"Dynamic" online courses will require new "normalized" content
Recently, online learning is gaining popularity, both as services for building individual learning trajectories, and microcourses, which make it possible to build trajectories more efficiently. Meanwhile, it seems to the author, it's time to take the next step and move from trajectories and micro-education to developing courses that will be formed dynamically, at the time of the client's request. However, such solutions will require the creation of fully updated “normalized” content (a knowledge base consisting of “training quanta”, with minimal redundancy of educational material).
For the past 15 years, I have been developing various multimedia courses in IT and STEM (mainly for university and corporate training). Accordingly, the hype around the massive open online courses (MEP) unfolded before my eyes. This format, which, in fact, boils down to the transfer to the online of the traditional university educational process, has so far eclipsed almost everything else and continues to grow in popularity. In this article, I would like to present my thoughts about the future of the MEP (namely, about turning modern “static” courses into “dynamic”) and show some pictures of the prototypes of “dynamic” courses that we began to develop in Nereputitor (I’ll just say that all screenshots made from the working version of the service, as long as it is not publicly available even in the form of an alpha version).
')
While the MEP was in short supply, almost all users perceived them with a bang, including in the corporate sector. Meanwhile, as the market became saturated, the disadvantages of the MEP format began to come to the fore, the main one of which, in my opinion, is, oddly enough, the redundancy of the material offered to users seeking to gain new knowledge and skills.
In school and university studies, whose programs have lined up over the years and decades, the problem of content redundancy is not so sensitive, since the MOEP was initially honed under existing curricula. Unlike corporate training, where the principle of “time is money” rules and where the result should be put at the forefront (namely, how well and quickly the employee mastered the necessary skill). At the same time, it is essential that in the corporate sector the initial knowledge and skills of employees are not as aligned as in the university. Therefore, in my opinion, it is from him that we should expect a request for a new training format that goes beyond the MEPs (especially since the money is concentrated just in corporate training, and with monetization of the MEPs, the situation, to put it mildly, is not healthy).
Let me explain the thesis about the redundancy of the MEP using simple examples.
Suppose a consumer (engineer, student, musician, etc.) wants to understand the basics of the spectra and begins to look for a course in Fourier analysis. Most likely, he will find the necessary information in courses on mathematical analysis, and with a lack of basic knowledge, he will be forced to refer to previous courses in mathematics.
Most likely, the consumer will have to filter the necessary information himself, moving in the opposite direction, i.e. from complex to simple. This is inefficient in terms of mastering the material, and leads to a waste of time. A variant of the course, immediately cleared of “extra” topics, would look much more attractive. In our example, important, but for this particular case, optional topics, such as differentials, limits, antiderivatives and L'Hôpitel's rule, etc., may well be excluded from the learning path, the minimal structure of which can be described something like this:
Of course, a natural question may arise: “how can the user deal with a certain integral if he“ does not need antiderivatives ”?”. However, in this case, he will be simplifiedly explained the concept of a definite integral through the corresponding area under the graph of the curve and the Riemann sum. Indeed, in this formulation of the problem, a strict explanation is not necessary, but in practice, for calculating the spectra, the integrals are still counted numerically, just across the area.
But how can topics be excluded from current, static courses that the user does not need at the moment (and it can take weeks for him to study)? The answer, in my opinion, is that - firstly, content of a different type is needed (“normalized”) and, secondly, an appropriate system for managing this content (knowledge base).
We give another example, now in the field of corporate education. Suppose there is a training system (either for internal employee training or for customer support), consisting of a set of interrelated lessons. Then for training, for example, warehouse workers could be limited to a small set of lessons (they are highlighted in color in the scheme). This is possible, again, subject to prior "normalization" of the content.
It is significant that in the first and second examples we are dealing with a compromise between the duration and depth of training. In fact, starting a conversation about courses created dynamically, we immediately recognize that to solve one and the same task (gaining certain knowledge, skills, etc.) you can generate not a single course, but a whole set of courses that differ from each other content, efficiency, size, cost and other parameters. Therefore, when creating tools for generating dynamic courses, you need to have an algorithm for comparing them to each other in order to be able to choose the best course (from the point of view of the current learning task).
Let's try (at least speculatively) to build a theory of the quality of education, starting from the requests of the corporate customer. Namely, we will put some indicator Q, which will be equal to 0 in case of absolutely useless training and 1 for absolutely successful, as the determining criterion of the quality of education. (Already at this stage it is clear that if we were to try to solve the problem mathematically and elegantly, then, apparently, we should use the terminology of the theory of fuzzy sets. This goes beyond the real - the practical orientation - of the article, therefore we restrict ourselves further to qualitative reasoning.)
