Blog

Category: Practitioner Case Studies | Reading time: 6 min&nbsp; Why Stories Are the Most Powerful Technology We Have Before I became a researcher, I was a student in Zambia watching my teachers tell stories.Not to fill time. To transfer something that a textbook could not carry. The best teachers I had did not start with the lesson. They started with a person. A moment. Something real that had weight and smell and consequence. And then the lesson lived inside that story forever. I could not forget it because the story would not let me.That is not an accident. That is design.Cognitive science has a name for it. Narrative transportation. When a person is genuinely inside a story — not reading about it but living it momentarily — their brain processes it differently than it processes information. It encodes it more deeply. It connects it to emotion. It makes it retrievable years later in ways that bullet points and frameworks simply do not.Every learning system I have ever studied, every database I have ever built, every AI framework I have ever designed — the ones that actually worked started with a story. A specific human being in a specific situation with a specific problem that needed solving. Not a use case. A person.&nbsp; The Teacher Who Got Too Close I spent seven years as an ICT and Guidance Teacher in Zambia's Ministry of General Education.And yes — sometimes that got me in trouble.Why are you bringing your students this close? People would ask. Colleagues. Administrators. The system itself, in its quiet bureaucratic way, reminding me that there was a correct distance between a teacher and a learner. A professional distance. A distance that protected the institution from liability and the teacher from entanglement and the student from the dangerous illusion that someone in a position of authority actually cared about them specifically.I never managed that distance very well.Every class I taught — it was beyond just being a class teacher. I was their friend. I wanted to see them. Not their grades. Not their attendance records. Not their performance against the national curriculum benchmarks. Them. Where they were going and whether they believed they could get there and what was sitting on their chest at 14 years old in a system that was not always designed to see them clearly.I wanted to know what made them lean in and what made them shut down. I wanted to know which student needed to be challenged hard and which one needed to be believed in first before they could be challenged at all. I wanted to know the whole person — because I had figured out early that the whole person was the only unit of analysis that actually mattered in a classroom.I guess I am doing the same thing for my kids.&nbsp; And Before You Come For the Technical Argument You can argue about schemas. You can argue about the relationships between data tables. You can argue about APIs and whether what I built is production-ready and whether Lovable is a real development tool or a shortcut and whether someone with my background has the right to call herself a builder.I know those arguments. I have sat in rooms where they were made. I know database architecture. I know how tables connect and what breaks when the logic between them is wrong. I have built systems that handled federal compliance requirements and served thousands of people and passed audits that would have exposed any structural weakness in the data model. I know what a poorly designed schema costs you six months into a build. I know the difference between a system that runs and a system that holds.I implemented this. Not theoretically. Not conceptually. Actually.The tool gave me speed. The knowledge was already there — built over years of systems work, doctoral research, and seven years of watching what happens when a learning experience is designed around the real human being in the room versus the imagined average one.So yes. You can argue about the technical specifications.Just know that the person you are arguing with has already thought about them — and then went ahead and built it anyway for reasons that have nothing to do with impressing you.&nbsp; The Letter &nbsp; Wanipa. Koko.This one is for you. Not for LinkedIn. Not for the algorithm. Not for the doctoral committee or the conference panel or the people who need to see credentials before they decide if I am worth listening to.For you.&nbsp; I have been away for almost four years.I want you to know that I have counted every single one of those days. Not dramatically. Not in a way that made me unable to function — because I could not afford to be unable to function. But quietly. The way you count things that matter too much to say out loud.9,893 miles. I looked it up the first time and then I could not stop looking it up. As if the number would change. As if checking it again would make it smaller.It never got smaller.&nbsp; I looked for something I could give you.Not a present. Not a video call where we both pretend the connection is good enough when it is not. Something real. Something that said — I know you. I have been paying attention across every timezone and every missed bedtime and every photograph that arrived on my phone at 2 in the morning Reno time which was morning in Lusaka and you were already up and already becoming more of who you are without me in the room.I know that Koko needs to move before she can sit still. I know that Wanipa needs to see the finish line before she will start running. I know the songs you like and the subjects that make you lean in and the exact moment when you have stopped learning and started performing patience for the adult in the room.I know you.No app knew you. I searched. I found apps built for children whose lives look nothing like yours. Children whose curriculum comes from somewhere else. Whose cultural references are not yours. Whose languages do not include the ones you move between like water.Every platform was built for someone else's child.And I am not interested in someone else's child. I am interested in you.&nbsp; So I built something.Saturday night. Your grandmother's voice somewhere in the background of my memory telling me that when you cannot find the door you build one. Ingrid Michaelson on repeat because her voice does something to the inside of my chest that makes hard things feel survivable.Everybody wants to love. Everybody wants to be loved.I built it the way I know how to build things — with the architecture first, the logic of how pieces connect, the decisions about what data matters and why and how it should flow. I used tools that closed the gap between knowing and building. And I put you inside every decision.Koko — the app knows you need a break before you know you need one. It will not wait for you to check out. It will shift before that.Wanipa — the app will show you exactly how far you have come. Every time. Because I know that is what keeps you going and I have been keeping that knowledge for years waiting for somewhere to put it.&nbsp; I am not going to tell you it is perfect.I am going to tell you it is yours.Built by your mother, from the other side of the world, on a Saturday night, with a song on repeat, because the distance is real and the love is realer and technology gave me a way to put the second one inside the first one and hand it to you.Every learning experience in that app is there because I know something specific and true about who you are. Not who I hope you will be. Not who the curriculum says you should be at your age. Who you actually are right now, today, in this specific season of your specific lives.That is all instructional design ever was supposed to be.I just needed you to exist for me to finally do it right.&nbsp; I am coming home.Until I do — open the app.Your mother built it.&nbsp; For Wanipa and Koko. Everything else is just work.&nbsp; Priscillar McMillan is a doctoral research assistant in Information Technology in Education at the University of Nevada, Reno, former UNICEF consultant, and founder of Kowa Agency. She builds learning systems, advises organizations on ethical AI implementation, and writes weekly on the decisions that will shape how the world learns. <a title="ZAMBIAN HOME SCHOOL APP" href="https://zambia-learn-home.lovable.app" target="_blank" rel="noopener" data-mce-href="https://zambia-learn-home.lovable.app">CHECK IT OUT HERE</a>

