Presentation is loading. Please wait.

Presentation is loading. Please wait.

IT803 Mixed-Initiative Intelligent Systems CIRCSIM – Tutor Presented by Bernard Yung Principal Investigators Martha W. Evens, Illinois Institute of TechnologyMartha.

Similar presentations


Presentation on theme: "IT803 Mixed-Initiative Intelligent Systems CIRCSIM – Tutor Presented by Bernard Yung Principal Investigators Martha W. Evens, Illinois Institute of TechnologyMartha."— Presentation transcript:

1 IT803 Mixed-Initiative Intelligent Systems CIRCSIM – Tutor Presented by Bernard Yung Principal Investigators Martha W. Evens, Illinois Institute of TechnologyMartha W. Evens Joel A. Michael, Rush College of MedicineJoel A. Michael Allen A. Rovick, Rush College of MedicineAllen A. Rovick Other Investigators and Alumni Stefan Brandle, Taylor UniversityStefan Brandle Byung-In Cho Korea Institute for Defense AnalysisByung-In Cho Reva Freedman, Northern Illinois University ***Reva Freedman Michael Glass University of Illinois at ChicagoMichael Glass Injoo Jeong East-West UniversityInjoo Jeong Jung Hee Kim North Carolina A & TJung Hee Kim Ramzan Ali Khuwaja,Ramzan Ali Khuwaja Gregory A. Sanders, NISTGregory A. Sanders Farhana Shah, ***Farhana Shah Chong Woo Woo, Kookmin UniversityChong Woo Woo Feng-Jen Yang Rowan UniversityFeng-Jen Yang Yujian Zhou WebEx Corp.Yujian Zhou ***main source of this presentation

2 Cardiovascular Physiology Overview of CIRCSIM-Tutor Mixed-Initiative Issues in CIRCSIM Architecture Control Task Awareness Communication Evaluation. Summary Lesson Learned References Presentation Outline

3 Fig Baroreceptor reflexes that help to maintain blood pressure homeostasis Human Anatomy & Physiology by Marieb, E. p733

4 Shah, 1997, p. 20 Evolution of CIRCSIM-Tutor

5 Domain: baroreceptor reflex - the negative feedback system that controls blood pressure in the human body Procedure: CIRCSIM describes a perturbation of the cardiovascular system to the students CIRCSIM then asks them to predict the qualitative changes in the most important parameters (tell whether they will go up or down or stay the same). CIRCSIM then launches a tutorial dialogue to help the student correct any mistakes or misconceptions. CIRCSIM-TUTOR

6 Sample Screen shot of CIRCSIM-Tutor.

7 Developed by the Dr. Rovick and Michael [Khuwaja et al., 1992]. Depicts the causal relationship between parameters involved in regulating the blood pressure. RAP - Right Atrial Pressure SV - Stroke Volume CC - Cardiac Contractility HR - Heart Rate MAP - Mean Arterial Pressure CO - Cardiac Output, TPR - Total Peripheral Resistance BR - Baroreceptor Reflex CNS - Central Nervous System Top Concept Map – 7 physiological variables

8 7 physiological variables Top level Concept Map Shah, Farhana. Recognizing and Responding to Student Plans in an Intelligent Tutoring System: CIRCSIM-Tutor Ph.D., Illinois Institute of Technology, 1997 CO = SV * HR

9 Shah, 1997, p31 CIRCSIM ARCHITECTURE

10 Shah, 1997, p.22 Prediction Table

11 Domain Knowledge Base Problem Solver Student Modeler Instructional Planner Input Understander Judger Screen Manager Controller Discourse Planner Text Realization Component CIRCSIM-Tutor (v. 3) modules:

12 Information Store Domain Knowledge Base Concept Map – contain causal relationships between various components of the CV system. Instructional Planner may switch system to use the two deeper levels of the concept map. Student model is consulted heavily by the instructional planner and the discourse planner. Lexicon Tutoring History Curriculum Discourse History Journal File

