Friends,
Been lurking, reading from the corner of my eye as I struggle with
manuscripts, too many and all overdue.
I may try to return to this later, but I want to clarify an issue
that may have gotten tangled about what learning is and how it works.
The issue is anchored in how we use the words "learning" and "designing."
The reason that I argue that machines and artifacts can neither learn
nor design is that they lack the properties of consciousness and
agency required for the meaning of these words as we use them.
Many kinds of systems, sub-systems, and even single entities (for
example, neurons) can learn in some sense. Learning in the larger
sense that I use the word takes place on a higher, integrative level.
Nearly everyone I know on this list recognizes the difference between
two kinds of learning when it comes to our teaching. Requiring
students to memorize, repeat, and imitate something may involve a
kind of learning, but it is inferior and mechanical. Asking students
to recognize, identify, and choose problems, then develop solutions,
is the kind of learning we encourage. There are more aspects to this,
and they can be stated in greater detail, but the point is clear.
Learning has to do with choice among alternatives at every level,
from problem seeking to deciding among possible relevant factors in
problem and in solution to choosing among alternative solution paths
and finally to choosing solutions. Choice requires agency.
Artifacts and machines can perform many of the different kinds of
functions. Machines may not be mechanical -- just as a sculpted model
of a painting is a machine, so a software program may serve as a
machine. The issue is whether machines can integrate information, AND
organize it as embodied knowledge, AND choose among alternatives
based on a DESIRED OUTCOME chosen by the learning agent rather than
programmed by an external agent. On this basis, only human beings can
learn.
It's true that part of this is about the future. Things may change.
Anyone who recalls Isaac Asimov's robot novels and stories will
recall a hundred variations on how entities that some consider
machines may actually be conscious agents who learn. In making my
statement that machines cannot learn, including automata of all
kinds, computers, and programs, as well as subsystems or neurons, I
speak of learning at a level of choice and conscious organization.
When machines can choose and CONSCIOUSLY organize, they will be able
to learn.
There are many ways to use the word learning in which all kinds of
systems and entities can and do learn. Define the term differently,
and I will agree that neurons learn, traffic systems learn, forests
learn, and computers certainly learn. Nevertheless, they do not learn
and they cannot therefore design at the level of conscious choice,
particularly not conscious choice involving ethics and phronesis --
the quality of applied or practical wisdom we seek in learning and in
design.
Perhaps I shall return to this fascinating thread. Depends on the
stack of papers that sits here, jeering, taunting, rustling and
whispering in the darkness. My manuscripts do not learn, but I am
convinced they are motivated by the desire to harass me.
Warm wishes,
Ken
--
Ken Friedman
Professor of Leadership and Strategic Design
Institute for Communication, Culture, and Language
Norwegian School of Management
Design Research Center
Denmark's Design School
email: [log in to unmask]
|