UIs That Aid
Learning and Doing
Alan Kay's sketch for the Dynabook, a laptop computer for children, c.1968
Another area of our work at Viewpoints Research concerns user interface design, specifically the design of a UI that can flexibly help its users learn content and applications by providing guidance, taking advice, and learning. We sometimes refer to this as a "mentoring" user interface.
This new user interface will be created by greatly enhancing the process of framework authoring: the making of projects and applications that allow end-users to make content, e.g. a teacher making a learning project for a child, who will then learn by making a model of an idea, or a new kind of simulation framework for adults, who will need to be helped to learn how to make simulations.
We want the content and applications to not just provide functionality, but also be able to teach the end-users how to learn and use the new framework, and be able to itself learn about the user's styles and motivations to create the best learning-and-doing environment.
Both the idea and need for a mentoring interface is not new. It
was a dream of the sixties and a number of attempts were made to
invent one, some of large scale (e.g. Plato , SOPHIE , etc.) More
recently there have been a number of major efforts in this direction,
some of them impressive: e.g. ACT-R by John Anderson and colleagues.
However, quite a bit of expertise must be gained by the curriculum
and application designers, and the ratio and costs of "student hours"
to "content authoring" hours is very high and expensive. An analogy
(that shouldn't be pushed too far) is between the early days of
computerized layout and typesetting, basically done by experts in
the older typesetting technologies, and the end-user revolution
caused by the inventions of the modern GUI and Desktop Publishing
at Xerox PARC. All of a sudden end-users could and did start creating
millions of documents with all the paraphernalia formerly used by
experts. Our main users will be teachers (and many applications
documenters) and this user community needs a much stronger easier
approach to creating explanations and guidance that can be automatically
manifested by the user interface. This is much more than a "training"
and "education" problem because it addresses needs that are as pervasive
as the need for a user interface itself: it is a deep everyday computing,
learning, productivity, cost (and even morale) problem for most
users and most institutions.
In practice, such a system must be very flexible about how it advises the end-user. For example, in some cases it will not be able to pin down just what went wrong (or it might think something went wrong but the end-user was just taking a valid but unanticipated path). However, a parallel and covering heuristic process will be to show the end-user an example of a desired end-state and let the user decide if they are close or not. One of the keys here will be an extremely comprehensive UNDO that can back up both the end-users' and system's attempts to good re-starting places.
Key psychological ideas behind this interface come from Montessori (rich playful environment), Vygotsky (zone of proximal development), Bruner (using multiple mentalities to learn with), the teaching theories of two master teachers (Tim Gallwey and Betty Edwards), and a variety of psychological typing theories about learning and motivational styles derived from the literature and from our 30+ years of experience. One part of the learning that the UI does is to learn what kind of user is trying to use it. The range of human styles and motivations is considerable and taking advantage of them is critical. This is something that every good teacher does, but (amazingly) has not been put in any existing UI.
We will treat this as a "total user interface design problem with an important
AI part" rather than "a standalone AI with few social skills"—
i.e. if we take a smart and knowledgeable person (smarter than any
AI that can be made) we will still not have a good teacher, whereas
a good teacher knowing less (but enough) content can bring out the
RI (real intelligence) in the student.
OLPC B1 Prototype
photo by Mike McGregor
The One Laptop per Child (OLPC) Initiative is the perfect motivation
for initially doing a comprehensive version for third world children
whom for many, the computer will be the