13 Acts as the actual means of communication between the system and the student. Displays and echoes the responses of the tutor and student Displays text and graphics on the computer screen for the student Keeps track of the variables and the time that the student and system take in responding Receives the student’s input via the keyboard and passes it to the Input Understander. CIRCSIM-Tutor Screen Manager

14 Produces an internal representation (logical form) from the student’s natural language responses using a lexical functional grammar as a basis for the parser. Has to deal with incomplete sentences such as noun phrase, or an adjective, or an adverb, or some other fragment, and spelling errors Uses a bottom-up parsing approach Uses Information Stores: Discourse History and the Lexicon CIRCSIM-Tutor Input Understander

15 The Instructional Planner is the central component of this architecture. Decides what, when, and how to teach the student.  what subject matter to focus on  how to convey knowledge/instruction to the student  when to intervene when the student is busy problem-solving Determines the course of action at each point during a tutoring session. Builds a tutoring history Measures how often the parser/understander fails to understand student input. Plans the curriculum and teach the selected goals Answers student questions/responds to other student initiatives CIRCSIM-Tutor Architecture Instructional Planner

16 CIRCSIM-Tutor Architecture Instructional Planner Interacts with the Input Understander, the Text Generator, the Student Modeler, and the Screen Manager in order to carry out the tutorial activities Uses domain knowledge, student model, and pedagogic knowledge in making decisions. The Instructional Planner that carries out the mixed initiative strategy. The system controller coordinates actions/ applies plans. It allows access to any procedure at any stage, whether direct response (DR), reflex response (RR) or steady state (SS).

17 Evaluates the response from the student encoded as a logical form, and declares whether it is correct or not by examining the response, the question, and the correct answer provided by the Problem Solver. Stores the information about the student reply in the student model. CIRCSIM-Tutor The Judger

18 Student modeler builds the student model containing the basic knowledge about the student and the model describes the tutor’s perception of what the student does or does not know about cardiovascular physiology. The student modeler uses two information stores: the Student Model and the Domain Knowledge Base. Calls the Domain Problem Solver. Interprets the student’s cognitive performance, and records or updates it in a student model that describes the current stage of the students’ knowledge from the tutor’s point of view. Interprets the student’s errors as well as correct responses. CIRCSIM-Tutor Architecture Student modeler

19 CIRCSIM-Tutor Architecture Student modeler Two major approaches used for student modeling: Overlay model [Carr and Goldstein, 1977] - designed to represent the student’s knowledge state as a subset of an expert’s knowledge state. Buggy model [Brown and Burton, 1978] - uses Bug libraries to determine common student misconceptions to be fixed. CIRCSIM-Tutor uses a unified approach.

20 Components: the main problem solver and the subproblem solver. When the screen manager takes qualitative predictions from the prediction table and sends them to the planner, it is the problem solver that generates correct predictions for all parameters in the prediction table and gives a solution path to be used to monitor the student’s problem solving approach. The subproblem solver is used for solving problems from any other module of the system or problems raised by a student query. CIRCSIM-Tutor Problem Solver

21 Translating the plan formulated by the discourse planner into natural language sentence or sequence of sentences. For example, (question (affected-by SV ?))  “What are the determinants of SV?” Two parts: the Discourse Planner and the Surface Generator. Discourse Planner Keeps track of the discourse content and conveys a discourse plan to the Turn Planner. Uses the Lexicon, the Discourse History, the Domain Knowledge Base, the Domain Problem Solver and the Student Model. CIRCSIM-Tutor Text Generator

22 CIRCSIM-Tutor Text Generator Turn Planner organizes sequence of subplans into a plan for a content turn. Surface Level Text Generator takes the resulting plan and turns it into a sequence of English sentences. Ramachandran and Evens [1995] introduced a User-Driven Lexical Choice, in which the user vocabulary is used as the basis for text generation. Uses both the Lexicon and the Discourse History.

