Machine Learning and Human Learning

EPOL 479 · Fall 2026, Term A · 4 credit hours

Towards Synthesis: Rethinking Learning in the AI Era

Both sections meet together as a single seminar and are held to the same expectations under this syllabus.

Instructor: Professor Liam Magee, Department of Education Policy, Organization and Leadership (EPOL), College of Education E-mail: lmagee@illinois.edu — by far the best way to reach me Office hours: by appointment only. Email me and we will find a time; I am glad to meet over Zoom at short notice.

Course description

Catalog description. Examines the differences between machine and human learning and the ways in which machines can complement human learning. It examines technical definitions of supervised and unsupervised machine learning, as well as broader views of mechanical intelligence able to replicate or exceed human intelligence. The course will also explore practical applications of learning analytics and artificial intelligence in learning management systems and other educational tools.

This section approaches that remit philosophically as well as technically, responding to a new learning reality shaped by advances in machine learning. It traces the long history of tools that externalize memory, from the stylus to the internet and the smartphone, and examines the reciprocal learning relationship between humans and AI systems. Through concepts such as technosymbiosis and symbiotic pedagogy, we ask how both forms of learning — machine and human — come together in synthesis.

Prerequisites: none.

Course outline

What does it mean for both humans and machines to “learn”? The usual picture has an active human learner using a passive machine. That picture does not survive contact with the present. People select the data, goals, categories, and evaluation criteria that shape computational systems; those systems in turn reorganize what people notice, practice, and come to know.

The course proceeds through eight concepts, each of which does double duty as a term in the philosophy of mind and a term of art in machine learning:

1. Synthesis — how a new relation forms in which both sides change
2. Experience — expectation, encounter, contradiction, revision
3. Recognition — identification, validation, compliance, reciprocity
4. Attention — selection in transformers and in brains
5. Consciousness — cognition, self-report, and the nonconscious
6. Alignment — whose values, translated by whom, into what rewards
7. Critique — immanent rather than merely external
8. Technosymbiosis — interdependence, and whether it is equal

The shared vocabulary is the course’s central problem, not its solution. That transformer “attention” and human attention bear the same name is a question to investigate, not a conclusion to assume. Throughout, we read Hegel’s Phenomenology of Spirit alongside contemporary machine learning literature, letting each estrange the other.

The emphasis is on developing defensible positions and recognizing the limits of your own, not on reproducing settled answers. There are none.

What I expect from you

This is a compressed eight-week course carrying four credit hours. You will engage in five kinds of activity:

Budget at least 20 hours per week in total, of which roughly 17 fall outside the seminar itself. A four-credit course delivered over eight weeks rather than sixteen does not contain less work; it contains the same work, twice as fast. Most of your learning will happen outside our three hours together. Come to seminar having read — the discussion is not a substitute for the reading, and it will not be legible to you if you have not done it.

Because we meet online, the asynchronous half of the course carries real weight. Readings, weekly responses, and AI dialogue practice happen between sessions on Canvas; the live seminar is where those threads are argued out.

Philosophy rewards rereading in a way that few things do. A passage of Hegel that is opaque on first pass is often tractable on the third. Plan for that.

What you should expect from me

• I will hold office hours as scheduled, and will find another time if you cannot make them.
• I will answer email within 24 hours on weekdays.
• I will return written work with substantive comments within one week.
• I will post clarifications when a reading or an assignment is causing broad difficulty.
• I will give you my own written commentary at two points — across Sessions 3/4 and Sessions 6/7 — so you are never guessing about where you stand.
• I will grade the quality of your thinking, not its agreement with mine. A well-argued position I disagree with earns full marks; an echo of my own view, poorly defended, does not.
• If nobody in the class managed something, I will treat that as a fact about my teaching rather than about you, and act accordingly.

Course objectives

By the end of this course, you will be able to:

1. Understand the historical relationship between technology and learning.
2. Explore the phenomenology of AI-mediated learning experiences.
3. Develop critical frameworks for evaluating AI in educational contexts.
4. Practice synthesis of human and machine learning approaches.
5. Distinguish the technical from the philosophical sense of shared terms such as attention, alignment, and learning, and assess where the analogy holds.
6. Construct an immanent critique of an AI system or institution using its own stated norms.

Texts, readings, and tools

Books to obtain:

• G. W. F. Hegel, Phenomenology of Spirit, trans. Terry Pinkard (Cambridge University Press, 2018).
• N. Katherine Hayles, Unthought: The Power of the Cognitive Nonconscious (University of Chicago Press, 2017).

All other readings — journal articles, ML papers, and excerpts — are provided through the course site. See the weekly index for the full reading list by week and references by week for complete citations.

Tools: the running record of your responses is kept in Google Docs; Canvas carries readings, announcements, Zoom links, recordings, and grades. You will also need access to at least one contemporary generative AI system, since a weekly chat session is part of the assessed cycle. Free tiers are sufficient. If cost or access is a barrier, tell me early and we will arrange an alternative — no student should be disadvantaged by this requirement.