Problem-forward, Global South angle Why No Homeschooling App Worked for My Zambian Children — So I Built One With AI

African American professor using touchpad while assisting her student on a class at the university.

<h3>Category: The Educator's Dilemma &nbsp; | &nbsp; Reading time: 6 min The Framing That Poisons the Conversation</h3>There is a story that circulates in higher education administration about AI adoption, and it goes roughly like this: the technology is ready, the students want it, the evidence supports it, and the obstacle is faculty resistance — professors who are too attached to traditional methods, too skeptical of innovation, too protective of their disciplinary authority to embrace what is coming. The solution, in this story, is better communication, more professional development, and patient culture change.I teach in a university. I study AI adoption in higher education. And I want to offer a different framing, one grounded in what the research on faculty AI adoption actually shows rather than what makes for a clean institutional narrative.Faculty are not, as a population, irrational resisters of useful technology. What the evidence shows is something more specific and more uncomfortable: faculty are making rational cost-benefit assessments under conditions of genuine uncertainty, inadequate support, and misaligned institutional incentives — and arriving at conclusions that look like resistance from a distance but look quite different up close.<h3>What Faculty Are Actually Weighing</h3>When a faculty member declines to integrate an AI feedback tool into their course, they are typically weighing a specific set of real considerations. Will this tool produce reliable, accurate feedback on the kinds of work my students produce? Will it disadvantage students who lack the devices, connectivity, or digital fluency to use it effectively? Will it create accessibility problems for students with disabilities? Will the time I spend learning the system, configuring it for my course, and troubleshooting student issues be offset by any actual improvement in learning? And underneath all of these: if this pilot fails or produces unintended consequences, who bears the professional cost?These are not irrational questions. They are exactly the questions a responsible educator should ask before deploying an untested technology into a learning environment where real students bear the consequences of getting it wrong. The research on faculty AI adoption consistently finds that skepticism is highest not among the least competent or least innovative faculty but among faculty with the deepest pedagogical knowledge — the ones who have thought most carefully about how learning works and are therefore most capable of identifying the ways a new tool might disrupt it.A 2024 EDUCAUSE survey found that faculty concerns about AI center heavily on academic integrity, equity of student access, reliability of AI outputs in discipline-specific contexts, and uncertainty about institutional support if something goes wrong. These are not technophobic concerns. They are legitimate professional concerns that institutions have frequently failed to address before asking faculty to absorb the risk of innovation.<h3>The Professional Development Mismatch</h3>The dominant institutional response to faculty AI hesitancy is professional development — workshops, webinars, lunch-and-learns, peer learning communities. Some of this is genuinely useful. Most of it misses the actual problem.The research on effective professional development for technology integration in higher education identifies several conditions that predict whether faculty will actually change their practice: sustained engagement over time rather than one-time events, pedagogically grounded rather than technically focused content, departmental or disciplinary community rather than institution-wide generic programming, and concrete support for the actual implementation work of adapting a specific course. Professional development that covers how to log into a platform and generate a report does almost nothing to address whether integrating that platform will improve learning in a specific disciplinary context. That question requires disciplinary knowledge the professional development facilitator frequently does not have.What faculty consistently report wanting and rarely receiving is something more substantial: time to experiment without the pressure of student performance consequences, access to disciplinary colleagues who have genuinely integrated AI into similar courses and can speak honestly about what worked and what did not, and institutional signals that imperfect pilots during a genuine learning period are acceptable outcomes. Instead, what they frequently receive is an implicit expectation of seamless adoption on an accelerated timeline, with professional development as a substitute for the structural support that would actually change behavior.