23 Control Issue CIRCSIM is designed to keep control of the conversation, to keep the initiative throughout. Tutor asks a question to start the problem solving process A: If student answers the question correctly B: then continue with a question about the next step else when answer is wrong system provides a hint and asks a follow-up question give a brief explanation and ask another follow-up question provide answer (go to B) What if the student takes the initiative: asks a question, proposes an explanation, or asks for confirmation?

24 Control Issue CIRCSIM should be able to handle student initiatives therefore strategies for recognition and response is needed. Source: human tutoring sessions Problem: How would you distinguish student initiatives from student’s answer to questions? Method: Analyse transcripts and identify student initiatives Classify student initiatives Classify tutor responses

25 A student initiative occurs when a student takes control of the tutoring session temporarily by saying something that forces the tutor to change the course of action and respond to the new situation; A student initiative is any attempt by the student to seize control for changing the course of the dialogue Definition Student Initiative

26 Student Initiatives classified in 4 dimensions (shah, p.106)

27 Student Initiatives Surface Forms Interrogative. An interrogative form is a sentence mainly recognized by word order or wh-element. K4-st-76-3: Would it be affected in a person who was not on an artificial pacemaker? Declarative. A declarative form makes a statement. The statement may reflect an assertion or state ideas. K4-st-84-1: I don’t think I understand the question.

28 Imperative. An imperative form makes a request or proposes a theory. Like the declarative form the imperative sentence is usually followed by a period. Very strong requests may be given an exclamation point. K11-si-18-1: Let me restate my question. Sometimes students hedge even an imperative with a question mark. K22-st-89-1: Let me start somewhere else? Other. These are the fragment forms comprising one word or more. We classify here strings that do not contain the syntactic structure of a sentence. K28-st-53-1: No Student Initiatives Surface Forms

29 Student Initiatives Surface Forms Silence/Pause. K7-tu-91-3: What happens in the reflex next. [big pause here] K7-ti-92-1: Need help? K7-st-93-1: Y Meaning can be associated with a pause:  student needs time to think over the task  student gets engaged in some other activity  student wants to disengage from the session  student gets confused or has some comprehension problem  student faces difficulty in reaching some information for expression  student is hesitant to answer

30 Student Initiative Communicative Goal/Intention. Requests for Information and Confirmation are the Most Common Categories. Request for Information. This plan/goal produces a direct interpretation of the initiative as a request that the tutor inform the student about the topic in focus. This can be satisfied by explaining the topic or releasing the information appropriate to the student's current goal. K2-st-49-1: I think I would like to further discuss the idea of RAP. K2-st-49-2: Unless compliance is involved, I still do not understand why the pressure in the right atrium decrease with an increase in right atrial filling.

31 Student Initiative Communicative Goal/Intention. Request for Confirmation. The student generates an explanation and asks for confirmation of this theory. Sometimes a simple yes or no is a sufficient response. K9-st-38-1: i.e. the change in sympathetic input changes the location of the Starling curve? K9-tu-39-1: Yes. More often the tutor responds more elaborately, especially when the student's explanation is wrong.

32 Student Initiative Communicative Goal/Intention. Conversational Repair. Both the tutor and the student express their thoughts in a way that is not always perfect or clear. The repair initiative is often a request for clarification or it may be a request for rephrasing or correction. K4-tu-83-6: How are the falls in TPR and in CC connected to the decrease in MAP? K4-st-84-1: I don't think I understand the question. <-- repair req K4-tu-85-1: What are the determinants of MAP? <--restate K4-st-86-1: MAP is determined by TPR and CO, so if the TPR is decreasing then the CO is decreasing too, given the act that CC is also decreased.

33 Student Initiative Communicative Goal/Intention. Inability to Answer. Sometimes the student does not know the answer and utters an explicit statement of her/his inability to give the answer (or refuses to answer, and just gives up). This initiative appears to be taken as a kind of giving up participating in the game, yet the student is obliged to utter something following the rules of the game. The tutor responds in the form of an explanation on the topic. K1-tu-56-3: What other neurally controlled structure is affected by the reflex and how? K1-st-57-1: Cardiac muscle, CC i. K1-tu-61-1: Think again sympathetic firing is being decreased. K1-st-62-1: I don't know. --con’t--

34 Student Initiative Communicative Goal/Intention. Inability to Answer (cont) K1-tu-60-1: When MAP goes up it increases basroceptor nerve impulse input to the CV centers. K1-tu-60-2: Sympathetic output TO ALL OF THE CV EFFECTORES is inversely related to the afferent input rate. K1-tu-60-3: Parasympathetic output to the pacemaker is direclty related to the input afferent rate. {PAUSE} K1-tu-60-4: Still stuck?

