To be successful, online graduate learners frequently require timely, personalized scaffolding and clear, actionable feedback; meanwhile, instructors must provide such support at scale without diluting rigor or equity. Recent reviews of artificial intelligence in education (AIED) indicate that generative and conversational systems are reshaping support ecosystems around adaptive tutoring, intelligent assessment, and profiling/prediction, with growing focus on design quality and theoretical grounding (Wang et al., 2024). Meta-analytic evidence on intelligent tutoring systems also suggests positive average effects on achievement, while cautioning that impact depends on comparison condition, measure type, and implementation fidelity, underscoring the need for principled integration and instructor oversight (Kulik & Fletcher, 2016). In practice, GenAI can deliver on-demand coaching and formative guidance in asynchronous contexts, but it also raises concerns about over-scaffolding, superficial or biased feedback, and the potential erosion of collaborative learning if systems are not deliberately designed (Hanshaw et al., 2024).
We present the LDT program as a program-level case of GenAI integration organized around three applications: AI Mentors, activity tailoring, and AI-driven planning tools. We outline how these uses personalize learning for online students while maintaining rigor and ethical practice.
Research indicates that, when implemented as conversational course assistants, AI Mentors afford on-demand, context-aware guidance; adaptive, individualized feedback; metacognitive prompting and plan critique; and social presence that mitigates isolation in asynchronous learning, thereby supporting self-regulated learning and professional communication (Hanshaw et al., 2024; Evmenova et al., 2024).
In ASU’s LDT program, AI Mentors are LMS-embedded, course-specific generative agents intentionally designed to augment human instruction and support. They engage learners in real-time to navigate academic challenges and to cultivate professional competencies, offering contextually relevant advice that helps students interpret expectations, refine plans, and sustain momentum precisely when asynchronous learners most need support.
Interactions are dialogic and iterative: the AI Mentor prompts goal clarification, critiques plans, and offers reflection nudges (what to keep, what to change, why). It translates criteria into plain-language checks and exemplars and scaffolds pre–post reflection to strengthen argumentation and civility. These uses align with research showing that conversational agents increase engagement through structured inquiry and context-aware questioning (Ilieva et al., 2023), and that GenAI can act as a “creative engine” by providing alternative explanations and perspectives (Henriksen et al., 2025). The Mentor delivers just-in-time scaffolding tied to outcomes and goals, providing specific guidance and encouragement. This is consistent with evidence that GenAI can individualize feedback, reduce isolation in asynchronous settings, and bolster self-regulated learning when used intentionally (Wang et al., 2024).
In LDT courses, student AI use follows a green–yellow–red framework. Green uses include ideation, concept clarification, formative feedback, organizing/formatting drafts, and learning to read/write code. Yellow uses require judgment and documentation; students verify facts and note limits. Red uses are prohibited: generating graded work, submitting unsupervised AI text as one’s own, writing peer replies, or inventing citations. For privacy, Mentors run within ASU’s managed, FERPA-aligned OpenAI instance; students are encouraged to avoid including personally identifiable or sensitive information in prompts. Because outputs may be biased or inaccurate, students are also encouraged to verify claims with primary sources, seek corroboration, and note uncertainties. Disclosure is required whenever AI contributes meaningfully. Typical permitted queries include clarifying theories or methods, brainstorming design scenarios, drafting checklists or timelines, and requesting formative feedback on outlines or code. These practices address risks of superficial/biased feedback and over-reliance, positioning the Mentor within a broader support ecosystem.
GenAI can clarify rubric criteria in plain language, produce parallel task variants and exemplars, adapt language and modality for multilingual and accessibility needs, and deliver specific, context-aware formative feedback, thereby enabling multiple means of engagement while preserving outcome alignment (Evmenova et al., 2024; Wang et al., 2024).
Another key application of GenAI in the LDT program is its role in mediating how rubrics are interpreted and applied, as well as in adapting course activities, to better align with diverse learner needs. By leveraging GenAI, educators can create multiple versions of an assignment reflecting diverse instructional design contexts, yet still assess all variations consistently with a single rubric. Additionally, GenAI also assists instructors in drafting targeted formative feedback.