<h3>The Incentive Structure Nobody Wants to Name</h3>There is a structural dimension to faculty AI adoption patterns that the professional development conversation consistently avoids. In most research universities and a significant proportion of teaching institutions, faculty reward structures center on research productivity, grant acquisition, and publication record. Teaching innovation — including the substantial time investment of thoughtfully integrating AI tools into courses — is rewarded weakly if at all in tenure and promotion decisions. Adjunct and contingent faculty, who teach the majority of undergraduate courses at many institutions, have no job security that would make voluntary investment in institutional AI initiatives a rational use of their limited professional time.Asking faculty to lead AI transformation under these incentive conditions is asking them to invest substantially in an activity that their institutions do not adequately value, at a risk level their institutions have not adequately mitigated, for a transformation whose benefits will accrue primarily to the institution's efficiency metrics and the vendor's contract renewal. The surprise should not be that some faculty decline. The surprise is that as many faculty engage as willingly as they do.Genuine institutional support for faculty AI adoption would look like course releases or reduced loads for faculty undertaking substantial curriculum redesign. It would look like AI integration contributions weighted explicitly in annual reviews and promotion files. It would look like multi-year pilot commitments that give faculty enough time to learn a system properly rather than a semester to demo it. And it would look like institutional acknowledgment that the first iteration of any genuine innovation will be imperfect, and that imperfection during learning is not a professional failure.<h3>What Sustainable Adoption Actually Looks Like</h3>The institutions showing the most durable and meaningful faculty AI adoption in the research literature are not the ones that ran the most professional development workshops or sent the most communication about AI strategy. They are the ones that treated faculty as design partners rather than adoption targets — institutions where faculty had genuine voice in which tools were selected, how they were configured, and what success looked like.Participatory models of AI integration, where faculty with disciplinary expertise co-design implementation with educational technologists and learning scientists, produce tools that are better calibrated to actual learning goals and faculty who are genuinely invested in making them work because they had a hand in making them. This is slower than top-down adoption mandates. It is also more durable, more equitable, and more honest about the complexity of the work.Faculty are not the obstacle to AI transformation in higher education. They are the professional community whose knowledge is most essential to getting it right. Institutions that understand this will build AI education systems that actually improve learning. Institutions that frame faculty judgment as resistance to be overcome will build AI education systems that improve efficiency metrics and leave the deeper questions unanswered.What would it look like if your institution's next AI initiative began with a genuine question to faculty — not 'how do we get you on board?' but 'what would you need in order for this to actually work for your students?'&nbsp; Priscillar McMillan is a doctoral research assistant in Information Technology in Education at the University of Nevada, Reno, and founder of Kowa Agency. She writes weekly on AI, learning systems, and the institutional decisions that will shape education for the next generation.

Faculty Are Not Resisting AI. They Are Responding Rationally to an Irrational Ask.

