Transcript lect3_classicsystems..

Lecture 3
Classic Intelligent Learning
Tools
26-Mar-16
Adaptive Learning Environments
1
What is an ALE?
“Tutoring and training systems which mimic tasks
traditionally done by teachers”
A fairly broad definition, to include:
Intelligent Learning Environments,
Intelligent Tutoring System
Adaptive Learning Environments,
Intelligent Computer Assisted Instruction,
and other intelligent interactive teaching/learning tools...
Blur distinction between tutoring and training:
use the terms “tutoring”, “teaching”, “education”
and “training” interchangably....
26-Mar-16
Adaptive Learning Environments
2
Hartley’s (1973) Framework
Representation of the knowledge of skill that is to be taught = Domain
Knowledge
eg the ability to solve problems in the domain, to judge/comment on
student's answer, or answer questions posed by the student.
Teaching actions = Teaching Strategies and Tactics
eg making positive/negative comments, providing examples, setting
problems, asking the student to explain, find counter-examples, step
through examples
Model of the student's current state = Student Model
ie history, capabilities, knowledge, beliefs, goals and motivation
Interface and Communication
eg discourse between student and system, choice of interface - WIMP,
graphics, text, speech, VR
26-Mar-16
Adaptive Learning Environments
3
1. Learning through
model building - the
LOGO experience
26-Mar-16
Adaptive Learning Environments
4
A distinction…
A. Simulations, Exploratory Environments,
Micro-worlds
= the investigation of views of a given
domain, which may differ from the learner's
own
B. Expressive or Modelling Environments
= the modelling of user's own beliefs, and
reflecting on and exploring own models
26-Mar-16
Adaptive Learning Environments
5
Learning through Model Building
Assumes learner is active and seeking stimulation
Making knowledge explicit:
• get learner to communicate beliefs
• get learner to model theories and test them
• get learner to reflect on learning
Learning through confrontation:
• student has belief of what happens in environment
• tests belief -> consequences in environment
• if consequences don't match belief, then (hope)
• cause student to review belief (=learn)
But, are all confrontations beneficial?
26-Mar-16
Adaptive Learning Environments
6
Model Building in LOGO
Allows student to explore their own models by building
programs written in LOGO, to improve e.g. programming,
maths, language, physics
Feurzeig and Papert (1969)
- Start with body awareness and relate body movements to
those of a small robot.
- Control the robot (turtle) with commands - see effect of
drawing device attached.
- If does not do what expect, try to act out what it does do
and correct it.
Then child uses LOGO to describe procedures and learns
problem-solving skills such as problem decomposition.
Aim is to then generalise to other similar problems
26-Mar-16
Adaptive Learning Environments
7
Logo Turtle Robot
3/26/2016
Adaptive Learning Environments
8
Examples of LOGO code:
FD 10
LEFT 90
BD 20
ADD 3 2
SUB 5 1
DIVIDE 6 2
REPEAT 4 (FD 10 LEFT 90)
RIGHT 50
MULTIPLY 7 3
Procedures:
DOODLE 'SIZE
1. FD :SIZE
2. LEFT 90
3. FD ADD (10 :SIZE)
4. RIGHT 120
5. BD MULTIPLY (:SIZE 5)
Write procedures to draw:
- equilateral triangle
- pentagon
- polygon of any number of sides
26-Mar-16
Adaptive Learning Environments
9
Examples of LOGO code:
TRIANGLE 'SIZE
1. REPEAT 3 (FD :SIZE LEFT 120)
PENTAGON 'SIZE
1. REPEAT 5 (FD :SIZE LEFT 72)
POLYGON 'SIZE 'N
1. REPEAT :N (FD :SIZE LEFT DIVIDE 360 :N)
26-Mar-16
Adaptive Learning Environments
10
LOGO pros/cons
PAPERT (MIT):
* if integrate in existing class this may prejudice any success
* take whole new approach to curriculum with it
* only need student and environment
v. EDINBURGH:
* need for some structure to support it - cannot just say
"explore, learn!"
