Instructional Design Models

Overview. Generative AI is shifting instructional design from a process where one designer authors one path for a course to one where the designer authors a systemthat adapts a path for each learner. The most useful frame here is not “AI replaces designers” but “AI takes over the parts of design that already wanted to be variable” (examples, scaffolds, feedback wording, difficulty, modality). The designer's job moves toward setting the constraints and guardrails inside which AI can vary the experience: learning objectives that survive variation, assessment criteria that survive variation, accuracy thresholds, source authority, and tone.

Key learnings from this unit.

1. AI literacy in adult learners is not a single skill. It is the layered combination of (a) safety habits that predate AI, (b) a mental model of what AI actually is, and (c) a workflow that includes verification. A course that teaches AI without the first and third layer just trains people to ship hallucinations with confidence.
2. The bottleneck for using AI in instructional design is not generation. It is evaluation. Knowing whether the AI's output is good. This is the same skill the learner needs and the same skill the designer needs, which is why this minicourse devotes a full module (Module 2) to predictability and benchmark literacy before anyone is asked to use AI for real work.
3. Industry benchmarks (MMLU, HumanEval, GPQA, Chatbot Arena) are an underused literacy resource in adult education. Plain-language benchmark fluency gives a non-technical adult a defensible answer to “is this AI any good at the thing I am about to ask it to do.”

Additional resources.

• Artificial Intelligence and the Future of Teaching and Learning: Insights and Recommendations . U.S. Department of Education, Office of Educational Technology (May 2023). The federal framing of AI in education, including a seven-recommendation policy stance. The most-cited starting point for an instructional-design conversation about AI today. Read the report (PDF).
• Guidance for Generative AI in Education and Research . UNESCO (September 2023). International framing, including a human-centered approach, age thresholds, and the institutional capacities required to adopt AI well. Pairs with the U.S. ED report because it foregrounds equity rather than productivity. Read the guidance.

Implications for instructional design. A designer working with AI in 2026 carries three responsibilities that did not exist five years ago. First, the designer must specify the evaluation criteria for AI-generated content because the cost of generation has collapsed and the cost of checking has not. Second, the designer must teach the learner to evaluate AI output; skipping this produces graduates who cannot tell when their tools are wrong. Third, the designer must design for both the AI-assisted and the AI-unassisted learner in the same course, because access is uneven and assuming everyone has the same tools recreates the digital divide. Getting Ready for AI takes the second responsibility as its primary mission and the third as a constraint. Every exercise has a no-AI version so the protective and conceptual habits land regardless of whether the learner uses AI at all.

Target audience.Working professionals on the non-technical side of the workforce who have been asked, by their employer or by their own ambition, to use AI tools well in their existing job, but who do not have an engineering, IT, or security background to draw from. The pathway is built to take them from “curious but cautious” to “competent and confident” across roughly fifteen hours of focused content, with a self-attestable certificate of completion that maps to 10 to 12 CPE credits across ISC2, CompTIA, ISACA, PMI, and SHRM. The LDTC 615 capstone build within this pathway is Module 2, Sensitive Data and AI-Safe Behaviors, which is what the IDD is scoped to from Unit 1 forward.

Learner profile.

Demographics.Adults aged 30–60, employed full-time or running a small business, in the United States, Canada, the United Kingdom, Australia, or other primarily English-speaking workforces. Gender, race, and geography are broadly representative of the white-collar / knowledge-worker labor force; the audience is not concentrated in one industry but skews toward operations, HR, marketing, finance, project management, healthcare administration, and education administration. Many are holders of a four-year degree; some are holders of graduate degrees; a significant minority hold one or more active professional certifications.

Background and prior knowledge. The learner has used a laptop and a smartphone professionally for at least a decade. They are comfortable with email, document editing, video calls, basic file sharing, and the common tools of their own profession (Excel, Salesforce, Jira, EHRs, the LMS at their school, etc.). They have heard of ChatGPT and Claude, may have tried one of them once or twice, and have read at least one article or seen one news segment about AI. They do not know how a large-language model works under the hood, and the pathway does not require them to. They do not know what full-disk encryption is, which is the gap that opens Module 1.

Skills and dispositions. Skill-wise, they can read for forty-five minutes without losing focus, follow multi-step instructions, write a one-page memo, and make a defensible decision with incomplete information , because they do all four every workday. Disposition-wise, they are cautious about AI for one of two reasons: either they are worried about getting it wrong in front of stakeholders (the operations / HR / finance pattern), or they are worried about getting it wrong in front of a regulator (the healthcare / education / compliance pattern). The pathway is built to convert that caution into informed practice, not to argue it away.

Reflection.