<h3>Category: Global South &amp; Equity &nbsp; | &nbsp; Reading time: 7 min What Seven Years in a Zambian Classroom Taught Me About EdTech</h3>I spent seven years as a Guidance Teacher in Zambia's Ministry of General Education, teaching ICT and career guidance in a system that was doing remarkable things with very constrained resources. I watched brilliant students learn on equipment that would be considered obsolete in any well-resourced setting. I watched teachers design genuinely creative pedagogy around infrastructure limitations that most EdTech researchers have never encountered and do not think to account for.When I read the AI in education literature now — as a doctoral researcher at the University of Nevada, Reno, and as a former UNICEF consultant — I am struck by how consistently that world is absent from the conversation. The studies are conducted at well-resourced universities in high-income countries. The platforms are designed for reliable broadband, modern devices, and students whose first language is English. The findings are presented as if they describe education universally, when they describe a particular slice of it — the slice that already has the most.This is not a peripheral equity concern. It is a fundamental validity problem. If the AI-in-education research base is built almost entirely on evidence from high-income contexts, then the frameworks, policies, and systems being designed from that evidence are not designed for most of the world's learners. They are designed for the world's most advantaged learners and exported everywhere else.What the Research Base Actually Looks LikeA systematic review of AI feedback and equity in higher education reveals that research examining Global South implementation contexts is critically underrepresented relative to the student populations these contexts serve. When Global South contexts appear in the literature at all, they tend to appear as sites of need — places where AI might solve access problems — rather than as sites of knowledge, where different pedagogical traditions, infrastructure realities, and community values might actually improve AI system design for everyone.The research that does exist from Global South contexts surfaces patterns that high-income country research consistently misses. Connectivity is intermittent rather than continuous, meaning AI systems that require persistent internet access fail not occasionally but routinely. Device diversity is substantially wider, with students accessing learning platforms on everything from low-end Android phones to shared community computers, meaning platform designs optimized for laptop browsers exclude large populations by default. English-language training data underlies most AI educational tools, producing systems that perform significantly less reliably for students working in Bemba, Swahili, Shona, or any of the hundreds of languages in which actual learning happens across the African continent.These are not edge cases. They are the modal conditions for a significant proportion of the world's higher education students. Treating them as afterthoughts in research design is not just an equity failure. It is a scientific failure, because it means the evidence base does not describe the phenomenon it claims to describe.The Infrastructure Assumption Nobody StatesEvery AI feedback system I have encountered in the literature operates on a set of infrastructure assumptions that are so foundational they are never made explicit. Reliable broadband. Consistent electricity supply. Devices capable of running contemporary web applications. Institutional IT infrastructure capable of integrating with third-party platforms. Data storage and processing capacity sufficient for real-time adaptive responses.In Zambia, in much of Sub-Saharan Africa, in significant portions of South and Southeast Asia, in rural and remote communities across every continent including the United States, these assumptions fail routinely. When they fail, the AI system does not degrade gracefully. It simply does not work. The student who needed the personalized feedback receives nothing, and the gap between them and the student whose infrastructure held opens a little wider.I have worked as a UNICEF consultant on technology and education in contexts where these infrastructure realities are not hypothetical planning considerations but daily operational facts. The difference between how EdTech is designed in those contexts and how it is designed in Silicon Valley or Cambridge is not just a matter of resource allocation. It is a matter of who is in the room when design decisions are made, whose experience is treated as the default, and whose experience is treated as a special case to be addressed later if resources allow.What Culturally Responsive AI Design Would Actually RequireDesigning AI feedback systems that work equitably across Global South contexts is not primarily a technical challenge. The technical solutions — offline-capable architectures, low-bandwidth optimization, multilingual natural language processing — exist or are in active development. The harder challenge is the epistemological one: who gets to define what good learning looks like, whose pedagogical traditions inform how feedback is structured, and whose evidence counts as evidence.Most AI educational feedback systems embed, invisibly, a set of assumptions about what learning looks like that reflect particular cultural and institutional contexts. The assumption that learning is primarily individual rather than collective. The assumption that the teacher's role is to transmit information rather than to model membership in a knowledge community. The assumption that correct answers are the primary signal of understanding rather than the quality of reasoning or the richness of explanation. These are not universal features of good pedagogy. They are features of a particular tradition — and building them into AI systems at the foundational level means exporting them globally as if they were.Research on AI in Global South EFL education offers an instructive example. Studies examining AI tools in English as a Foreign Language contexts across African and Asian settings consistently find that tools designed around native-English pedagogical norms underserve learners whose relationship to English is complex, multiple, and embedded in multilingual realities that monolingual AI systems cannot model. The finding is not that AI cannot support EFL learning. It is that AI designed without that complexity in mind does not.The Contribution Global South Contexts Make to the FieldI want to be direct about something that the equity framing sometimes obscures. The argument for centering Global South contexts in AI education research is not only that it is fair to those students — though it is. It is that those contexts surface design requirements, failure modes, and pedagogical insights that make AI systems better for everyone.Teachers who have designed effective pedagogy under significant infrastructure constraints understand something about the relationship between technology and learning that teachers who have always had reliable tools do not. Students who navigate multilingual learning environments develop cognitive capacities that monolingual assessments do not capture. Communities with strong oral knowledge traditions have epistemological frameworks that challenge and enrich the text-centered assumptions baked into most AI educational tools.When Global South researchers, educators, and students are co-designers rather than afterthought beneficiaries of AI education systems, the systems that result are more robust, more flexible, and more honest about what they are actually doing and for whom. That is not charity. That is better science.What would it require for your institution's AI education initiatives to treat Global South implementation evidence not as a special case but as a primary source — and to build that standard into vendor selection, research partnerships, and pilot design from the beginning?&nbsp; Priscillar McMillan is a doctoral research assistant in Information Technology in Education at the University of Nevada, Reno, and founder of Kowa Agency. She writes weekly on AI, learning systems, and the institutional decisions that will shape education for the next generation.

