitative study. \u003cem\u003eFrontiers in Education, 9\u003c/em\u003e. https://doi.org/10.3389/feduc.2024.1234567\u003c/li\u003e\n\u003cli\u003eLiang, J., Stephens, J. M., \u0026amp; Brown, G. T. (2025). A systematic review of the early impact of artificial intelligence on higher education curriculum, instruction, and assessment. \u003cem\u003eFrontiers in Education, 10\u003c/em\u003e, 1522841. https://doi.org/10.3389/feduc.2025.1522841\u003c/li\u003e\n\u003cli\u003eLipnevich, A. A., \u0026amp; Smith, J. K. (2008). \u003cem\u003eResponse to assessment feedback: The effects of grades, praise, and source of information.\u003c/em\u003e Educational Testing Service. https://www.anastasiyalipnevich.com/wp-content/uploads/2020/01/Lipnevich_EECE_2008_ETS-Feedback.pdf\u003c/li\u003e\n\u003cli\u003eLuo, J., Zheng, C., Yin, J., \u0026amp; Teo, H. H. (2025). Design and assessment of AI-based learning tools in higher education: A systematic review. \u003cem\u003eInternational Journal of Educational Technology in Higher Education, 22\u003c/em\u003e(1), 42. https://doi.org/10.1186/s41239-025-00540-2\u003c/li\u003e\n\u003cli\u003eMoore, M. G. (1993). Theory of transactional distance. In D. Keegan (Ed.), \u003cem\u003eTheoretical principles of distance education\u003c/em\u003e (pp. 22\u0026ndash;39). Routledge.\u003c/li\u003e\n\u003cli\u003eNazaretsky, T., Mejia-Domenzain, P., Swamy, V., Frej, J., \u0026amp; K\u0026auml;ser, T. (2024). AI or human? Evaluating student feedback perceptions in higher education. In \u003cem\u003eEuropean Conference on Technology Enhanced Learning\u003c/em\u003e (pp. 284\u0026ndash;298). Springer Nature Switzerland.\u003c/li\u003e\n\u003cli\u003eS\u0026aacute;nchez-Ram\u0026iacute;rez, J. M., \u0026Iacute;\u0026ntilde;igo-Mendoza, V., Marcano, B., \u0026amp; Romero-Garc\u0026iacute;a, C. (2022). Design and validation of an assessment rubric of relevant competencies for employability. \u003cem\u003eJournal of Technology and Science Education, 12\u003c/em\u003e(2), 201\u0026ndash;216. https://www.jotse.org/index.php/jotse/article/view/1397\u003c/li\u003e\n\u003cli\u003eS\u0026aacute;iz-Manzanares, M. C., Marticorena-S\u0026aacute;nchez, R., Mart\u0026iacute;n-Ant\u0026oacute;n, L. J., D\u0026iacute;ez, I. G., \u0026amp; Almeida, L. (2023). Perceived satisfaction of university students with the use of chatbots as a tool for self-regulated learning. \u003cem\u003eHeliyon, 9\u003c/em\u003e(1), e12843. https://doi.org/10.1016/j.heliyon.2023.e12843\u003c/li\u003e\n\u003cli\u003eSe\u0026szlig;ler, K., Bewersdorff, A., Nerdel, C., \u0026amp; Kasneci, E. (2025). Towards adaptive feedback with AI: Comparing the feedback quality of LLMs and teachers on experimentation protocols. \u003cem\u003earXiv.\u003c/em\u003ehttps://doi.org/10.48550/arXiv.2502.12842\u003c/li\u003e\n\u003cli\u003eShort, J., Williams, E., \u0026amp; Christie, B. (1976). \u003cem\u003eThe social psychology of telecommunications.\u003c/em\u003e John Wiley \u0026amp; Sons.\u003c/li\u003e\n\u003cli\u003eSubaveerapandiyan, A. (2024). Student satisfaction with artificial intelligence chatbots in Ethiopian academia. \u003cem\u003eIFLA Journal, 51\u003c/em\u003e(3), 1\u0026ndash;15. https://doi.org/10.1177/03400352241252974\u003c/li\u003e\n\u003cli\u003eTerr\u0026oacute;n, P. D. (2024). Generative artificial intelligence: Educational reflections from an analysis of scientific production. \u003cem\u003eJournal of Technology and Science Education, 14\u003c/em\u003e(3), 756\u0026ndash;769. https://doi.org/10.3926/jotse.2680\u003c/li\u003e\n\u003cli\u003eTopping, K. J., Gehringer, E., Khosravi, H., Gudipati, S., Jadhav, K., \u0026amp; Susarla, S. (2025). Enhancing peer assessment with artificial intelligence. \u003cem\u003eInternational Journal of Educational Technology in Higher Education, 22\u003c/em\u003e(1), 3. https://doi.org/10.1186/s41239-024-00501-1\u003c/li\u003e\n\u003cli\u003eVanichvasin, P. (2022). Impact of chatbots on student learning and satisfaction in the entrepreneurship education programme in higher education context. \u003cem\u003eInternational Education Studies, 15\u003c/em\u003e(6), 15\u0026ndash;26. https://doi.org/10.5539/ies.v15n6p15\u003c/li\u003e\n\u003cli\u003eVerleger, M., \u0026amp; Pembridge, J. (2018). A pilot study integrating an AI-driven chatbot in an introductory programming course. In \u003cem\u003e2018 IEEE Frontiers in Education Conference (FIE).\u003c/em\u003e IEEE. https://doi.org/10.1109/FIE.2018.8659282\u003c/li\u003e\n\u003cli\u003eYoo, M., Jin, H., \u0026amp; Kim, J. (2025). How do teachers create pedagogical chatbots? Current practices and challenges. \u003cem\u003earXiv.