EduMining

One of the main challenges of my research is that of validation. While trying to understand online learner’s motivation through their log files, I’ve came to the stage where the automatic tool for doing that is almost ready. The mechanism is clear, the code is written, and only one tiny little screw is missing in this machine: validation…

The measuring tool we’ve been developing should give us the answer to the following question: What is the level of motivation of a student during a given learning period? And it seems that the suffix of that question is the core of the problem. Unlike other questions that might be answered aftera certain period of time is over (e.g., What is the student’s grade? What is the student’s attitudes towards a completed task?), the affective state should be measured continuously and constantly. Actually, this is why I was atracted to measure it through log files in the first place.

So, what is the best practice for collecting information from online learners during their learning sessions? More precisely, the questions is that: What is the best practice for collecting information from online learners during their learning sessions, without disturbing the learning and with the highest chances of getting an accurate and reliable feedback (as much as any human feedback is accurate and reliable)? Well, now that this question is loudly asked, it’s time for me to go looking for answers…

Data mining has many applications. Every data mining freshman must have heard the cool story of how Wal-Mart used association rules for increasing their selling rates. This story, which according to many is only an urban legend, demonstrates what we all now know: Data mining is good for e-commerce.

Take, for example, Amazon.com. In Web Mining background presentations I give to M.A. students in courses discussing (research of) Web-based Learning Environments, I usually demonstrate to them the “surfing experience” in this huge Website. First, I login with my real username/password combination, and go over the list of recommended items. Here is a glance to the first three items in it (click thumbnail to enlarge):

The list is built mostly of history novels, history of science non-fictions, and some networking gadgets. Then, after a few keystrokes during which I change the username/password, I reveal the Mr. Hyde in me, and although the system is still greeting me with my real name, a totally different list of recommendations appears (click thumbnail to enlarge):

Suddenly, Origami and napkin folding (that’s pretty cool!) populate the wish-list I presumably wish to list. Since both accounts were used by the same (real) person and from the same machine (i.e., same IP), it’s clear enough that Amazon has been using the usage data (i.e., log files) of its customers and surfers in order to “help customers find what they want and purchase it in a way that is simple and convenient” (i.e., increase the store’s selling numbers; quote by Amazon’s Senior Manager for Data Mining, taken from here).

Well, although it might be funny (or not), this little example demonstrates the well-known power of data mining. However, what I’m most curious about are the less-known-sometimes-strange potential implications of it.

Take, for example, the future! Moreover, think about how depressing it will be. Now, when you’re already bad-mooded, it’s the right time to cheer you up with some great news: Data mining has shown that although medium-term future will be quite confusing, long-term future will be good! This fascinating result was presented by Alberto Pepe of UCLA (paper was co-authored with Johan Bollen LANL) in AAAI 2008 Spring Symposium on Emotion, Personality, and Social Behavior (held three weeks ago at Stanford University, Palo Alto, CA).  The full paper can be reached via arXiv.org.

Well, thinking about it, data mining has to do with forecasting, so even depression forecasting doesn’t seem too weird. But, what about politics? Well, as been reported a few weeks ago, there is a strong connection between those two, and it is basically data mining application which brought NY Governor Eliot Spitzer to resign. The first link in the chain which finally brought to the 1000$-an-hour-prostitution-affair revelation was, as it turns out, a warning produced by a fraud-detecting software which uses data mining techniques to point out irregular money transactions.

* * *

So, now I’m really encouraged. When I first told friends and peers that I’d like to use data mining methods for extracting motivation of online learners (and affective aspects of online learning in general), they looked at me in an odd way. “How do you think you might evaluate one’s motivation without seeing him, without hearing him, without talking to him?”

Well, now I’m sure I’m not the only dreamer. If politics and depression may benefit from data mining, why shouldn’t affective learning research? It seems that my feelings regarding it had been phrased already:

You may say I’m a dreamer, but I’m not the only one

(Imagine, John Lennon)

Well, this is a great excuse to meet again with this wonderful song…

Our last post has probably encouraged the editor of KDnuggets, and 3 years after the first Data Mining Haiku Competition was held, the current issue of KDnuggets News (n06, March 25, 2008) is announcing the second competition. The Haiku presented here is already a competitor.

So, wake up and sharpen your pencils or warm up your keyboard-connected-fingers, and start thinking on the 5-7-5 rythm.