The AI in Education Conversation Is Missing Most of the World

Category: AI &amp; Learning Outcomes &nbsp; | &nbsp; Reading time: 6 min The Measurement Problem Hiding in Plain SightPick up almost any study on AI feedback and learning outcomes and you will find a version of the same sentence: the intervention reduced cognitive load, as evidenced by improved performance scores. Read it twice. The cognitive load conclusion is drawn from the performance outcome — not from any direct measure of what was actually happening in the learner's mind during the learning process.This is not a minor methodological footnote. It is a structural problem that means we do not actually know, with any empirical precision, how AI feedback systems interact with human cognitive capacity during learning. We know what students scored. We do not know what was happening in their working memory while they were scoring it. And those are different questions with different implications for how we design AI systems that work.My literature review on AI feedback and learning retention identified this as one of the most significant gaps in the current evidence base. The field has a sophisticated theoretical framework — Sweller's Cognitive Load Theory — and almost no direct measurement of whether that framework is actually operating the way researchers assume in AI-enhanced environments.What Cognitive Load Theory Actually ClaimsCognitive Load Theory, developed by John Sweller in the late 1980s and extensively built upon since, proposes that human working memory has a limited capacity and that learning is most effective when instructional design manages the demands placed on that capacity deliberately. Three types of load are distinguished: intrinsic load, arising from the inherent complexity of the material itself; extraneous load, arising from poor instructional design that forces cognitive work unrelated to the learning goal; and germane load, arising from the cognitive processing that actually builds schemas and produces learning.The theoretical promise of AI feedback systems is compelling within this framework. Well-designed AI can reduce extraneous load by providing immediate, targeted corrections that eliminate the need for learners to search for feedback through laborious channels. It can manage intrinsic load by adapting problem difficulty to learner readiness, keeping students in the zone where challenge is productive rather than overwhelming. It can increase germane load by requiring students to engage actively with feedback rather than passively receiving grades.This is an elegant argument. The problem is that it is largely theoretical. When researchers claim that AI feedback reduces cognitive load and therefore improves learning, they are typically working backward from learning outcomes to infer the mechanism — not measuring the mechanism directly and then observing its effect on outcomes. The difference matters enormously for design decisions.What Direct Measurement Would Actually Tell UsDirect cognitive load measurement during AI-assisted learning would use tools that capture what is happening in real time as students engage with feedback. Eye-tracking measures where and how long learners fixate on different elements of a learning interface, revealing attention patterns and cognitive effort. Response time analysis captures how long students take to process feedback and formulate subsequent responses, with longer response times often indicating higher cognitive demand. Physiological measures — galvanic skin response, heart rate variability, pupil dilation — provide continuous signals of cognitive and emotional arousal during learning.These methods exist. They are used extensively in human-computer interaction research, in usability studies, and in cognitive psychology laboratories. They are rarely used in AI education research, partly because they require specialized equipment and expertise, partly because large-scale field studies are harder to instrument than controlled laboratory experiments, and partly because the field has been content to infer mechanisms from outcomes rather than observe them directly.What direct measurement would reveal that outcome-based inference cannot is the temporal dynamics of cognitive load during AI feedback interactions. Does load spike when feedback arrives? Does it peak at different moments for novice versus experienced learners? Does multimodal feedback — combining text and visual elements — reduce load or increase it by splitting attention? Does the design of the feedback interface itself impose extraneous load through cluttered presentation or confusing navigation? We do not know the answers to these questions with empirical confidence, because we are not measuring the right things.The Productive Failure ConnectionOne of the most interesting findings in recent learning science literature involves what researchers call productive failure — the phenomenon where students who struggle with problems before receiving instruction sometimes show better learning outcomes than students who receive instruction first. The mechanism appears to involve the cognitive effort of the struggle itself: attempting to solve a problem without adequate resources activates relevant prior knowledge, surfaces misconceptions, and primes the learner to extract meaning from subsequent instruction more efficiently.AI feedback systems, almost by design, tend to minimize struggle. Immediate feedback interrupts the productive difficulty of working through confusion. Hints and scaffolding reduce the cognitive demand of problem-solving. Adaptive difficulty adjustment ensures students are rarely stuck for long. If cognitive load during struggle is actually germane load — the kind that builds understanding — then AI systems optimized to reduce all load may be inadvertently reducing the cognitive work that matters most.This is not an argument against AI feedback. It is an argument that the relationship between cognitive load and learning in AI-enhanced environments is more complex than the dominant narrative suggests, and that we need direct measurement to understand it rather than theoretical inference from incomplete data.What This Means for How We Build and Evaluate AI SystemsIf institutions are making procurement decisions about AI feedback systems based on the assumption that these systems reduce cognitive load and therefore improve learning, they are building on a foundation of theoretical inference rather than empirical evidence. That does not mean the inference is wrong. It means we do not actually know whether it is right — and for educational technology making claims about how human minds learn, that uncertainty should matter.The practical implication is that AI feedback systems should be developed and evaluated with cognitive processes, not just learning outcomes, as primary variables. This requires partnerships between learning scientists who understand measurement methodology, educational technologists who understand system design, and practitioners who understand classroom reality. It requires instruments that capture what is happening during learning, not just what happened afterward. And it requires a research culture willing to complicate the success narrative when the mechanism does not match the theory.Until we measure what we claim to know, we are designing AI learning systems by inference and hoping the theory holds. Sometimes it will. The question is whether that is good enough for the students whose learning depends on getting it right.What would it change about how AI feedback tools are selected and evaluated at your institution if cognitive load during learning — not just post-learning performance — were a required evidence standard?&nbsp; Priscillar McMillan is a doctoral research assistant in Information Technology in Education at the University of Nevada, Reno, and founder of Kowa Agency. She writes weekly on AI, learning systems, and the institutional decisions that will shape education for the next generation.