35 Student Initiative Communicative Goal/Intention. Challenge. This kind of response reflects some sort of disagreement with what the tutor has said. The use of clue words like "but", especially at the start of the sentence, often indicates that the initiative is taken as a challenge to the tutor's preceding utterance. This act may happen as a result of not accepting the truth of the tutor's previous statement completely. K20-tu-46-2: But you forgot that the real pacemaker is dead and this guy's HR is determined by the broken artificial pacemaker. K20-st-47-2: OOPS. (cont)

36 Student Initiative Communicative Goal/Intention. Challenge (cont) K20-st-47-2: BUT I WAS JUST READING EARLIER TODAY IN SMITH AND KAMPINE ABOUT HOW SANS CAN 'TURN ON' OTHER AREAS AND INFLUENCE HR WITHOUT ACTING FIRST ON THE SA NODE K20-tu-48-1: It happens sometimes (extopic pacemaker) and sometimes it doesn't. K20-tu-48-2: The description of this patient is asking you to assume that his HR is solely under the control of the artificial pacemaker.

37 Student Initiative Communicative Goal/Intention. Support: shows agreement with the tutor's claim. The student supports the tutor's point of view by accepting the knowledge the tutor is trying to give. The act of support is an indication of the student’s alignment of thought with the tutor. K4-tu-59-1: Let me remind you of the vascular function curve. K4-tu-59-2: It shows the relationship between central venous P (same as RAP) and CO when CO is the independent variable. K4-tu-59-3: DO you remember that? K4-st-61-1: Yes. I guess I do now. K4-st-60-2: A decrease in CVP would be in response to an increased CO.

38 Student Initiative Communicative Goal/Intention. Time Delay/Extension: Sometimes all that the tutor sees is a student pause. The student is busy in working the problem out and needs time to come up with a correct answer. The tutor's offer of help is the typical response to student pauses. K16-tu-17-1: Make your next prediction please K16-st-18-1: {Pause} K16-tu-19-1: Do you need any help to make a prediction at this point K16-st-20-1: I am thinking … K16-st-20-2: I just need a second more K16-tu-21-1: Ok

39 Student Initiative Communicative Goal/Intention. Compare and Contrast. It happens very often that the student confuses two parameters or state of affairs and asks the tutor to explain the difference between them. It is a subtype of request for information. K10-tu-61-1: Let’s put in the correct order, RAP (the dependent variable) is inversely proportional to CO (the independent one). K10-tu-61-2: OK? K10-st-62-1: What’s the difference? K10-tu-63-1: If RAP is the independent variable and it goes up, you get increased filling and increased SV (i.e.> CO). K10-tu-63-2: That’s Starling’s Law.

40 Focus of Attention or Content. Initiatives are not fully understood until their focus has been determined. Focus/content options: Rules of the Game Problem-Solving Algorithm Language Issue Causal Reasoning

41 Focus of Attention or Content. Language Issues. The proper use of the sublanguage --> reason for a natural language interface. The knowledge of correct terminology and its appropriate usage is essential for meaning negotiation. K12-tu-45-3: Does venous return go up immediately? K12-st-46-1: Does the rate of blood removal from the central veins mean that blood entering the right atrium, if so ithink venous return does go up immed. K12-tu-47-1: We need to get our terminology straight. K12-tu-47-2: Venous return means blood returning from the systemic circulation to the heart. K12-tu-47-3: That does not go up immediately.

