 |
Description  |
|
|
TECHNICAL FIELD
This invention relates to computer based intelligent tutoring methods and
systems.
BACKGROUND ART
Many customers form their initial impression of service and product
providers through their telephone interactions with Customer Service
Representatives (CSRs). Indeed, for service providers in particular, CSRs
are called when a new customer wants services, or when an existing
customer has a problem or wants information or a change of service. As a
result, CSRs function as "ambassadors" to the service provider's customers
and their performance can have a direct impact on customer satisfaction as
well as market share.
The CSR's job is a complicated one in which they are expected to handle all
manner of customer calls regarding the provider's products and services.
This means that they must simultaneously carry on a consultative
conversation with the caller, manipulate service order and billing
software to find out information about the caller, enter information
regarding service registration, rapidly look-up information about
availability, compatibility and capabilities of the different products and
services from reference documentation, understand all of the features and
incompatibilities of the offered services, and at times prepare mailings
of information for the customer. These services and products, as well as
the information about them, are frequently updated, producing an ongoing
learning problem in order to "stay current". Because of this complexity,
it may take a year or more of training and on-the-job experience for CSRs
to become fully proficient.
At present, CSR training is directed predominantly to traditional learning
activities such as lectures and discussions rather than actual job
practice and training. In fact, trainees spend only about 1/4 of their
class time practicing their job using exercises such as role plays and
taking actual customer calls. However, both these techniques are less than
optimal. Since instructors can only observe and coach on role play at a
time, role plays are often done with little instructor interaction.
Moreover, role plays are often done without access to a phone and computer
terminal, the two essential components of the CSR's work environment. The
result is that role playing lacks realism, minimizing the ability to
prepare the trainee for the job that is to come.
Moreover, time spent by CSRs taking actual customer calls is structured so
that trainees take only one specific type of call during a session, such
as billing inquiries. As a result, while CSR trainees may receive dozens
of calls in an average work session, they will actually handle only a
limited number of those calls since calls other than a billing inquiry
will be transferred to a regular CSR.
Computer-assisted learning systems have been developed to address some of
the problems associated with traditional learning activities such as
lectures and discussions. A typical computer-assisted learning system is
illustrated in Haga et al. U.S. Pat. No. 5,211,563 ("the Haga '563
patent"). The system of the Haga '563 patent allows a trainee to access
teaching materials in computer storage through a central processor via
input and display devices. While such a system may free instructors to
concentrate on activities other than lecturing, it merely supports student
training and is unable to tutor or interact with the student as would a
traditional instructor.
As a result, computer-based training programs have also been developed that
deliver instructions to a student trainee. Computer-based training
programs, however, deliver such instructions staticly and uniformly. Thus,
while again freeing instructors from lecturing, computer-based training
programs still lack the dynamics associated with traditional instructors.
Therefore, a need exists for an intelligent tutoring system having
dynamically organized instructional programs that employ independent
representations of domain, instructional, and student knowledge enabling
it to provide individualized instruction much like that provided by a
personal human tutor. Such an intelligent tutoring system would provide
real time, context-appropriate and cost-effective training enabling
learners to perform appropriate domain tasks in the right manner and at
the proper time. In so doing, such an intelligent tutoring system would
decrease the time required to migrate learners from novice to expert,
while increasing the number of trained personnel successfully reaching a
more knowledgeable level.
An intelligent tutoring system would achieve these goals by dynamically
creating and revising individual instruction plans, actively teaching
difficult and abstract concepts and skills, guiding and assisting students
during exploratory learning in a simulated environment, and tailoring
training scenarios to the student's learning progress. More specifically,
such a system would apply state-of-the-art knowledge regarding artificial
intelligence, cognitive science, and multimedia to intelligently coach
trainees to perform the job of the CSR.
In such a system, trainees would exercise their customer interaction skills
by working through typical customer interactions in a tutoring environment
that simulates their actual working environment. Trainees would study
multimedia information, such as animations and video segments,
prerequisite to specific types of customer interaction skills. Instruction
would be trainee initiated but would also assess trainee performance and
use such assessments to make recommendations about what to study or
practice next, determine how to apply different instructional methods,
initiate. interventions during procedural training sessions, and provide
the trainee with performance feedback. Finally, such a system would also
allow instructional designers to adjust instructional and student modeling
parameters to further individualize the delivered instruction.
DISCLOSURE OF INVENTION
Accordingly, it is the principle object of the present invention to provide
an improved computer based intelligent tutoring method and system.
According to the present invention, then, a computer based method and
system are provided for tutoring a student in an interactive application.
The method of the present invention comprises selecting a mode for an
adjustable teaching parameter, generating a student model, and monitoring
a student interactive task based upon the teaching parameter and the
student model. The method further comprises generating an updated student
model based upon a student response to the student interactive task
generated, and monitoring a student interactive task based upon the
teaching parameter and the updated student model.