* need additional guidance too
Evaluation:
- syntax of LOGO - language may be harder than domain to
be taught
- no clear evidence of generalisable problem solving skills
through LOGO
+ some evidence of improved learning in the domains where
LOGO used
26-Mar-16
Adaptive Learning Environments
11
2. Modelling Knowledge
SOPHIE
(Brown et al., 1975, 1976, 1982)
26-Mar-16
Adaptive Learning Environments
12
Encoding Domain Knowledge
1.
•
•
Black Box Models:
Reason about domain without actually coding knowledge e.g.
mathematical models
Generate the correct input-output behaviour over range of tasks; can
use as a judge of correctness.
BUT means by which provides behaviour not available for inspection
or use.
2. Glass Box Expert Systems:
Develop an expert system
•
Knowledge Acquisition from human, code and apply
•
May not correspond to way human applies it
3. Cognitive Models:
Make a simulation (at some level of abstraction) of how humans
uses knowledge
26-Mar-16
Adaptive Learning Environments
13
Black Box Models:
Sophie (Brown et al., 1982)
Students troubleshooting electronic circuits:
try out ideas, have them critiqued, receive advice
Fault inserted for student to debug.
The system can:
 answer hypothetical questions (what if?)
 evaluate hypotheses (consistent?)
 list all possible hypotheses (generate and test)
 evaluate a new measurement (redundant?)
Representations:
• simulation based mathematical model
• procedural specialists that use model
• semantic net of declarative facts
26-Mar-16
Adaptive Learning Environments
14
SOPHIE I
Performs correct behaviour, but….
Mathematical simulation model - so means by which
behaviour is produced is not available for inspection
Does not have student model, nor explicit knowledge of
troubleshooting
Makes no use of kind of causal reasoning performed by
expert troubleshooters
26-Mar-16
Adaptive Learning Environments
15
Illustration of Sophie (Brown et al., ‘75)
Student: What could be wrong?
SOPHIE:Here are some faults
which
measurements you observed:
the beta of Q3 is low
the beta of Q4 is low
or R22 has the wrong resistance
explain
the
Student: What would the output voltage be if the beta of Q3 were 8?
SOPHIE: If the beta of Q3 were 8 under the present
setting, the output voltage would be 10.5 volts.
In a working circuit it is 19.9 volts.
Student: Is the emitter current of Q4 correct?
SOPHIE:The emitter current of Q4 is -61 milliamps.
In a working circuit the emitter current of Q4
is -49 milliamps. That value seems to be
incorrect.
Student: Could R22 be low?
SOPHIE: Is it OK if
resistance of R22?
I
use
25
ohms
for
the
Student: Yes
SOPHIE:That fault agrees with all the measurements
you have made.
26-Mar-16
Adaptive Learning Environments
16
Towards a Glass Box Model
SOPHIE II - an articulate expert - 1976
Not qualitative simulation, but improvement:
• Troubleshooting expert added to lab.
• Able to demonstrate strategies, though cannot reason
about student strategies
Troubleshooting expert reasons qualitatively:
 makes causally meaningful measurements
 explains strategic decisions as proceeds
 uses decision tree, pre-stored explanations attached to
nodes
So we can start with a simulation, and then add further
knowledge and reasoning on top of it
26-Mar-16
Adaptive Learning Environments
17
SOPHIE II excerpt (Brown et al, 1976)
SOPHIE-II: How does the OUTPUT CURRENT of the CONSTANT CURRENT SOURCE
compare with a working circuit?
Student:
OK
SOPHIE-II: The current source is working then. The modules now left are
CL,VL,OPF and DARL. I think I will make my next measurement on the
Darlington amplifier. I want to know whether the BASE CURRENT of Q4
is OK.
Student:
HELP
Referee: Let’s use SOPHIE to find the answer.
>> Is the base current of Q4 OK?
SOPHIE-I: The base current of Q4 is 0.0 amps, which seems to be
incorrect because in a working circuit it is .031 milliamps.
Referee: So, you see, the BASE CURRENT of Q4 is LOW in the
faulted circuit. Let’s go on….
SOPHIE-II: Low input current to the Darlington amplifier would
certainly result in lower than normal voltage at its output under
moderate current loads. I think then, the Darlington amplifier is
OK, since it is following its input. Now, since we have already
seen that the CSS output current is OK, perhaps either the current
limiter or the voltage limiter must be drawing excessive current
26-Mar-16
Adaptive Learning Environments
18
from
the CSS.
