RU2294144C1 - Education method for training and diagnosing learning capability - Google Patents

Abstract

FIELD: medicine.

SUBSTANCE: method involves testing trainee, controlling actions directed to solving tasks set by means of computer system and representing information. Action control is carried out in operation-by-operation mode with distance to target being calculated as sum of given encouragements on correct actions and penalties on wrong actions. The trainee receives help by setting feedback between the trainee and computer system as an image on monitor screen representing distance to the target and producing analog task sequences as dynamically varying virtual problem medium images. Training process results are recorded on line by means of computer system.

EFFECT: enhanced effectiveness in controlling trainee activity and following dynamic training process characteristics.

4 cl, 5 dwg

Description

The invention relates to the field of education, in particular to methods of training with tracking the student’s activities, management of this activity, and is intended to diagnose the dynamic characteristics of the process of forming skills for solving problems in various subject areas.

A device is known for conducting psychometric studies, including explicit and implicit observation levels, with the possibility of comparing them, which makes it possible to isolate diagnostically significant parameters-correlators of the psychophysical state of the operator from the sensor signals in the process of solving information problems (RU 29457, U1, A 61 B 5/16 2003.05.20).

This device allows you to only monitor the activities of the subject, without correlating and not directing this activity.

A well-known automated system for monitoring performance is characterized by the availability of databases, including test sources, information about students, a logical unit that is capable of calculating the performance of each student, calculating average performance, comparing results, setting a weighted average score on an established scale for each person tested and determining the ratings of the tested groups (RU 20603, U1, G 09 B 19/00, 2001.11.10).

The known system only monitors the statistical characteristics of the learning process and does not allow monitoring of the dynamic characteristics of the learning process to solve problems. In addition, the system provides for work only with databases and does not provide for modeling tasks.

A training device is known for diagnosing an integral assessment of training, including a space navigation navigator, models of an onboard complex control system, a model of sensors, a model of a spacecraft’s motion, an image generator of the Earth and the ISS station, control sticks (RU 37429, U1, G 09 B 9/16 2004.04.20).

The proposed training device has a narrow professional focus. This device does not allow for operational monitoring of the learning process and logging activities. Learning is also not monitored as a characteristic of the training process.

Closest to the claimed one is a method of monitoring the operator’s professional suitability, which consists in installing sensors on the vehicle’s controls and information display system, testing by comparing the time of the operator’s actual reactions to external factors and the vehicle’s operation, monitoring the operator’s actions for correctness by n difficulty levels, with an increase in the level of complexity during the testing process (RU 2199272, C1, A 61 V 5/16, A 61 V 5/18, 2003.02.27).

This method involves only monitoring the correctness of operations and is not intended for training and management of the learning process with the generation of tasks and tracking the characteristics of the learning process, such as learning ability, the level of insufficient specific learning ability, the magnitude of the inductive threshold and others.

The objective of the present invention is to provide a method that allows you to teach and diagnose learning with the management of the learner, in the learning process and tracking the dynamic characteristics of the learning process.

The technical result of the invention is expressed in increasing the degree of interaction of the student and the automated system.

The objective of the invention is solved in that in a method of training and diagnostics of learning, including testing a learner, monitoring actions to complete the tasks carried out using a computer system with the ability to display information, the actions are monitored step by step, with simultaneous feedback between the student and the computer system in the form of helping the learner by encouraging correct and penalizing incorrect actions and generating a sequence of analog tasks, results about learning process is recorded using a computer system in on line mode.

Assistance to the student can be performed at different frequencies.

On-line mode can record the construction (solution) of tasks.

In on line mode, a time series of events (actions) can be recorded.

On-line mode can record the construction (solution) of tasks.

In on line mode, a time series of events (actions) can be recorded.

The inventive method is illustrated by drawings, where

figure 1 is a structural diagram of the implementation of the method of training and diagnosis;

figure 2, 3 shows a series (a-c), experimentally obtained graphs of the reward function of poorly and well-trained students;

4, 5 are graphs of the functions of the value of the state of students with insufficient and good learning.

The method of training and diagnostics of learning is as follows. The training takes place at the student’s computerized workplace, which is located in the system and is networked with the teacher’s computerized workplace. Since human training occurs as a result of his interaction with the environment, the proposed method using a computer system simulates the ways of such interaction, offering a real virtual environment instead. In solving the problem, the student performs some actions or operations. The environment, in which both the tasks and the “teacher” are included, transmits signals to the student (by means of displaying on the monitor screen), which are estimates of the actions performed by the student. These estimates are considered binary: the correct action is plus 1; incorrect action 0. To solve the problem, the student can choose actions from some given set D = (d1, d2, ..., dn). The learner's goal is to obtain a solution to the problem. Correctly performed action brings the student closer to the goal, incorrectly performed - removes from the goal. At the same time, the student receives timely signals with an assessment of activity. The value of the assessments of actions that can be designated 1 and 0 is formed by the “teacher”. In terms of the reward function, a score of 0 corresponds to a penalty of minus 1. When a student completes a sequence of actions and evaluates each of them, a total gain is formed that is equal to decreasing the distance (in actions) to the goal. The medium (in this case, a computer training system) transmits information about the size of the current total winnings to the student. The student made the right or wrong action, the environment determines it, the student knows about it only with a certain probability. Even before the action is taken, the student makes a choice as to what he will do. It is at the stage when the process of choosing an action is going on that there is a chance that he will receive encouragement or punishment. In the process of solving problems, the probability of evaluating and, to a greater extent, self-assessment of actions in the direction of increasing the likelihood of taking the right actions that are not penalized changes.