42 Focus of Attention or Content. Causal Reasoning The focus is on a parameter or a relation or a mechanism. The student is required to predict the changes regarding the parameters in the given prediction table as a part of the problem solving procedure. Patterns of errors in the predictions or in the dialogue or questions in the form of initiatives inform the tutor about the missing knowledge. K16-tu-37-2: Does changing the length of the muscle change its CONTRACTILITY? K16-st-38-1: I think I am getting contractility mixed up with stroke volume… K16-st-38-2: Contractility is the force of contraction that I think goes up with increased heart rate, but I am not sure how

43 Focus of Attention or Content. Causal Reasoning The focus of many of the student initiatives is on the relations: either causal relations or equations that imply causal relations. How a change in one parameter causes a change in another or causes some effect falls under the notion of causality. The understanding of the underlying causes and effects is important to recognizing the function and behavior of the parameters. These relations are essential to the causal reasoning that our tutors want the students to learn.

44 Focus of Attention or Content. Problem-Solving Algorithm In problem-solving, the students are concerned with how the qualitative result is produced, and the tutor teaches about the sequence of computational steps that must be performed to get the desired value. The central issues are the primary variables and sequential changes in other variables in logical order. In the following example the focus is on understanding the orderly steps required in the steady state (SS) phase. K17-tu-56-1: I am a bit confused over your order of prediction. K17-tu-56-2: Since CO is determined by SV, how could you predict CO first and SV later? K17-st-57-1: I was just going down the list of seven variables and adding the magnitude of change (+ or -) from the DR and RR columns, as CIRCSIM demonstrated. K17-st-57-2: I don’t understand your question?

45 Focus of Attention or Content. Rules of the Game The student needs to be familiar with the phases and the order of the columns in the prediction table. S/he is expected to know what these phases mean. K16-tu-41-2: Does sympathetic stimulation change during the DR phase? K16-st-42-1: Does dr mean diastolic relaxation? K16-tu-43-1: NO! K16-tu-43-2: The DR occurs during the period of time before any reflex response to the perturbation of the system take place.

46 Degree of Certainty Hedging Almost any speech act can be hedged, although imperatives are not hedged as often as declaratives or interrogative sentences in our data. In our transcripts we see many types of hedges in the form of adverbs like “maybe”, “perhaps”; in the form of verbs like: “I think”, “I guess”; in the form of auxiliary verbs as “may”, “might”, “can not”; in the form of adverbial adjectives like: “I am not sure”, “I am not comfortable”; in the form of informal expressions like: “sort of”, and “most often”, question marks. The following examples are illustrative: K3-tu-53-1: The venous return may not change for a couple of minutes but what about the rate at which blood is being removed vfrom the central blood compartment? K3-st-54-1: That rate would increase, perhaps increasing RAP???

47 Example Illustrating the 4 dimensions of Student Initiatives K3-tu-54-1: The venous return may not change for a couple of minutes but what about the rate at which blood is being removed vfrom the central blood compartment. K3-st-55-1: That rate would increase, perhaps increasing RAP??? Surface Form: Other Goal: Request for Confirmation Focus: Causal Reasoning Hedged: Yes

48 Reasons for building natural-language based ITSs - which allows student initiatives Open-ended questions force students to think more deeply. Open-ended questions permit students to focus in on their specific problems. …but there are problems with unrestricted student initiatives: Communication Issue

49 . Student’s utterance can be too difficult to understand at the purely mechanical levels of spelling, syntax and basic semantic processing.. Understanding a statement at a literal level does not mean that we can understand the student’s model of the domain. More so if the student.s domain model is invalid.. Even if we can understand what the student is telling us, we may not have a constructive response available.. Even if we have a constructive response available, responding to the student initiative may not help the tutor achieve its agenda. Communication Issue Problems with Unrestricted Student Initiatives

50 Communication Issue Reducing Unwanted Student Initiatives 1. Ask short-answer questions instead of open-ended ones. Reducing the size and complexity of the expected response reduces the chance of misunderstanding a student utter- ance and the attendant frustration on the part of the student. 2. Ensure that each turn ends with an explicit request. With an explicit request, it is more likely that the student will answer the question rather than change the topic. It also provides students with an unambiguous indicator of when it is their turn to respond.

