Some Rough Thoughts on Learning in the 21st Century

1/11/2024

Joe Iuliano, Assistant Head of Academic Affairs

Human Learning

“Learning is the change in the behavior of an organism that is a result of prior experience. Learning theory seeks to explain how individuals acquire, process, retain, and recall knowledge during the process of learning. Environmental, cognitive, and emotional influences, along with prior experiences, play a vital role in comprehending, acquiring, and retaining skills or knowledge. Motivation plays an important role in enabling the process of learning and is said to be the driving force where activity is started and sustained to achieve a target.” — NIH National Library of Medicine: Learning Theories (https://www.ncbi.nlm.nih.gov/books/NBK562189/ on 1/10/2024)

The NIH—what, the National Institute of Health offers insight on learning? Go figure (but be sure to choose one of the learning methods below. And yes, NIH does offer insights into the science of teaching…so here we go. According to this hopefully still prestigious government agency, the five most prevalent theoretical learning models today are, in very brief form, the following (the parenthetical asides are mine, of course):

Behaviorism—“learning occurs by linking stimuli and responses” (‘Sit Ubu, sit. Good dog… here’s a treat.’)
Cognitivism—“learning occurs through the processing of information internally rather than merely responding to an external stimulus” (‘Touche!’ says Piaget, ‘Metacog that!’)
Constructivism—“individuals learn by constructing new ideas, and an understanding of the world is based on prior knowledge and experiences” (‘My knowledge is my power and yours is yours…sort of.’ “Can we fix it?” asks Bob the Builder.)
Connectivism— “learning is through the formation of connections between each other as well as their roles, hobbies, and other aspects of life” (“Only connect!” says E.M. Forster in the front leaf of Howard’s End; if it works for love why not for a love of learning?)
Humanism—“learning is a natural desire with the ultimate goal of achieving self-actualization (‘Reach for the top!’ says Maslow.)

And there are probably as many as 25+ more learning theories out there too. For teachers, once a theoretical perspective is adopted (or a combination of them), it’s on to curriculum design and delivery and pedagogy to make that happen. Well…and relationship building, understanding learning styles, and igniting motivation because this is human learning.

I want to acknowledge that, as a human, I learned a thing or two while writing the text above. Such learning, human learning, is, I believe, existentially necessary and serendipitous for individuals, communities, and humanity in general.

Now the harder part.

Machine Learning

“…machine learning is the process of teaching a computer system how to make accurate predictions when fed data.”

While the two main methods of machine learning are reasonably understandable, I am not entirely confident that I get how this works, but here’s something like it:

supervised machine learning — teaching/learning by example through the labelling of galaxies of data (an exponentialized version of Wallace Stevens’ poem “Thirteen Ways of Looking at a Blackbird”)
unsupervised machine learning — “tasks algorithms with identifying patterns in data, trying to spot similarities that split that data into categories” (my carbon dioxide detector studies hard to become IBM’s Watson)
semi-supervised machine learning — “the approach mixes supervised and unsupervised learning” (this is not exactly cats and dogs consorting, but kind of like and I can’t explain either one of these circumstances briefly)

and evolutionary computation, expert systems, deep learning, mathematical models, neural networks, etc.

[definitions from “What is machine learning? Everything you need to know, Nick Heath, from https://www.zdnet.com/article/what-is-machine-learning-everything-you-need-to-know/ on 1/10/2024]

I have a lot to learn here as this is quite complicated. If we can teach a machine to learn through a variety of complex methodologies, and I can teach a human to learn through a variety of pedagogical and coaching practices, and if I can teach and learn myself through reading, listening, writing, studying ideas and data, then perhaps I can gain a greater understanding of how machine learning really works—and I am intrinsically motivated to do so. Thus from motivated human learning comes an understanding of machine learning; from educated, motivated human intelligence comes artificial motivated machine intelligence for… the creation of greater human intelligence, yes?

As an intrinsically motivated human learner, I’ve got a few possibly dis-associated questions about all this:

Are those constant critics of the human learning system (they’ve been around for centuries) going to show up in equal numbers to criticize the machine learning system (because, like the human learning system it has human teachers too?— until, unlike the human system, they don’t) Or will the machine learning system be more effective (by some metric to be determined) because it’s machine based?

Haven’t the teachers of machines been taught by the teachers of humans in that human learning system (schools) that has been so heavily criticized so regularly? Hmmm…

Will machines be intrinsically motivated to learn, extrinsically motivated to learn, both, neither? Does artificial intelligence incorporate motivation?

I went and looked up motivation in machine learning immediately after I encountered this concept because it was a compelling question for me (admittedly, the first two questions above represent my being nudgy). Motivation—sometimes its necessity and sometimes, it’s “for the sake of”—is essential to effective learning so I wonder if it can somehow be incorporated in machine learning. I found a paper titled, “Intrinsic and Extrinsic Motivation in Intelligent Systems,” authored in 2020 by Henry Lieberman of MIT’s Computer Science and Artificial Intelligence Laboratory. I have begun my reading and will use a bit of human intelligence to pursue some more learning of my own, but I will also get some great help with this learning endeavor through this year’s Bissell Grogan Symposium Keynote Presentation and Workshops. With a stable of outstanding expert speakers and workshop leaders addressing a range of AI applications from the arts to medicine, this year’s Symposium will provide students and faculty with a thought-provoking, motivating human learning opportunity, one delivered through a variety of those frameworks mentioned at the outset of this writing. Through this experience, Brimmer’s student will 'human learn' their way to a better understanding of how machine's learn and what they and we can do with this learning to imagine, question, and create in hour human world. Let the learning continue!

Some Rough Thoughts on Learning in the 21st Century

Brimmer and May School