We Think We Know How AI Affects Cognitive Load. The Data Says Otherwise.

<figure id="figure-4s7x1323emml2fahl"><img id="4s7x1323emml2fahl" src="https://cdn.durable.co/covers/2cZc0N0BwvrYnYBZC1nDr0VzgapC6UZael4RjZuYZNDB18L1U9dxUmjuLKAFrHVC.jpg" alt="" title="" class="object-cover bg-gray-100 rounded-sm md:rounded-md lg:rounded-lg" style="aspect-ratio: 16/9; object-position: center center;" data-corners="default" data-aspect="16:9" data-caption="" data-positionX="undefined" data-positionY="undefined" data-fullWidth="false"/></figure>The Problem Nobody Is MeasuringThere is a version of the AI-in-education story that looks like success from every angle. Students complete assignments. Grades improve. Engagement metrics trend upward. Institutions publish case studies. And then, six months later, those same students sit in the next course, the one that assumes they retained what they learned, and something has quietly gone missing.This is the transfer problem, and it is the most important thing the AI-in-education debate is not talking about. We have invested enormous energy in the question of whether AI helps students perform better in the short term. We have invested almost nothing in the question of whether that performance represents learning that actually moves — that carries forward into new contexts, new problems, new disciplines.I study AI adoption in higher education at UNR, and I teach BUS 101. In both roles I keep running into the same gap: the research on AI feedback and learning outcomes almost universally measures retention of similar problems within weeks of instruction. That is a narrow definition of learning, and it is doing quiet damage to how institutions interpret their own AI pilots.What Transfer Actually MeansTransfer, in the cognitive science literature, refers to the ability to apply knowledge or skills acquired in one context to a meaningfully different context. Near transfer involves applying learning to situations closely resembling the original. Far transfer involves applying learning to situations that look substantially different — different problems, different disciplines, different professional settings.The distinction matters because most AI feedback systems are optimized for near-transfer performance. An intelligent tutoring system in introductory statistics helps students solve the specific problem types the system covers. It gives immediate feedback on those problems, and students get better at those problems. What it does not reliably do — and what the research largely does not measure — is whether students can take statistical reasoning into a business case, a policy brief, or a research design they have never seen before.The Wharton AI tutoring study that I wrote about in an earlier post is the clearest public evidence of this gap. Students using AI performed 48% better on assisted practice and 17% worse on unassisted tests. That test was still closely related to the practice material. The transfer gap, measured across genuinely novel contexts, would almost certainly be wider — and we do not have the data to know by how much, because researchers rarely follow students far enough forward to find out.Why Institutions Are Not Asking This QuestionPart of the answer is structural. Measuring far transfer requires following students across time, across courses, and sometimes across institutions. That is expensive, logistically complicated, and difficult to publish in clean experimental designs. Measuring whether a student who used an AI writing tutor in freshman composition can construct a persuasive argument in a senior capstone is genuinely hard research. Measuring whether they scored higher on the next essay in the same course is easy research. The field has optimized for what is easy to measure.Part of the answer is also incentive-driven. EdTech vendors selling AI feedback platforms have no commercial interest in funding studies that might reveal their products produce performance gains that do not transfer. Institutional administrators under pressure to demonstrate AI ROI have no political interest in commissioning research that complicates the success narrative. The people with the resources to fund transfer research are often the people who benefit most from its absence.And part of the answer is conceptual. There is a quiet assumption embedded in much AI-in-education discourse that learning is essentially the accumulation of correct responses — that if a student has answered enough questions correctly with the help of AI feedback, something durable has been deposited in their understanding. Cognitive science has known for decades that this is not how human memory works. Correct performance and durable understanding are related but distinct, and the conditions that produce one do not automatically produce the other.What the Evidence Does SayThe literature on retrieval practice offers the clearest evidence-based path forward. When students are required to actively retrieve information from memory — rather than receiving it through re-reading or AI-generated summaries — retention and transfer both improve substantially. The mechanism is not mysterious: the act of retrieval strengthens memory traces in ways that passive reception does not. AI systems that incorporate retrieval practice into their feedback loops, requiring students to attempt recall before receiving assistance, show more durable learning than systems that provide assistance on demand.Spaced practice — distributing learning across time rather than concentrating it immediately before assessment — similarly improves both retention and transfer. AI systems that introduce deliberate spacing into their adaptive sequencing, revisiting earlier concepts after intervals rather than moving linearly through curriculum, show stronger long-term outcomes. The irony is that neither retrieval practice nor spaced repetition is technologically difficult to implement in AI systems. They are pedagogically demanding because they require institutions to build curricula around learning science rather than administrative convenience.The deeper implication is this: AI feedback that is designed around performance optimization will produce performance. AI feedback that is designed around transfer will produce understanding. These are different design goals, and achieving the second requires institutions to be explicit about which one they are actually pursuing.The Accountability QuestionI want to end on something that I think the research has not yet reckoned with honestly. If students are leaving courses with AI-assisted performance gains that do not transfer to subsequent learning, the cost of that gap does not show up immediately. It shows up later, in the next course, in the internship, in the first year of a job, in the graduate program that expected preparation students technically completed but did not genuinely acquire.By then, the AI feedback system that produced the gap has been credited with the performance improvement and evaluated against that credit. The transfer failure is invisible to the data that institutions track, because institutions do not track it. This is how a technology can simultaneously improve short-term outcomes and degrade long-term educational quality — and why the absence of transfer research is not a neutral gap. It is a gap that systematically favors certain conclusions over others.What would it change about how your institution evaluates its AI investments if transfer — not just completion and grades — were a required outcome metric? And who in your organization has the standing to make that demand?&nbsp; Priscillar McMillan is a doctoral research assistant in Information Technology in Education at the University of Nevada, Reno, and founder of Kowa Agency. She writes weekly on AI, learning systems, and the institutional decisions that will shape education for the next generation.