The computer based intelligent system of the present invention for tutoring
a student in an interactive application comprises means for selecting a
mode for an adjustable teaching parameter, means for generating a student
model, and means for monitoring a student interactive task based upon the
teaching parameter and the student model. The system further comprises
means for generating an updated student model based upon a student
response to the student interactive task generated, and means for
monitoring a student interactive task based upon the teaching parameter
and the updated student model.
These and other objects, features and advantages will be readily apparent
upon consideration of the following detailed description in conjunction
with the accompanying drawings.
BRIEF DESCRIPTION OF DRAWINGS
FIG. 1 is a traditional intelligent tutoring system architecture known in
the art;
FIG. 2 is a Voice Messaging Service task block diagram for the intelligent
tutoring method and system of the present invention;
FIG. 3 is a hierarchical diagram of the intelligent tutoring method and
system of the present invention;
FIG. 4 is a conceptual diagram of the intelligent tutoring method and
system of the present invention;
FIG. 5 is a top level function diagram of the intelligent tutoring method
and system of the present invention;
FIG. 6 is a subset of a student model representation for the intelligent
tutoring method and system of the present invention;
FIG. 7 is a knowledge base hierarchy diagram for the intelligent tutoring
method and system of the present invention;
FIG. 8 is a subset of a discourse grammar representation for the
intelligent tutoring method and system of the present invention;
FIG. 9 is an in-topic question probability graph for the intelligent
tutoring method and system of the present invention; and
FIG. 10 is an out-of-topic question probability graph for the intelligent
tutoring method and system of the present invention.
BEST MODE FOR CARRYING OUT THE INVENTION
The intelligent tutoring method and system of the present invention is a
domain-independent platform for teaching CSRs the procedures of their job,
including conversing with a customer and using order entry software. The
present invention involves working through either abstract or concrete
simulations of on-the-job scenarios and, when desired, reviewing related
declarative material. The method and system simulate both the conversation
and the order entry software, monitor the student's performance, provide
feedback on their performance, provide hints on expert responses during
problem solving sessions, and employ several strategies to ensure that
students are continually but not overly challenged, including tailoring
the style of instruction and choice of scenarios to the individual and
skimming over well known parts of a scenario during problem solving.
To work independent of the domain, the method and system of the present
invention maintain a strict division between the general tutoring
knowledge and the domain specific knowledge. Moreover, to employ
intelligent tutoring strategies, all of the structures of the knowledge
base are linked, thereby enabling a student's problem solving performance
to be integrated throughout the knowledge base.
Referring first to FIG. 1, a traditional intelligent tutoring system
architecture is shown, denoted generally by reference numeral 20. As seen
therein, an intelligent tutoring system employs a basic architecture
consisting of a domain model (22), an instructional model (24), a student
model (26), and a user interface (28). Domain model (22) comprises a
representation of the knowledge to be tutored to the student and is also
used as the standard for evaluating student performance. Instructional
model (24) comprises a representation of the knowledge of how to tutor the
student. More specifically, instructional model (24) includes
instructional methods to be employed in the tutoring system and how they
are employed. Student model (26) comprises a dynamic representation of the
student's state of knowledge. Finally, user interface (28) comprises a
communication channel between the tutor and the student.
The intelligent tutoring method and system of the present invention will be
described herein in conjunction with teaching CSRs how to operate, sell
and register customers for a residential Voice Messaging Service (VMS)
application. From this description, those portions of the intelligent
tutoring method and system of the present invention that are reusable for
other applications will be readily identifiable. A typical VMS answers
incoming calls placed to the subscriber when the called number is busy or
does not answer. VMS allows subscribers to record their own personal
greeting or use a prerecorded greeting. When messages are waiting, the
subscriber is alerted by a special "stutter" dial tone. Messages can then
be retrieved by calling a special access number using a push button phone
and entering a personalized security code.
Referring now to FIG. 2, a VMS task block diagram is shown, denoted
generally by reference numeral 30. As seen therein, all customer contacts
begin with the CSR soliciting (32) the caller's name and telephone number,
which is then entered into a service order software database via a
computer workstation to pull up the customer's account information.
Thereafter, an experienced CSR will engage in a short conversation with
the customer to determine (34) the nature of the call.
In the VMS domain, extensive task analyses have indicated that there are
essentially six types of customer calls. Calls from a customer who already
has VMS include those related to handling (36) a complaint about VMS,
disconnecting (38) a customer's VMS, changing (40) a customers VMS in some
way, and adjusting (42) a customer's VMS bill. Customer calls may also be
related to direct requests (44) for VMS, as well as non-VMS matters (46).