Three checks govern use. First, an equity/language review ensures rubric phrasing avoids bias, deficit framing, or culturally narrow exemplars (Bae & Bozkurt, 2024; Su & Yang, 2023). Second, an outcome-alignment check confirms that any activity variant assesses the same learning targets (Mathew & Stefaniak, 2024; Kumar et al., 2024). Third, exemplar anchoring requires that AI-generated exemplars be human-reviewed and paired with transparent commentary, so students can see why their work meets (or falls short of) a particular level. Together, these practices personalize assessment while preserving rigor and fairness.
GenAI-enabled planning assistants can break complex tasks into milestones and checklists and scaffold self-regulated learning by strengthening planning, time management, and self-monitoring (Lee & Moore, 2024; Ng et al., 2024).
In ASU’s LDT program, the LDT Program Assignment Organizer Prompt Generator helps learners translate major assignments into a concrete daily plan. Students select productive hours, due date, and assignment text; with one click, they receive a tailored prompt to paste into ChatGPT, which returns a custom, sequenced schedule. The result is a lightweight, student-specific plan that keeps rubric expectations visible while sequencing work into manageable units.
Beyond tooling, the LDT program embeds planning into course task structures. Consistent with research that planning and organizing can be intentionally designed into coursework (Navayuth & Yurayat, 2022), LDT courses routinely scaffold smaller assignments that culminate in a capstone artifact. For example, in LDT 508: Design of Accessible and Inclusive Digital Learning, students develop an accessible educational website through a sequenced series of tasks. This sequenced approach promotes task initiation and completion, sustaining momentum across the term, while the Organizer tool provides day-to-day pacing and visibility of progress.
Across the three applications, our design priorities center on access, inclusion, and responsible use. We emphasize plain-language, multiple representations, and flexible evidencing so learners can demonstrate competence through varied modalities. GenAI can rapidly produce alternative explanations, exemplars, and task variants, supporting UDL (Evmenova et al., 2024). For multilingual and international cohorts, GenAI enables language-level adjustments, idiomatic refinements, and culturally responsive examples, expanding participation without lowering standards (Zlotnikova & Hlomani, 2024). In addition, conversational systems can scaffold progressive challenge and reflective practice, which is consequential for learners developing self-regulated learning skills (Ilieva et al., 2023; Ng et al., 2024).
As a program, we share a commitment to responsible use practices. Routine bias and language checks are applied to rubric phrasing, exemplars, and feedback to reduce deficit framing and cultural narrowness (Bae & Bozkurt, 2024; Su & Yang, 2023). In the design of student-facing AI tools, we use ASU’s managed OpenAI instance, which provides a walled garden to protect student information. Additionally, our program requires transparent disclosure when AI contributes meaningfully to student work, consistent with recommendations to maintain academic integrity and clarity about AI’s role (Lee & Moore, 2024; Mathew & Stefaniak, 2024).
Finally, we position GenAI to augment instructor presence and feedback. AI systems can handle routine clarifications and formative nudges, thereby freeing faculty time for higher-value guidance while maintaining the human judgment component central to assessment and mentoring (Hanshaw et al., 2024; Kumar et al., 2024). This stance aligns with field syntheses that highlight adaptive/personalized tutoring and chatbots as productive when they are anchored in pedagogy and oversight, not as stand-alone solutions (Wang et al., 2024). In sum, equity-minded design, transparent governance, and deliberate instructor orchestration make GenAI a lever for inclusion rather than a source of new inequities.
Taken together, the LDT case illustrates how a coherent, program-level approach to GenAI can expand timely scaffolding for online graduate learners while preserving rigor, inclusion, and academic integrity. By integrating the use of AI Mentors, rubric and activity tailoring, and AI-enabled planning tools, ASU’s LDT program has leveraged GenAI to personalize learning, differentiate feedback, and scaffold self-regulated work habits. This model’s value lies not in automation, but in augmenting instructor presence and strengthening executive-function supports for students. As institutions adapt these practices, systematic documentation and faculty development will be essential. Future work should pair use analytics with learner-reported experiences to evaluate differential impacts across disciplines and contexts, refining guardrails and design patterns that sustain equity-minded personalization at scale.
EdTech Archives