It is the Hebrew month of Adar now, the main commandment for which is “Be Happy!”. So, I decided to be.

Three years ago, Data Mining Haiku competition was held by KDnuggest, the winners of which had proven (again) that humor may bring together two disjoint fields, like science and poetry.

During the last weeks, I’ve been very busy with analyzing some big log files, and this morning - maybe it is the happiness atmosphere of Adar - a Haiku came into my head. Since clearly I can’t send it to the 2005 competition, and since Haiku - just like scientifical research - is more valuable when presented to the public, I’ve decided to publish it here.

I don’t know if a Haiku should be entitled, however I call this one: Preprocessing.

Cleaning, reducing
and ignoring outliers.
Only one case left.

A few days ago (February 20, 2008), an article was published in the front page of the Israeli daily “Ha’aretz”. The article, which in its English version was titled: “Looking at the Monkey in the Mirror”, was titled in Hebrew (with my translation to English): “The Israelies Learn About Themselves - in the Zoo”. Both these titles relate to the self-observing, which is the main finding of the fascinating research covered in the article. The research was taken by Dr. Yaron Idan, an Israeli Sociologist, who wished to portray Israeli visitors to zoos. Here is the main finding, as reported in the newspaper (the resaech was not published yet):

Contrary to the popular perception among zoo directors in Israel and abroad, the visitors hardly ever come to learn about the animals, Yaron says. They come mainly to learn about themselves. During visits they play a kind of “mirror game.” “The animals serve as a kind of mirror in which the faces of human beings are reflected. The visitors see the advantages and disadvantages of human beings. The discussion surrounding the animals always comes back to the human and his characteristics,” says Yaron. By observing the animals, the visitors engage with matters of parent-child relations, sex and gender, nutrition and external appearance.

It took me only a few seconds to discover that Yaron and I share not only the need to understand people’s behavior, but also some principles of how to do it. Here is a short description of Yaron’s methodology (again, from the newspaper article):

Armed with a baseball cap, notebook and a pair of sharp ears, Yaron eavesdropped on zoo visitors without them knowing they were part of his research.

“Wait a minute!” I screamed towards the newspaper as soon as I’ve read this paragraph, isn’t this wrong? Apparently, some of the talkbackists (in the Hebrew version of the online article) screamed about that eavesdropping issue also. A few seconds lated I’ve calmed down. The talkbackists haven’t.

Well, first of all, let’s remember that Yaron didn’t do anything new methodology-wise. Unobtrusive methods are in use in Social Sciences for decades, and among the various forms of implementing this idea one may find: used condom collecting for measuring safe sex¹, obituary analysis for investigating changes in gender steretypes ², garbage sorting for finding consumption patters³, and many other examples. However, one of the simplest (and less bizarre) modes of unobtrusive research is observation on people who does not know they are being observed. The zoo resaerch mentioned above demonstrates such an unobtrusive observation-eavesdropping resaerch, but if it was a virtual zoo - no baseball cap and sharp ears the resaercher had needed.

Online discussion boards are a fertile ground for research of human interaction on the Web. Being public, organized and archived, Internet communities have been the source for many studies, and although being on the Web - there is no significant difference between unobtrusively reading those forums’ scripts and unobtrusively eavesdropping zoo visitors. Both cases should follow some basic ethical rules. Here is, for example, a paragraph from Research Ethics Application Guidance Notes, taken from the School of Psychology, Queen’s University Belfast (Ireland):

Some research involves unobtrusive observation of naturalistic situations where it is not possible to inform participants or gain consent prior to the observations. In these situations you need to state how the privacy and individual confidentiality will be preserved.

That means it is o.k. to analyze text extracted without prior consent from the writers or eavesdropped discussion - as long as the Resaerch Authority (or its corresponding responsible staff member) had approved it. Is this the case for Usage Analysis also?

On one hand, usage analysis of learners’ log file records may be treated as an unobtrusive method for objectively determining their actions; replacing each learner’s identification value with a random number might solve the privacy problem. But on the other hand, data mining might be pretty privacy harming, since its results may directly effect the person the traces of whom is being mined; online learners being categorized according to their behavioral patterns (not to mention online buyers being categorized in order to increase the amount of money they spend). For understanding that data mining is much more a complicated case than “simple” unobtrusive methods, it is recommended to read the hypothetical scenario of Lee’s loan application and the consequences of the information he agreed to give the bank⁴ (the whole article is inspiring and worth reading).