The appropriate behavior of the student is determined by the probability of choosing the right action. A student who does not have appropriate behavior chooses the right action with a probability of 0.5. This corresponds to a maximum of entropy. If the student acts appropriately in the learning process, then the probability of the correct action will increase. At the same time, the entropy of the results of its activity decreases. The environment with which the student interacts can help reduce informational entropy; otherwise, there is feedback between the student and the computer system by transmitting information about the current total gain or the reward function. This information plays the role of reinforcing the right actions.

The actions performed by the subject in solving problems form a time series of events. Operational control carried out by a computer system allows you to get a reward function. Adaptation of a computer system to a student’s possible level of training is carried out by varying the frequency of prompts determined by the level of training. The prompts received by the test subject from the system play the role of reinforcements.

The time series of events (actions) is recorded in computer memory and is experimental material that allows you to measure changes in the execution of actions.

The diagram illustrating the method shown in Fig. 1 shows the interaction of the student and the training and diagnosing system: numbers and letters denote - 1, 2 - control modules, 3 - task generator, 4 - an analytical module that records and analyzes information about student activity 5 - a module that remembers the results of the student’s activity, information about the student’s activity, x _в — task, U — control action, Z — disturbance (environmental influence), x — controlled quantity, R ₁ , R ₂ — optimality criteria, F — functionals, y , φ - corrective actions (reflect the reward function and the state value function), P is a switching switch.

The following are examples illustrating the implementation of the method.

Example 1. A student, solving a problem, performs n actions. Of these, n ₁ are the correct actions and n ₂ = nn _{1 are the} wrong actions. The first task allows you to determine the likelihood of reinforcements that the student will receive in the second task. The probability of reinforcement is q = n ₂ / n ₁ . This is the probability that the system will display to the student information about the value of the reward function. In solving the second problem, the student also performs actions that can be right and wrong. The average number of reinforcements during the solution of the task is ν = n * q. The policy of the student performing actions depends on the state he has achieved, which is characterized by the value function of the state I (i) = 1-H (i),

where i is the task number, H (i) is the informational entropy of the student's activity.

Information entropy is determined by the relative frequencies of correct and incorrect actions in task i

H (i), = plog ₂ -qlog ₂ q,

where p is the relative frequency of correct actions, q is the relative frequency of incorrect actions. The value of the state value function depends on the results of solving the previous task.

Thus, the reward functions allow you to determine the value of the state value function, which must be estimated to obtain a higher reward.

Example 2. A student is trained to solve the tasks of constructing graphs of a function that is generally written as

The student constructs the graph of function (1) by transforming the graph of the function y = F (x), which is given on the coordinate grid of the display working field. To do this, the student must carry out the transfer operations of the graph y = F (x) along the OX and OY axes, the compression and decompression operations of the graph y = F (x), the inversion of the graph so that the graph of the desired function is obtained (1). The system interface has corresponding buttons for controlling both the position and the graph of the function y = F (x). The memory of the machine records: actions or operations performed by the student; time taken to complete the action; the correctness (+1) or incorrectness (-1) of the performed actions; the value of the function is the value of the state and other products of the student’s activity.

The value function of the student’s state φ (t) at time t _{i + 1} = t ₁ + Δt _{i + 1}

where S (t) is the student's reward function, it is equal to the sum of the rewards (+1) and penalties (-1) received by the student in the process of completing i + 1 tasks. Feedback coefficient ƒ (t) is formed by the function F

The feedback coefficient determines the type and frequency of assistance that the learning system provides to the student. The frequency of help or reinforcement of the student’s activities depends on the achieved value of the state value function. As φ increases from zero to 1, the average frequency of reinforcements decreases from one reinforcement by two actions to zero.

Comparison of the experimentally obtained series (a-c) of the graphs of the reward function of poorly and well-trained students, shown in Figs. 2 and 3, where the abscissa axis shows the time to complete the tasks - t, on the ordinate axis - the reward function - P-HP allows you to track the dynamics learning process. A student with insufficient learning in mathematics (figure 2), performing significantly more tasks than a well-trained student (figure 3) cannot learn to complete all the tasks correctly. In his actions, the random component is still great (random action). As a result, as can be seen from the graph depicted in Fig. 4, where the number of completed tasks is plotted on the abscissa axis and the state value function is plotted on the ordinate axis, the maximum value of its state value function did not exceed 0.5 (5th level) . A well-trained student stops quickly enough to make erroneous actions, the time for completing assignments is reduced, and accordingly, the state value function reaches a maximum value of 1, which corresponds to the 10th level of achievement, which is reflected in the graph shown in Fig. 5.

Thus, the inventive method allows you to educate and diagnose learning with the management of the learner, in the learning process and tracking the dynamic characteristics of the learning process.

Claims (4)

1. A method of training and diagnostics of learning, including testing a learner, monitoring actions to complete the tasks carried out using a computer system with the ability to display information, characterized in that the monitoring of actions is carried out step by step, with the computer system calculating the distance to the target, which is the amount received as a result of encouraging the right actions and penalizing the wrong actions, and at the same time assisting the student with feedback waiting for the learner and the computer system in the form of a display on the monitor information about the distance to the target and generating a sequence of tasks analogs, in the form of a dynamically changing display of the virtual environment problematic, the results of the learning process is recorded by the computer system in the on line mode. 2. The method of training and learning diagnostics according to claim 1, characterized in that the computer system produces information about the distance to the target with different frequencies. 3. The method of training and learning diagnostics according to claim 1, characterized in that on-line mode record the construction (solution) of tasks. 4. A method of training and learning diagnostics according to any one of claims 1 to 3, characterized in that on-line mode records a time series of events (actions).

Images