In the case of calls in which the customer is complaining about service or
billing, the CSR's task is to initiate the appropriate work order (48,50).
For calls where the customer is requesting some change to the service, the
CSR's task is more complicated. In such cases, the CSR must determine the
customer's reasons for wanting a change or removal of service (52,54),
discuss with the customer the implications of the changes (56),
potentially attempt to dissuade the customer from removal of the service
(58), and initiate the appropriate change or removal order (60).
Finally, in the case of calls that have nothing to do with VMS, the CSR
must seek an opportunity to discuss VMS capabilities and benefits with the
customer (62,64). Such a process requires that the CSR have skills in the
areas of evaluating customer information for VMS opportunities,
transitioning the conversation to VMS, explaining VMS benefits and
features, and responding to specific customer concerns. Once the caller
becomes interested in VMS, or in the case where callers making a direct
request for VMS, the CSR then must assess the capability of VMS with the
subscriber's other services and equipment (66), and finally register the
customer for VMS (68).
During this entire process, the CSR must simultaneously converse with the
customer to problemsolve and provide information, manipulate phone calls,
correctly enter data into service registration and customer billing
software while maneuvering through numerous screens and fields, understand
all features and. incompatibilities of an offered service, look-up
information about service availability and capabilities from reference
documentation located on their desks, and prepare mailings of VMS
information for the caller. Thus, an integral part of the CSR's job in
registering customers for VMS is interacting with service registration
software.
Referring next to FIG. 3, a hierarchical diagram of the present invention
is shown, denoted generally by reference numeral 70. As seen therein, the
present invention comprises three primary components: a tutor (72), an
authoring system (74), and a teaching parameters editor (76). Tutor (72)
teaches course materials as described in a knowledge base (78), stores and
uses a student's performance in a student model (80), and makes tutoring
decisions based on adjustable teaching parameters (82).
The authoring system (74) permits instructional designers and domain
experts to create and edit the knowledge base which, as described in
greater detail below, contains information such as the course topics, the
conversation discourse grammar, the discourse rules, audio and textual
forms of the conversation, the application description, and the
application commands. As will also be described in greater detail below,
the teaching parameters editor (76) allows instructors and instructional
designers to modify various instructional behaviors, such as how to
recommend topics and conversations, how skimming and scaffolding during
practice occur, and how factors of performance are weighed in student
modeling. Thus, the functions of the present invention can be broken down
into two major groupings: (i) tutor functions usable by CSR trainees; and
(ii) author/editor functions usable by instructional designers and domain
experts.
As again will be described in greater detail below, tutor (72) itself
comprises study and exercise functions. Using the study function, trainees
can study lessons presented in multimedia format on information
prerequisite to the domain of study. Using the exercise function, trainees
are presented with a number of different methods for working through
typical CSR tasks or problems in an environment that closely emulates
their actual work environment. Similarly, authoring system (74) and
teaching parameters editor (76) themselves comprise conversation knowledge
and instructional strategy functions, respectively, designed to support
the instructions delivered by tutor (72). The conversation knowledge
function enables instructional designers and domain experts to build
knowledge bases of conversations for use by the exercise function of the
tutor (72). The instructional strategies function enables instructional
designers to adjust the parameters which the tutor (72) uses to make it
tutoring decisions.
Referring now to FIG. 4, a conceptual diagram of the method and system of
the present invention is shown, denoted generally by reference numeral 90.
As seen therein, the method and system of the present invention comprise
five-high level functions. First, a study function (92) supports trainees
studying interactive multimedia lessons on information prerequisite to
VMS. Second, an exercise function (94) supports trainees working through
typical conversations where the goal is to handle some customer request
regarding VMS, or transition a non-VMS call to a VMS opportunity. The
method and system further comprise a student model (96), which dynamically
represents the trainee's performance including all activities performed or
observed by the trainee and the time required to perform them.
Still further, the method and system comprise a coach (98) which guides
instructional activities using input from the student model (96), as well
as domain and instructional knowledge. More specifically, the coach (98)
provides an introspective evaluation of the trainee's skill level for a
number of topics on a five point scale which are preferably denominated as
"untried", "needs practice", "almost", "good", and "excellent". Coach (98)
also makes recommendations about what to study or practice. Ultimately,
however, instruction is trainee-initiated. That is, the trainee selects
both the mode and the topic to study or practice. Coach (98) further
applies various instructional strategies during a contact rehearsal based
on the individual's trainee assessment, provides instructional assistance
during contact rehearsal, and provides trainee performance feedback at the
end of a session.
Finally, the method and system of the present invention also comprise a
user interface (100) which controls the interaction with the student in as
natural a manner as possible by creating an environment that emulates the
trainee's actual work environment in both appearance and behavior. As will
be described in greater detail below, the method and system of the present
invention are also supported by conversation knowledge and lesson files
library databases (102,104).