Evaluation

Think of the course as six-plus-one iterations on your own initial impulses, with a mix of self- and other-assessment. You start by writing down what you already think. Each week you revise it in the light of a new concept. At the end you synthesize. You will keep the whole sequence in a Google Doc, so that the revisions stay visible to you and to me — the record of the changes is the work.

Baseline response (not assessed)

In the first fortnight, write 200–300 words — longer is fine, but not more than 500 — in your own words, on:

• How would you define machine learning, in your own words and based on your own understanding of the term, today?
• How would you define human learning?
• What are the similarities between the two? What are the differences?
• What can we say about the relationship between the two? In what ways does machine learning build upon human learning? And in what ways can human learning benefit — or not — from machine learning?
• What is the most important question you want to pose and have answered about AI in relation to learning and education? Put another way: what are you yourself hoping to learn here, and how would you answer that question provisionally?

This sounds like a survey, and in a way it is. More importantly it sets a baseline of your own understanding, which the rest of the course revisits. It is not graded.

1. Formative self-assessments (60%)

Six weekly cycles, one for each of the concepts in Sessions 2–7. Each week you:

1. take notes from the readings, the lecture, and the group discussion;
2. hold a chat session with your preferred AI system;
3. copy your previous responses, then revise and extend them;
4. bring the result to peer review in class;
5. close with a short reflective note, and grade your own contribution.

Why self-assessment. It is subjective, obviously. Two arguments for it anyway. The first is Hegelian, and this is a Hegel course: I can only judge the facade of your learning, its outward appearance. Only you know whether you have learned something. Aligning the assessment of the course with its content is itself a Hegelian move. The second is that in an age of AI we want learning to become more accountable, not less — and accountability that you exercise on yourself is the kind that survives contact with a tool that can write for you.

Rubric. There isn’t a conventional one, which I know is uncomfortable. I suggest grading the derivative: what did I know last week, what do I know now, and how would I grade that rate of change? I reserve the right to upgrade marks that strike me as excessively punitive.

2. Peer commentary (10%)

Two rounds — across Sessions 3/4 and again across Sessions 6/7 — in which you read and comment on a peer’s revised responses. This is a hurdle requirement: the 10% is awarded on completing it in good faith, and the commentary itself is not graded on quality. I will provide my own commentary at the same two points, so you have an outside reading of your work twice before the final report.

3. Final synthesis report (30%)

A summative piece that synthesizes the course materials — a report, an essay, or a curriculum design. It builds directly on the formative work: the six iterations are the raw material, and this is the seventh. Its exact shape is something we will settle together during the course; the form should follow what you have actually been arguing. Due Friday 16 October at 11:59 PM Central, the last day of Term A.

Use of AI in this course

You are studying these systems, so you should use them — a weekly AI chat session is a required part of the formative cycle. What I ask is that the use be visible and reflective rather than concealed.

• Permitted and encouraged throughout: using AI to explore ideas, test arguments, find sources you then verify, and stress-test your drafts. This is course material.
• Required: where AI shaped a submitted piece of work, append a short note describing how. A sentence or two is enough. This is a scholarly disclosure, not a confession, and it does not affect your grade.
• Not permitted: submitting generated prose as your own unexamined writing. Not because it is cheating in some abstract sense, but because the point of the assignment is the change in you that writing produces. Outsourcing that leaves you with a document and no education — and in a course assessed on the rate of that change, it is self-defeating as well as a breach.
• Verify every citation. Generated bibliographies routinely contain plausible references to work that does not exist. An uncheckable citation will be treated as an error of scholarship.

Beyond this, the University’s Academic Integrity Policy applies in full.

Weekly schedule

The seminar meets on Mondays. There is no class on Monday 7 September (Labor Day); that week runs as a reading week, and Critique and the closing Synthesis session are combined in the final seminar.

Each seminar runs three hours, with ten-minute breaks between them:

The final session is a double one in substance: it closes the critique thread and then draws the eight concepts together, so budget extra reading time for that week.

Grading scale

Policies

Attendance. Attendance means joining the live Zoom seminar. With only seven seminars, each one is roughly a seventh of the term, so an absence has real weight. Tell me in advance if you must miss one and we will arrange for you to make up the discussion.

Late work. The weekly cycle only works if it is weekly — a self-assessment written three weeks late is a memory exercise, not a revision, and the peer commentary rounds depend on your classmates having something to read. Tell me before the deadline if you are going to be late and we will sort it out. The final report is the one hard deadline: Term A grades are due to the University shortly after the term ends, so extensions on it must be arranged in advance. If something serious is going on, talk to me — I would much rather grant an extension than receive nothing.

Students with disabilities. To obtain accommodations, contact Disability Resources and Educational Services (DRES) as early in the term as possible, and speak with me so we can implement them promptly. If you have a condition that may require accommodation but are not yet registered with DRES, please come and see me regardless.

Recording. Seminars are recorded and posted to Canvas. The recording is a resource for revision and for genuine conflicts, not a substitute for attending — the argument only happens if you are in it.

Mental health. Seeking help is a sign of strength. The Counseling Center and McKinley Health Center offer free, confidential services to enrolled students.

The official University-formatted version of this syllabus, built on the UIUC template, is kept alongside this file in syllabus/. All submissions are made through Canvas.