3. Modelling Knowledge
Guidon – a Glass Box Model
(Clancey, 1979, 1982, 1983)
26-Mar-16
Adaptive Learning Environments
19
Glass Box Models: Guidon
General Method is Knowledge Acquisition:
1. knowledge engineer + domain expert
2. identify problem area and scope, formalise key concepts
3. formulate system to implement K, test and refine
iteratively
GUIDON: tutor built on top of MYCIN expert system, which
diagnoses bacterial infections (Shortliffe, 1976)
- uses MYCIN rule base + extra annotations, e.g. canned
text explaining how some lab test is performed.
- filter out medically irrelevant parts of a rule
Can provide articulate, human-like representation of
knowledge underlying expertise in domain,
BUT time consuming and may not match novice learners'
domain knowledge....
26-Mar-16
Adaptive Learning Environments
20
Guidon: mixed initiative dialogue
1.
2.
Case selected, initial data presented to student
Student acts as diagnostician, asks questions, gathers
further data, proposes hypotheses and requests help
3. Student's behaviour compared with MYCIN's
4. If student actions suboptimal, or asks for help, GUIDON
intervenes
5. Guidon guides student's reasoning in solving case,
using MYCIN's rules as 'issues’
Dialogue of intervention guided by tutoring (T) rules, uses
MYCIN's reasoning in explanations
- Separates domain and pedagogical knowledge (c.200
rules)
Socratic tutoring style response to student hypothesis
Attaches certainty factor values to domain rule: student
‘seems to know’, ‘is able to apply’,‘has actually used’ rule
26-Mar-16
Adaptive Learning Environments
21
Excerpt from dialogue
Guidon: You should now proceed to ask questions that will enable you to make
a hypothesis about the organisms (other than those seen on cultures or
smears) which might be causing the infection.
{Student has initial data to go on at this point}
** HOW IS THE FACT THAT THE PATIENT IS BURNED RELEVANT?
Guidon: The fact that the patient is seriously burned will be used to
conclude that Pseudomonas-Aeruginosa is one of the organisms which might
be present (in addition to those seen on cultures or smears) - Rule 578
** HELP
Guidon: Try to determine the type of the infection: bacterial, fungal, viral
or TB.
** WHAT IS THE PATIENT'S WBC?
Guidon: The white bloodcell count from the patient's peripheral CBC is 1.9
thousand.You should now proceed to ask questions that will enable you
to make a hypothesis about the organisms(other than those seen on
cultures or smears)which might be causing the infection.
{Student has initial data to go on at this point}
** HOW IS THE FACT THAT THE PATIENT IS BURNED RELEVANT?
Guidon: The fact that the patient is seriously burned will be used to
conclude that Pseudomonas-Aeruginosa is one of the organisms which might
be present (in addition to those seen on cultures or smears) Rule 578
** HELP
Guidon: Try to determine the type of the infection: bacterial, fungal, viral
or TB.
** WHAT IS THE PATIENT'S WBC?
Guidon: The white bloodcell count from the patient's peripheral CBC is 1.9
thousand.
26-Mar-16
Adaptive Learning Environments
22
Consider domain rule 578:
IF
(1)
(2)
(3)
(4)
the infection needing therapy is meningitis
organisms were not seen on the stain of the culture
the type of the infection is bacterial
the patient has been seriously burned
THEN
Pseudomonas Aeruginosa might be one of the
organisms (other than those seen on cultures or
smears) causing the infection (0.5)
26-Mar-16
Adaptive Learning Environments
23
What is happening in this extract?
Rule 578 applies to the patient in question,
Student asks for help,
GUIDON:
• chooses a rule to discuss (one just mentioned)
• chooses a way to present it - tells student to work on
untackled subgoals in the rule (type of infection)
This becomes the next topic
Next question relevant so GUIDON does not intervene.
26-Mar-16
Adaptive Learning Environments
24
Incorporating Teaching Knowledge
T-rules in GUIDON:
A: guiding discussion of a goal
* the student asked about the relevance of burning: became the focus
topic, following t-rule succeeded
IF
the recent context of the dialogue mentioned
either a deeper subgoal or a factor
relevant to
the current goal
THEN define the focus rule to be the domain rule that
mentions this topic.