51 A tutor response involves sentences which realizes some tutorial or dialogue goal passing judgment on the student’s answer giving a hint asking the next question The human tutoring transcripts was used as the source of rules that determine machine’s decision on what to teach next how to teach it how to adjust the tutoring to various student resouses what topics should be elicited from the student what topics should be ignored language of individual sentences Tutoring Tasks

52 Classification of Tutor Responses (shah, p.76)

53 Classification of Tutor Responses 1. Hinting (or reminding) - a strategy that stimulates the recall of inert knowledge or activates the inferences needed in the completion of a task. Hints prompt the student to remember and organize the information that s/he already possesses. 1.a. Convey Information Hint (CI-Hint) 1.b. Point to Information Hint (PT-Hint) Definition / Classification Tutor Responses

54 The CI-Hints convey information directly to the student. K3-tu-51-2: And how might RAP change in the DR period? K3-st-52-1: The RAP would not change in the Dr period. K3-st-52-2: The RAP will not increase for about 1-2 minutes-the time it takes for the blood pumped out of the left ventricle to return to the right atrium. K3-tu-53-1: The venous return may not change for a couple of minutes but what about the rate at which blood is being removed from the central blood compartment? K3-st-54-1: That rate would increase, perhaps increasing RAP??? K3-tu-55-1: You are correct the rate of removal of blood would increase because CO is going up. K3-tu-55-2: But if you take blood out of the central venous compartment faster than it is returning, what happens to the central venous (I.E. RAP) pressure? Tutor Responses 1.a. Convey Information Hint (CI-Hint)

55 The PT-Hints point to pertinent information but do not explicitly convey information to the student. How the tutor points to a piece of information can be seen in the following example (in K5-tu ): K12-st-76-2: So i think that tpr goes down K12-tu-77-1: Your answer is right but for all the wrong reasons. K12-tu-77-2: Let's get back to this reflex. K12-tu-77-3: What is it trying to do in this patient? The tutor understands the nature of the student initiative that contains a hedge and reflects the misconception of the student. He recognizes the error in her previous knowledge, i.e., the situation of RAP during a minute or two. The tutor responds by acknowledging the correct part of the student answer and pointing to a hint on the incorrect part he thinks is a source of her mistake. Tutor Responses 1.b. Point to Information Hint (PT-Hint)

56 A multiturn dialogue segments in which the tutor communicates by posing a sequence of questions in order to enhance student reasoning capability. K13-st-48-1: If Stroke Volume decreases every time, Heart Rate increases, how can Cardiac Output ever increase K13-tu-49-1: What are the two parameters that determine Cardiac Output?. K13-st-50-1: Stroke Volume and Heart Rate K13-tu-51-1: Can you write the simple equation that relates Cardiac Output, Stroke Volume and Heart Rate? 2. Directed Line of Reasoning (DLR)

57 K13-st-52-1: Stroke Volume times Heart Rate = Cardiac Output K13-tu-53-1: Good. K13-tu-53-2: Then if Cardiac Output is up and Stroke Volume is down what must have happened? K13-st-54-1: Heart Rate increases more than Stroke Volume decreases K13-tu-55-1: Right, that's exactly what happens. 2. Directed Line of Reasoning (DLR) [cont]

58 An acknowledgement tells the student whether an explanation is correct or not. provides evidence of understanding provided in response to the student’s act. acknowledgment may be positive or negative depending upon the follow up question or theory presented by the student. K6-st-60-1: Does the direct affect steady state more than the reflexes? K6-tu-61-1: Yes. 3. Acknowledgement

59 EX. Negative acknowledgment, followed by an elaboration. K12-tu-33-1: By what mechanism will it increase? K12-st-34-1: If you increase pressure will you momentarily increase resistance K12-tu-35-1: No. K12-tu-35-2: You may be thinking of autoregulation. K12-tu-35-3: That's slow. K12-tu-35-4: Remember that we're dealing with the short period before you get a reflex response. K12-tu-35-5: Is this what you had in mind? K12-st-36-1: Yes i guess i am not sure then what happens to tpr 3. Acknowledgement