In such a fashion, the method and system of the present invention are
consistent with several features of the minimalist approach to training
and learning. Specifically, the present invention employs task-based
training, and allow trainees to start immediately on meaningful and
realistic job tasks in any order. Moreover, the present invention keeps
the amount of passive instruction to a minimum. Only prerequisite
information that cannot be conveyed to the trainee during active contact
rehearsal is conveyed through guide (92). In addition, information in
guide (92) is presented in interactive, multimedia formats to increase the
level of involvement by the trainee. Finally, the present invention
contains explicit training on errors and error recovery to support the
recognition and recovery from error, thereby making the learning materials
more robust and complete, and training learners in error recovery skills.
Referring next to FIG. 5, the top level functions of the intelligent
tutoring method and system of the present invention are shown, denoted
generally by reference numeral 110. As seen therein, those functions
available from a tutor top level (112) comprise a study the guide function
(114), a rehearse conversations function (116), and an examine contact
flow function (118). Guide (114) supports the delivery of study materials
in interactive multimedia formats on topics identified as prerequisite to
operating, selling and registering customers for VMS. Guide (114) operates
as an interactive, multimedia "magic book" in which students can browse
and study information prerequisite to a topic in a variety of different
media and forms including text, digitized audio, graphics, photographs,
animations and digitized video.
Each piece of multimedia is contained in separate files. Audio segments are
recorded and stored on hard disk using a hierarchical file structure that
maps onto either chapters in the guide (114) or conversations, which will
be described in greater detail below. Video segments are recorded and
transferred onto laser disks from which they are digitized. These
digitized segments are also stored on hard disk using a hierarchical file
structure that maps onto either chapters in the guide (114) or
conversations, which will again be described in greater detail below.
Lessons in the guide (114) have been produced to a grain size that supports
entry into the entire lesson and topic level, or entry into parts of the
lesson as they apply to parts of conversations. This second capability is
particularly useful in enabling trainees to review information from part
of a lesson in the guide (114) to support their conversation rehearsals
during tutoring.
Referring now to FIGS. 4 and 5, the function of exercise customer
interaction skills (94) employs rehearsing conversations (116) or
examining contact flow (118) as two different ways of rehearsing customer
contacts. During rehearse conversation (116), tutor (112) supports
trainees working through customer contacts representative of those faced
on the job. As trainees work through the contacts, immediate feedback and
hints (120) are available when trainees have difficulties. Moreover, at
the end of each contact rehearsal, trainees are provided summary feedback
(122) including assessment of their current knowledge state.
Rehearse conversations (116) is the most complex part of the present
invention's instructional process. The rehearse conversation (116)
instructional environment and philosophy are similar to that employed in
known intelligent tutoring systems which seek to produce a situated
learning environment in which trainees can acquire and exercise their
knowledge and skills. However, the intelligent tutoring method and system
of the present invention goes beyond known intelligent tutoring systems by
providing proactive tutoring in the domains for which it was developed.
In that regard, in the present invention, an apprenticeship approach to
teaching the skills and knowledge necessary for competent domain
performance is used as a model of proactive tutoring while rehearsing
customer contacts. The apprenticeship approach involves using methods like
observation, coaching and successive approximation rather than didactic
teaching. In addition, apprenticeship approaches embed the learning skills
and knowledge in their social and functional contacts.
CSR-customer conversations correspond to exercises in conventional textbook
instruction. The complexity associated with their rehearsal arises from
the desire that trainee's work at the edge of their confidence in the
context of phone conversations. Working at the edge of confidence means
practicing only the task that is the current focus of attention while not
redoing those tasks already known, nor doing those tasks not yet known,
even though such tasks may arise. naturally during the course of a
conversation. In that regard, the present invention utilizes a top down
approach to instruction, putting more general before more specific
objectives, and global before local skills. Thus, the present invention is
capable of showing a global situation (the whole conversation) without
requiring that the trainee perform the whole conversation.
More specifically, CSR-customer contacts in the present invention have two
major components. One component is a dialog interaction between the
customer and the CSR, while another component is an application
interaction between the CSR and some service order software. Thus, a
contact can be characterized as a set of situation-action rules. Typical
situations may include verbal situations such as customer statements,
requests or questions, as well as operational situations such as service
order software output or configurations. Typical actions may include
verbal actions such as responses by the CSR to customer statements,
requests or questions, or to service order software information,
operational actions such as commands or data entered into the service
order software by the CSR, as well as cognitive actions such as those
focused around information gathering, information processing, and decision
making.
As an example of how a contact is rehearsed, contacts begin with a customer
call which the trainee hears over the same headset they use on the job.
The trainee initiates an action in response to the situation which, when
correct, prompts another contact situation an so on until the conv | | |