Your Students Can Do It With AI. The Question Is Whether They Can Do It Without.  Category: AI & Learning Outcomes

Artificial Intelligence and Human Learning Concept with Digital Brain

Category: The Educator's Dilemma &nbsp; | &nbsp; Reading time: 6 min At a recent education and technology conference, a panel moderator asked a room full of educators what they needed to feel confident integrating AI into their classrooms. The responses came fast: more training, clearer guidelines, better tools, more time. What struck me was not the content of the answers. It was the exhaustion underneath them.These were not resistant teachers. They were not technophobes clinging to chalkboards. They were professionals who had already absorbed remote learning, learning management system overhauls, post-pandemic re-enrollment crises, and chronic underfunding — and were now being handed one more transformation and told to lead it.I study AI adoption in higher education. I also teach in it. And I want to name something directly: when institutions point to teachers as the primary change agents for AI integration, they are not empowering educators. They are offloading an institutional responsibility onto the most structurally vulnerable people in the system.<h3>What We Are Actually Asking Teachers to Do</h3>To understand the weight of what is being asked, it helps to break AI integration into what it actually requires at the institutional level. I think about this in three distinct pillars — and the failure to keep them separate is where most AI-in-education strategies collapse.Pillar One is safe adoption — ensuring that AI tools being introduced into learning environments are vetted for data privacy, age-appropriateness, equity of access, and alignment with learning goals. This is a policy and procurement function. It requires legal review, IT governance, and institutional risk assessment. It is not a classroom teacher's job.Pillar Two is curriculum redesign — rethinking what we ask students to do, how we assess them, and what learning outcomes we are actually trying to build, given that AI now exists as a permanent feature of the cognitive landscape. This is a curriculum development function. It requires instructional designers, faculty governance, and serious pedagogical debate. It can involve teachers, but it cannot be carried entirely by them on top of full teaching loads.Pillar Three is systemic overhaul — the deeper question of what education is for in an AI-saturated world, and whether the structures, incentives, and metrics of our institutions are designed to answer that question honestly. This is a leadership function. It requires institutional courage, not just institutional communication.What most institutions are doing right now is collapsing all three pillars and handing the rubble to classroom teachers with a professional development workshop and a ChatGPT account.<h3>The Burnout Data Is Not Background Noise</h3>This matters more urgently because educator burnout is not a peripheral concern — it is the context in which AI integration is being demanded. A 2024 RAND Corporation survey found that nearly half of teachers reported frequent job-related stress, with workload consistently cited as the primary driver. In higher education, adjunct and contingent faculty — who now make up the majority of the teaching workforce — frequently carry heavy course loads with minimal institutional support and no job security that would incentivize volunteering as innovation pilots.Asking these educators to simultaneously teach, grade, learn new AI tools, redesign their curricula, and serve as the human face of an institutional transformation is not a growth opportunity. It is a system externalizing its costs onto its most precarious workers.This is not new. Education has a long history of institutional inertia dressed as teacher empowerment. When standardized testing regimes expanded, teachers were told to innovate within the new constraints. When learning management systems arrived, teachers were told to redesign their courses. When the pandemic hit, teachers were told to pivot to online delivery in a weekend. The pattern is consistent: the institution protects its structure, and the teacher absorbs the disruption.