60 The tutors choose to summarize often in all types of tutoring dialogue. They often use this strategy of reinforcing important concepts in responding to student initiatives. K13-st-56-3: I am not sure if 120bpm is fast enough to cause that. K13-tu-57-1: Probably not. K13-tu-57-2: But more to the point, both tpr and cc change only when the reflex alters the activity in the ans autonomic nervous system). K13-tu-57-3: And since dr is BEFORE the reflex can act, both must be 0 in dr. K13-tu-57-4: Let's go on to the next column. 4. Summary

61 The tutor decides that student does not understand how the protocol is supposed to work. This response incorporates some instruction for the student directing her/him how to proceed. K13-st-24-1: Cc increases maybe K13-tu-25-1: No maybe's allowed. 5. Instructions in the "Rules of the Game"

62 The tutor is concerned about teaching correct usage of physiology language. Indeed this is one of the most important reasons for implementing a natural language dialogue in CIRCSIM-Tutor. K12-st-46-1: Does the rate of blood removal from the central veins mean that blood entering the right atrium, if so i think venous return does go up immed. K12-tu-47-1: We need to get our terminology straight 6. Teaching the Sublanguage

63 The tutor responds according to the correct logical order of multiple initiatives taken by the student. A major goal of the tutor is making sure that the student understands how to solve problems. K12-st-62-2: I'm just hesitant to say what comes first. K12-st-62-3: I'll go with tpr i to slow blood flow back to heart (i don't really like this idea) K12-tu-63-1: Well let's see if we can get at the first question I asked and then we'll come back to TPR. 7. Teaching the Problem Solving Algorithms

64 The tutor encourages the student in active learning through self explanation. This also helps the tutor to update his model of the student. For example: K5-st-102-2: But I'll bet that's not right. K5-tu-103-1: Well you're right in your bet. K5-tu-103-2: Stroke Volume decreases because Cardiac Contractility decreases. K5-tu-103-3: That doesn't mean that RAP has to be decreased! K5-tu-103-4: Let me remind you again of the vascular function curve. K5-tu-103-5: Does that help? 8. Probing the Student's Inference Processes

65 K5-st-104-1: RAP I. K5-tu-105-1: Would you explain. K5-tu-105-2: You're right but I just want to hear what you'r thinking. 8. Probing the Student's Inference Processes (cont)

66 When the tutor notices a delay on the student side, he offers his help. This is another tutor tactic to help the student in active learning. This response works as a rejoinder for the pause initiative. K5-st-45-1: I don (big pause here) K5-ti-46-1: Need help? 9. Help in Response to Pause

67 Used when tutors decide to avoid or put off discussion and bring the dialogue back to issues of higher priority. Used when the tutors do not understand what the student is driving at. K16-st-46-2: Is sympa stimulation the only factor influencing cc? K16-tu-47-1: It is in the experiment we are discussing today. K16-tu-47-2: All of your other DR predictions were correct, so please read page 6 so we can go on. 10. Brushing Off

68 Repair is done to avoid misunderstanding and correct misconceptions If the misunderstanding is not noticed at once, the conversation may break down at a later stage. So it is very important to make an attempt to resolve the issue immediately. K2-st-29-1: I am not familiar with a "curce" relationship K2-tu-30-1: Sorry I mistyped it should be "curve", the vascular function curve. 11. Conversational Repair

69

70 Response to student.s previous statement Acknowledgment of student.s statement (e.g. yes, no, you.re right ) Content-oriented reply (e.g. rebuttal or statement of support) New material Next part(s) of current schema Question for student to answer Basic Structure of a Tutor’s turn.

71 Correct-neural (V): 1. Make sure student knows that V is controlled by the nervous system. 2. Make sure that student knows that current stage is pre-neural. 3. Make sure student knows that correct value of V is no- change. Example of tutor tasks

72 .