Range of technological proficiency, and how I'll keep the content engaging and accessible across digital literacy levels.Within the target audience, technical proficiency varies by roughly two standard deviations. One end of the curve is the HR generalist whose entire tool stack is browser-based and who has never opened a terminal in her life. The other end is the operations manager who has built complex spreadsheet models with macros and uses Zapier without help. The pathway needs to land for both. My core mechanism for that is the design decisions spec's plain-English voice rule (Sweller, 1988, on intrinsic cognitive load): every technical term is introduced with an everyday analogy first, then named. The lock icon, then TLS (design-decisions spec, 2026). Concept density is capped at three new ideas per concept lesson and one per exercise lesson. UDL multiple means (CAST, 2018) is preserved by pairing every prose lesson with a sibling exercise lesson and an upcoming module overview video so the same idea lands in three modalities. The capstone is verifiable by demonstration, not by recall, which lets a learner of any digital literacy show competence by completing one real artifact in their own work life. Built-in differentiation: every lesson opens with a one-screen placement check; learners who already know the content can skip ahead to the exercise.

Strategies and modifications for cultural relevance and accessibility.Cultural relevance starts in the scenario design. Module 1's data-states lesson uses a generic email scenario instead of an industry-specific one, because the audience spans many industries and workplace cultures; later modules let the learner choose a context (workplace, small business, or personal). The language is plain English, Flesch–Kincaid Grade 8–10, avoiding U.S.-only idioms. Accessibility is built to WCAG 2.2 AA per the design decisions doc: 16.8:1 ink-on-background contrast, 17/18px base font, keyboard-navigable focus rings, captions plus open transcripts on every video, and decorative-versus- functional alt text per the WAI-ARIA Authoring Practices (W3C, 2024). The single optional gamification visible on the pathway hub is a charcoal progress bar and a capstone badge state. No streaks, no leaderboards, no time pressure. Following Hamari and Koivisto (2014) on gamification effects across populations and the Self-Determination Theory frame (Deci & Ryan, 1985), which is especially load- bearing for adult learners who do not respond well to extrinsic-reward designs.

Mechanisms to collect and analyze feedback, and how I will use it for iterative improvement.The course ships with an xAPI minimum-viable statement set per the design decisions doc, recording lesson opens, objective views, exercise attempts and completions, mastery-check outcomes, and total time on lesson. The xAPI data lets me see where learners drop, retake, or score low , three signals that point to a content problem before any survey does. On top of xAPI, every mastery checkpoint includes an optional two-question post-survey (confidence in applying the objective, one thing the lesson could have done better), with responses stored against the lesson slug. Quarterly, I review the aggregate (drop rate, mastery-score distribution, time-on-lesson distribution, post-survey text) and promote one improvement per module per quarter to the next ship. This loop is documented in the design decisions doc and is the same iterative-design rhythm (Allen's SAM) that I commit to in the ADDIE section above. Improvements that the data does not justify do not ship, because gut-feel revisions are how instructional designers add complexity without adding outcomes.

References.

• Allen, M. W. (2006). Creating successful e-learning. Pfeiffer.
• Allen, M. W. (2012). Leaving ADDIE for SAM. ATD Press.
• Anderson, L. W., & Krathwohl, D. R. (Eds.). (2001). A taxonomy for learning, teaching, and assessing: A revision of Bloom's taxonomy of educational objectives. Longman.
• Branch, R. M. (2009). Instructional design: The ADDIE approach. Springer.
• Branson, R. K., Rayner, G. T., Cox, J. L., Furman, J. P., King, F. J., & Hannum, W. H. (1975). Interservice procedures for instructional systems development (TRADOC Pam 350-30, NAVEDTRA 106A). U.S. Army Training and Doctrine Command.
• CAST. (2018). Universal Design for Learning guidelines (version 2.2). https://udlguidelines.cast.org
• Deci, E. L., & Ryan, R. M. (1985). Intrinsic motivation and self-determination in human behavior. Plenum.
• Hamari, J., & Koivisto, J. (2014). Measuring flow in gamification: Dispositional flow scale-2. Computers in Human Behavior, 40, 133–143.
• Kirkpatrick, D. L. (1996). Great ideas revisited: Revisiting Kirkpatrick's four-level model. Training & Development, 50(1), 54–57. (Original work published 1959.)
• Sweller, J. (1988). Cognitive load during problem solving: Effects on learning. Cognitive Science, 12(2), 257–285.
• W3C. (2024). WAI-ARIA Authoring Practices 1.2. https://www.w3.org/WAI/ARIA/apg/
• Wiggins, G., & McTighe, J. (2005). Understanding by design (2nd ed.). ASCD.