It is obvious that Q (.) Is a complex function that depends both on the characteristics of the subject of the student - we denote them symbolically by the vector x 0 , and the parameters of the course C offered to him. With vector C we will denote the content of the online course, i.e., roughly speaking, the list of its member sections. You can also generalize the definition of C , considering it as a learning trajectory (which does not change the essence, since C is still a set of courses consisting of sections, tests, etc.). In other words (if we add the dependence of the quality of education on the time t spent on it), we can write down:
Q = f ( x 0 , C , t)
If we take into account that to achieve the same goal (gaining a certain knowledge or skill) different sets 1 , 2 , etc. can be offered, then the problem of choosing the best course C opt can be written as an optimization problem
f ( x 0 , C opt , t) ~ max.
As additional conditions, you can consider the limit on the time of study (which also depends on the course):
T ( C opt ) <T 0
and / or course price limit
$ ( C opt ) ≤ $ 0
etc.
Setting an optimization problem reverses the idea of ​​what an online course is. For me, it is obvious that the current MoEPs, in essence, static courses, will be replaced by such dynamic courses that will be formed at the time of the client's request.
The described approach, which is in general obvious (and is used in some form or another), has an obvious drawback: the “building blocks” of which courses C must be composed (I will call them a “training quantum”), somewhere take. The existing bulky courses of the MEP format are almost unsuitable for this role, and the micro-courses that gained popularity at one time may also not be suitable (as they usually do not describe relationships with other quanta — what exactly the client should know to master this quantum, and that he does not need to know)
Not to be unfounded, I will give an example of a prototype of our service. Normalized content is small lessons, minimally related to each other. A sample quantum is a minute clip + calculation (a Mathcad document and / or half a page of text). Training quanta are located in the knowledge base related to the graph.
Examples of quanta are:
Thus, there are no ready-made courses, as such, in the knowledge base; they are collected dynamically from quanta on a search query. At the same time, all content is not necessarily a single developer, quanta can be links to external sources.
Let us return to the example already mentioned, in which a client (engineer or student) orders a course on Fourier analysis. Depending on the settings, the client will receive something like this trajectory (or microcourse):
In addition to, in fact, the functions of online learning, it is reasonable to offer practical trajectories that represent how certain tasks are solved (code development, design and mathematical calculations, etc.) using one or another software. For example:
That is, depending on the client’s request, he will be offered a minimum rate collected dynamically.
Summarize. Dynamic approach to the generation of training courses will allow
For the past 15 years, I have been developing various multimedia courses in IT and STEM (mainly for university and corporate training). Accordingly, the hype around the massive open online courses (MEP) unfolded before my eyes. This format, which, in fact, boils down to the transfer to the online of the traditional university educational process, has so far eclipsed almost everything else and continues to grow in popularity. In this article, I would like to present my thoughts about the future of the MEP (namely, about turning modern “static” courses into “dynamic”) and show some pictures of the prototypes of “dynamic” courses that we began to develop in Nereputitor (I’ll just say that all screenshots made from the working version of the service, as long as it is not publicly available even in the form of an alpha version).
')
Modern MEPs are “static” courses.
While the MEP was in short supply, almost all users perceived them with a bang, including in the corporate sector. Meanwhile, as the market became saturated, the disadvantages of the MEP format began to come to the fore, the main one of which, in my opinion, is, oddly enough, the redundancy of the material offered to users seeking to gain new knowledge and skills.
In school and university studies, whose programs have lined up over the years and decades, the problem of content redundancy is not so sensitive, since the MOEP was initially honed under existing curricula. Unlike corporate training, where the principle of “time is money” rules and where the result should be put at the forefront (namely, how well and quickly the employee mastered the necessary skill). At the same time, it is essential that in the corporate sector the initial knowledge and skills of employees are not as aligned as in the university. Therefore, in my opinion, it is from him that we should expect a request for a new training format that goes beyond the MEPs (especially since the money is concentrated just in corporate training, and with monetization of the MEPs, the situation, to put it mildly, is not healthy).
Examples
Let me explain the thesis about the redundancy of the MEP using simple examples.
STEM courses
Suppose a consumer (engineer, student, musician, etc.) wants to understand the basics of the spectra and begins to look for a course in Fourier analysis. Most likely, he will find the necessary information in courses on mathematical analysis, and with a lack of basic knowledge, he will be forced to refer to previous courses in mathematics.
Most likely, the consumer will have to filter the necessary information himself, moving in the opposite direction, i.e. from complex to simple. This is inefficient in terms of mastering the material, and leads to a waste of time. A variant of the course, immediately cleared of “extra” topics, would look much more attractive. In our example, important, but for this particular case, optional topics, such as differentials, limits, antiderivatives and L'Hôpitel's rule, etc., may well be excluded from the learning path, the minimal structure of which can be described something like this:
Of course, a natural question may arise: “how can the user deal with a certain integral if he“ does not need antiderivatives ”?”. However, in this case, he will be simplifiedly explained the concept of a definite integral through the corresponding area under the graph of the curve and the Riemann sum. Indeed, in this formulation of the problem, a strict explanation is not necessary, but in practice, for calculating the spectra, the integrals are still counted numerically, just across the area.