<h3>What Genuine Institutional Leadership Looks Like</h3>I am not arguing that teachers should be excluded from AI integration conversations. Quite the opposite — their classroom-level knowledge is irreplaceable data for any institution serious about designing AI adoption that actually works. But there is a difference between including teachers as informed contributors and conscripting them as unpaid implementation leads.Genuine institutional leadership on AI looks like a few specific things. It looks like dedicated release time for faculty who are asked to redesign courses — not a summer stipend that amounts to poverty wages for weeks of professional labor. It looks like an AI governance committee with real authority, not an advisory group whose recommendations get thanked and shelved. It looks like curriculum offices that take on the structural redesign work rather than sending teachers to a three-hour workshop and calling it transformation. And it looks like institutional leaders willing to have the harder conversation: that some of what we currently call education is assessment theater, and AI has simply made that visible.The EDUCAUSE 2024 Horizon Report identifies AI as the defining challenge for higher education in this decade. The report is careful and well-researched. What it cannot compel is the institutional will to respond at the level the challenge actually demands.<h3>The Accountability Gap</h3>Here is the structural problem underneath the pattern. When a teacher's AI integration effort fails — when students game the tool, when learning outcomes drop, when the new pedagogy produces confusion rather than clarity — the accountability lands on the teacher. The institutional decision to deploy without design infrastructure, without governance, without genuine curriculum support, disappears into the background.This accountability asymmetry is what makes the current arrangement not just ineffective, but unjust. Teachers are being asked to carry reputational and professional risk for decisions that were made well above their pay grade, in procurement offices and administrative retreats, with little input from the people who will live with the consequences.Changing this requires institutions to do something genuinely difficult: accept that AI integration is a governance challenge first, a curriculum challenge second, and a classroom challenge third. The order matters. When you reverse it — as most institutions currently do — you get burned-out teachers, inconsistent student experiences, and AI pilots that produce impressive slide decks and underwhelming learning outcomes.The educators I have spoken with do not need another professional development workshop. They need institutional partners who are willing to do their share of the work — and leaders with enough intellectual honesty to admit that the current arrangement is not empowerment.It is abdication with better branding.What would it look like if your institution's AI integration strategy was held accountable to educator workload data, not just student performance metrics? And who in your organization is positioned to ask that question out loud?&nbsp; Priscillar McMillan is a doctoral research assistant in Information Technology in Education at the University of Nevada, Reno, and founder of Kowa Agency. She writes weekly on AI, learning systems, and the institutional decisions that will shape education for the next generation.

Why Telling Teachers to Lead the AI Transition Is Educational Malpractice Category: The Educator's Dilemma | Reading time: 6 min

Priscillar Simon Banda

Priscillar McMillan is a high-impact Change Manager, ERP Implementation Lead, and Technical Program Manager with deep expertise in modernizing systems, aligning cross-functional teams, and driving transformation across regulated industries.

She has led full lifecycle implementations of enterprise systems — from client onboarding platforms and claims operations to medical compliance trackers and government reporting modules. Whether deploying ERP solutions or leading multi-department rollouts, Priscillar specializes in guiding organizations through the technical, operational, and human side of change.

Her track record includes:
	•	Leading cross-functional teams across IT, compliance, medical, and client services
	•	Managing system integrations tied to HIPAA, ORR-5, and federal grant compliance
	•	Driving complex ERP rollouts with minimal disruption and high adoption
	•	Building reporting and automation tools to enhance program visibility and executive decision-making
	•	Training and transitioning 100+ users across multiple departments through system migrations

Priscillar’s work spans FinTech, HealthTech, public sector, and nonprofit environments. She is fluent in leading enterprise-wide change, designing scalable operations, and aligning tech execution with leadership strategy. Known for her clear communication and structured thinking, she bridges the gap between executive vision and system delivery.

If your organization is navigating digital transformation, system migration, or program expansion — she is the technical operator and change architect you want in the room.


Priscillar Simon Banda

Blog

Problem-forward, Global South angle Why No Homeschooling ...

Faculty Are Not Resisting AI. They Are Responding Rationa...

The AI in Education Conversation Is Missing Most of the W...

We Think We Know How AI Affects Cognitive Load. The Data ...

Your Students Can Do It With AI. The Question Is Whether ...

Why Telling Teachers to Lead the AI Transition Is Educati...