* domain rule 578 became focus of the dialogue.
* If no obvious choice of domain rule to suggest, uses t-rules which assess
interestingness:
26-Mar-16
Adaptive Learning Environments
25
Some rules covering the student’s
hypothesis (after Clancey, 1979)
26-Mar-16
Adaptive Learning Environments
26
An example of a tutorial rule for student
modeling (Clancey, 1979)
T-RULE 6.05
IF (1) The student’s hypothesis does include values
that can be concluded by this domain rule, as well
as others, and
(2) The hypothesis does not include values that can
only be concluded by this domain rule, and
(3) Some other values concluded by this domain
rule are missing from the hypothesis
THEN Define the belief that the domain rule was
considered by the student to be -0.70.
26-Mar-16
Adaptive Learning Environments
27
Student Modelling
Overlay Models: Pros and Cons
Advantages:
+ easy to implement
+ widely used
+ can be effective
Disadvantages:
- student may follow different problem solving approach
- may hold different beliefs that are not a subset of the
domain knowledge (e.g. misconceptions)
- cannot predict what user knows based on partial
knowledge
- does not represent order new information learned in
26-Mar-16
Adaptive Learning Environments
28
Knowledge Modelling
Some Other Problems:
Suitable domain for expert system does not imply suitable domain for
ITS: focus on EXPERTISE not on LEARNER (student as subset of
knowledge base)
Rules may encode expert knowledge but control stucture/ reasoning
strategy not same:
• forces MYCIN's top-down strategy on user;
• can reject user's reasonable hypothesis
Also, rules may be too complex for novices:
• hard to understand/remember/make sense of;
• no distinction between different types of conditions,
e.g. which most critical/or easiest to test or eliminate
Cost-effective Expert System may not be effective for ITS
26-Mar-16
Adaptive Learning Environments
29
e.g. rule 507
IF
(1) the infection needing therapy is meningitis
(2) organisms were not seen on the stain of the culture
(3) the type of the infection is bacterial
(4) the patient does not have a head injury, and
(5) the age of the patient is between 15 and 55 years
THEN the organisms that might be causing the infection are
diplococcus-pneumoniae(.75) and neisseriameningitidis(.74)
Mixes test data, age, strategic knowledge, meta
interpreter knowledge, initial data
26-Mar-16
Adaptive Learning Environments
30
Neomycin and Guidon 2
Collected protocols of experts' diagnosis and teachers
articulating reasoning: teachers' explanations more
general, not specific to medical domain
Re-configured MYCIN to get explicit model of "diagnostic
thinking”
• separation of strategic K from domain facts and rules
• metarules representing hierarchical reasoning strategy +
notion of hypotheses
Changed some ordering of conditions in rules
Organised rules into types of information: general
principles; common world realities; definitional and
taxonomic relations; causal relations; heuristic rules.
Wider range of diseases covered; decrease in number of
questions; justifications and explanations in terms of
strategic goals and r.e. specific hypotheses
26-Mar-16
Adaptive Learning Environments
31
References(also see course webpage):
Brown, J.S. and R.R.Burton, (1978) Diagnostic models for procedural bugs in
basic mathematical skills, Cognitive Science, 2, pp.155-192
Brown, J.S. & VanLehn, K. (1980). Repair theory: A generative theory of bugs in
procedural skills. Cognitive Science, 4, 379-426.
Burton, R.R. (1982) Diagnosing bugs in a simple procedural skill, in (eds.)
D.Sleeman and J.S.Brown, Intelligent Tutoring Systems, Academic Press,
pp.157-184.
Clancey, W.J. (1983) GUIDON, Journal of Computer Based Instruction, 10:(1+2)
8-15.
Clancey,W. (1986) Qualitative Student Models', in First Annual Review of
Computer Science, ACM, pp. 381-450.
Shortliffe, E.H. (1976) Computer-Based Medical Consultations: MYCIN. New
York: American Elsevier.
VanLehn,K. (1987) Learning one sub-procedure per lesson, Artificial Intelligence,
31, 1, pp.1-40.
Wenger, E. (1987) Artificial intelligence and tutoring systems: computational and
cognitive approaches to the communication of knowledge. San Francisco:
Morgan Kaufmann.
26-Mar-16
Adaptive Learning Environments
32