But how can topics be excluded from current, static courses that the user does not need at the moment (and it can take weeks for him to study)? The answer, in my opinion, is that - firstly, content of a different type is needed (“normalized”) and, secondly, an appropriate system for managing this content (knowledge base).
Business courses
We give another example, now in the field of corporate education. Suppose there is a training system (either for internal employee training or for customer support), consisting of a set of interrelated lessons. Then for training, for example, warehouse workers could be limited to a small set of lessons (they are highlighted in color in the scheme). This is possible, again, subject to prior "normalization" of the content.
It is significant that in the first and second examples we are dealing with a compromise between the duration and depth of training. In fact, starting a conversation about courses created dynamically, we immediately recognize that to solve one and the same task (gaining certain knowledge, skills, etc.) you can generate not a single course, but a whole set of courses that differ from each other content, efficiency, size, cost and other parameters. Therefore, when creating tools for generating dynamic courses, you need to have an algorithm for comparing them to each other in order to be able to choose the best course (from the point of view of the current learning task).
The focus is on the quality of education.
Let's try (at least speculatively) to build a theory of the quality of education, starting from the requests of the corporate customer. Namely, we will put some indicator Q, which will be equal to 0 in case of absolutely useless training and 1 for absolutely successful, as the determining criterion of the quality of education. (Already at this stage it is clear that if we were to try to solve the problem mathematically and elegantly, then, apparently, we should use the terminology of the theory of fuzzy sets. This goes beyond the real - the practical orientation - of the article, therefore we restrict ourselves further to qualitative reasoning.)
It is obvious that Q (.) Is a complex function that depends both on the characteristics of the subject of the student - we denote them symbolically by the vector x 0 , and the parameters of the course C offered to him. With vector C we will denote the content of the online course, i.e., roughly speaking, the list of its member sections. You can also generalize the definition of C , considering it as a learning trajectory (which does not change the essence, since C is still a set of courses consisting of sections, tests, etc.). In other words (if we add the dependence of the quality of education on the time t spent on it), we can write down:
Q = f ( x 0 , C , t)
If we take into account that to achieve the same goal (gaining a certain knowledge or skill) different sets 1 , 2 , etc. can be offered, then the problem of choosing the best course C opt can be written as an optimization problem
f ( x 0 , C opt , t) ~ max.
As additional conditions, you can consider the limit on the time of study (which also depends on the course):
T ( C opt ) <T 0
and / or course price limit
$ ( C opt ) ≤ $ 0
etc.
Setting an optimization problem reverses the idea of ​​what an online course is. For me, it is obvious that the current MoEPs, in essence, static courses, will be replaced by such dynamic courses that will be formed at the time of the client's request.
Need “content normalization”
The described approach, which is in general obvious (and is used in some form or another), has an obvious drawback: the “building blocks” of which courses C must be composed (I will call them a “training quantum”), somewhere take. The existing bulky courses of the MEP format are almost unsuitable for this role, and the micro-courses that gained popularity at one time may also not be suitable (as they usually do not describe relationships with other quanta — what exactly the client should know to master this quantum, and that he does not need to know)
Not to be unfounded, I will give an example of a prototype of our service. Normalized content is small lessons, minimally related to each other. A sample quantum is a minute clip + calculation (a Mathcad document and / or half a page of text). Training quanta are located in the knowledge base related to the graph.
Examples of quanta are:
- e number and exp (x) function
- exp (z) - complex argument
- Euler formula exp (z) = sin (x) + i * cos (x)
- etc.
Thus, there are no ready-made courses, as such, in the knowledge base; they are collected dynamically from quanta on a search query. At the same time, all content is not necessarily a single developer, quanta can be links to external sources.
Let us return to the example already mentioned, in which a client (engineer or student) orders a course on Fourier analysis. Depending on the settings, the client will receive something like this trajectory (or microcourse):
In addition to, in fact, the functions of online learning, it is reasonable to offer practical trajectories that represent how certain tasks are solved (code development, design and mathematical calculations, etc.) using one or another software. For example:
That is, depending on the client’s request, he will be offered a minimum rate collected dynamically.
Summarize. Dynamic approach to the generation of training courses will allow
- store knowledge base: quantum of learning quanta
- generate learning trajectories (from these quanta) by the search query of the client
- in passing, consider the economy and the assessment of time-consuming (since certain fragments of the course may be paid).
UPDATE: Video Course on Dynamic Courses
Source: https://habr.com/ru/post/300566/
All Articles