73 . Correct answer. Wrong answer. Physiological near-miss: a step toward the correct answer. Linguistic near-miss: linguistically close but not exact answer. Student adds new information, e.g. an explanation.. Student changes the topic. Classification of Student Input

74 . Does not require a deep understanding of the student’s (possibly buggy) domain model.. Does not require reasoning about the student’s plan.. Does not require the use of stacked discourse contexts, as would be required, for example, if the student asked a hypothetical question. Limitations of Circsim Tutor (1997)

75 Evaluation Issue - some generalization based on Turn Planning Evaluation of Turn Planner user acceptability of text modifications using a user survey of the dialogue. 2 groups of students using 2 versions of CIRCSIM-Tutor, with and without turn planning. 3 types of textual modifications made by the turn planner - adding discourse markers, - modifying the referring expressions for physiological variables, - modifying the acknowledgments of the student’s answer

76 Evaluation Issue Turn Planner Goal in using turn planning: more coherent, more fluent, and easier to understand tutorial dialogue. Measures for comparison: Pre-test and post-tests of student knowledge and problem-solving ability. Measures of student’s interaction with the tutoring system (e.g., number of erroneous variable preditions or time on task.) Student survey regarding user acceptability

77 Lessons Learned Documentation and Analysis of Simulated interaction between students and tutors through a computer interface is very useful in helping build a realistic system. To manage complexity and user expectation, it is necessary to place an upper bound on the type of student initiatives/responses. (It is still not possible to use natural language without restrictions). Experimental methods can be use to evaluate ITS. Use of hints is very helpful in promoting learning in ITS.

78 References Evens, Martha W., Stefan Brandle, Ru-Charn Chang, Reva Freedman, Michael Glass, Yoon Hee Lee, Leem Seop Shim, Chong Woo Woo, Yuemei Zhang, Yujian Zhou, Joel A. Michael, and Allen A. Rovick. CIRCSIM-Tutor: An Intelligent Tutoring System Using Natural Language Dialogue. Twelfth Midwest AI and Cognitive Science Conference, MAICS 2001, Oxford, OH, pp Freedman, Reva. Degrees of Mixed-Initiative Interaction in an Intelligent Tutoring System. AAAI 1997 Spring Symposium: Computational Models for Mixed Initiative Interaction. Kim, Jung Hee, Reva Freedman, and Martha W. Evens. Responding to Unexpected Student Utterances in CIRCSIM-Tutor v.3: Analysis of Transcripts. Proceedings of the Eleventh Florida Artificial Intelligence Research Symposium, Sanibel Island, FL, 1998, Menlo Park, CA: AAAI Press, pp Marieb, Elaine N. Human anatomy & physiology. San Francisco : Benjamin Cummings, c2001. Shah, Farhana. Recognizing and Responding to Student Plans in an Intelligent Tutoring System: CIRCSIM-Tutor. Ph.D., Illinois Institute of Technology, 1997

79 Shah, Farhana, Martha W. Evens, Joel Michael, and Allen Rovick Classifying Student Initiatives and Tutor Responses in Human Keyboard-to-Keyboard Tutoring Sessions. Discourse Processes vol. 33 no. 1 (2002) Yang, Feng-Jen, Michael Glass, and Martha W. Evens. Evaluation of the Turn Planner in CIRCSIM- Tutor. 17th International Conference on Advanced Science and Technology, ICAST-2001, Chicago, pp Zhou, Yujian, Reva Freedman, Michael Glass, Joel A. Michael, Allen A. Rovick, Martha W. Evens. What Should the Tutor Do When the Student Cannot Answer a Question? Proceedings of the Twelfth International Florida AI Research Society Conference (FLAIRS-99), Orlando, FL, 1999, AAAI Press, pp References


Download ppt "IT803 Mixed-Initiative Intelligent Systems CIRCSIM – Tutor Presented by Bernard Yung Principal Investigators Martha W. Evens, Illinois Institute of TechnologyMartha."

Similar presentations


Ads by Google