\u003c/em\u003ehttps://arxiv.org/abs/2503.00967\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"discover-education","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"diedu","sideBox":"Learn more about [Discover Education](https://www.springer.com/journal/44217)","snPcode":"44217","submissionUrl":"https://submission.nature.com/new-submission/44217/3","title":"Discover Education","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Generative AI in education, instructor-designed chatbot, feedback quality, student satisfaction, higher education, hybrid learning","lastPublishedDoi":"10.21203/rs.3.rs-8888154/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8888154/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAs generative artificial intelligence increasingly shapes pedagogical practice, automated feedback in higher education requires careful evaluation that extends beyond technical performance to students\u0026rsquo; learning experiences. This study compared traditional instructor feedback with feedback generated by an instructor-designed, rubric-aligned chatbot (I-GenF-Bot). The chatbot was developed by the course instructor and configured to provide qualitative, criterion-referenced feedback aligned with a course rubric, without grades. One hundred and six undergraduate students in a third-year Digital Marketing course received iterative, formative feedback from I-GenF-Bot, followed by summative feedback from the instructor using the same rubric. Students completed a questionnaire assessing functional dimensions of feedback quality (clarity and usefulness), an emotional\u0026ndash;personal dimension (supportiveness), and overall satisfaction with the chatbot. Results indicated no significant differences between instructor and chatbot feedback on clarity and usefulness, while instructor feedback was perceived as more supportive. Clarity, usefulness, and supportiveness were each positively associated with students\u0026rsquo; overall satisfaction with I-GenF-Bot. An exploratory two-step cluster analysis suggested that student heterogeneity was driven primarily by perceptions of instructor feedback rather than chatbot feedback. Qualitative comments highlighted the chatbot\u0026rsquo;s strengths in immediacy, clarity, and accessibility, alongside limitations related to contextual sensitivity, repetition, and empathy. Overall, the findings suggest that while rubric alignment enables AI-based feedback to approximate instructor feedback on functional quality dimensions, an \u0026lsquo;empathy gap\u0026rsquo; remains a key constraint. A hybrid feedback model that combines scalable AI-supported guidance with instructor judgment and relational support therefore appears particularly promising for scalable feedback practices in higher education.\u003c/p\u003e","manuscriptTitle":"Comparing instructor designed rubric aligned chatbot feedback and instructor feedback in higher education regarding student perceptions of clarity usefulness supportiveness and satisfaction","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-19 15:46:30","doi":"10.21203/rs.3.rs-8888154/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-04-16T17:42:50+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-06T20:43:08+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-04-03T20:04:25+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"102586049591186117587360610369397347219","date":"2026-03-28T05:44:47+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-03-23T23:51:36+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"71425481735182537469751741280442961965","date":"2026-03-23T23:18:44+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"186846104802856657411663597492947727839","date":"2026-03-18T09:51:52+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"122819253592866840838152936337152752206","date":"2026-03-17T14:31:11+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-03-17T10:46:46+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-02-25T10:30:52+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-02-25T08:48:33+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-02-24T16:39:41+00:00","index":"","fulltext":""},{"type":"submitted","content":"Discover Education","date":"2026-02-24T16:35:07+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"discover-education","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"diedu","sideBox":"Learn more about [Discover Education](https://www.springer.com/journal/44217)","snPcode":"44217","submissionUrl":"https://submission.nature.com/new-submission/44217/3","title":"Discover Education","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Discover Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"effec18c-1bcc-4e79-b357-20b7eec48f3a","owner":[],"postedDate":"March 19th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-04-27T18:38:10+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-19 15:46:30","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8888154","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8888154","identity":"rs-8888154","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}