What  many education resaerchers claim regarding those ethical issues, is that they - contrary to the bank or to the commercial Websites like Amazon.com - have only good intensions. They don’t want (so they say) to increase the amount of money spent for purchasing, but to enhance the effectiveness of the learning. But is it possible that those good intentions might, unknowingly and not intentionally, invase someone’s privacy? Unfortunately, the answer is “Yes”. Paraphrasing the well-known saying, it is sometimes true that the way to heaven is paved with unethical manners.

* * *

I thought this post will present some solutions, but it has been quite long enough with only presenting the introduction to the problem. Future posts (and, of course, the responses area) will be a good place for delving into this issue with presenting possible privacy harming (and other ethical concerns) and some suggested solutions.
 

References

1. Lister, N. A., Smith, A., Binger, A., & Fairley, C. K. (2003). A novel research approach in sex on premises venues (SOPV): objective measure of sexual behaviour and low level intrusion to patrons. Sex Transm Infect, 79(1), 53-55. [Available online as for February 28, 2008] [↩]
2. Rodler, C., Kirchler, E., & Hölz, E. (2001). Gender stereotypes of leaders: An analysis of the contents of obituaries from 1974 to 1998. Sex Roles, 45(11-12), 827-843. [Available online as for February 28, 2008] [↩]
3. Wallendorf, M., & Reilly, M. D. (1983). Ethnic Migration, Assimilation, and Consumption. The Journal of Consumer Research, 10(3), 292-302. [↩]
4. Tavani, H. (1999). Informational privacy, data mining, and the Internet. Ethics and Information Technology, 1(2), 137-145. [Availabe online as for February 28, 2008 [↩]

Since I don’t (yet) have a laptop, I couldn’t live-blog yesterday from the Learning in the Technological Era: The Third Chais Conference on Instructional Technologies Reserach, organized by Chais Research Center for the Integration of Technology in Education, a research unit of The Open University of Israel. Therefore, you may treat this as post-live-blogging, or maybe even dead-live-blogging…  This conference became, since its very first occurance (two years ago), the main stage in Israel for academic discussions regarding the various aspects of the Learning and Technology meeting points. Researchers from allover the country (which is not so big after all) gather to hear and to be heard, to share and to comment, and especially to feel the (real/offline) communityness. Young researchers, and even graduate students, find it a suitable place for fertile deliberation regarding their studies.

If this conference indeed reflects in some way the Israeli learning and technology resaerch community, here are some interesting numbers for giving a perspective regarding this community:

• 15 panels (plus 1 workshop) were held within 3 parallel sessions
• 47 papers were presented within those panels
• additional 17 posters were presented
• in total, there were 137 authors (including papers, posters and workshop)

You can watch an English version of the program here. (For the Hebrew-enabled, all the papres avaialbe here, later to be added presentations and video recording of the lectures.)

However, I guess the most important number will be that: only 1 presentation reported about a research that has something to do with the Data Mining methodology (this is, of course, as far as my eyes could scan all the titles and abstracts of the presentations I didn’t see during the day; please correct me if I’m wrong)! A little confession should be given at this point: no one of us (in the EduMining group) had sent a paper to the conference this year, and this is mainly due to schedule (i.e., other deadlines) limitations.

* * *

After attending a dozen of presentations, it came to me that something quite ackward was common to a subset of them. About five empirical studies (both quantitative and qualitative), which examined the use of computer-based learning units, used some kind of pre-post analysis of students’ performance and/or understanding. The research questions of all of those studies refered to the effectiveness of the certain examined learning unit without any comparative view of other computer-based units or traditional units.

Don’t unerstand me wrong - I’m not saying that any pre-post kind of research should take the comparative approach. Not at all! These studies choose different kinds of case study approach, and this was a reasonable choice taken by the researchers, considering their purposes and backgrounds. Almost all of them were asked about it during the 5-minutes-after-the-presentation-Q&A, and according to them - they are not going to do any comparative research, since it is not required according to their research interests.

I was just wondering why did they choose to evaluate the effectivity of the computer-based learning unit by itself, without examining other possibilities. While drinking coffee and talking about it with some colleagues, I brought up my interpertation to this almost-phenomenon.

Although a rapid growth in the use of computers in schools had started already in the 1970s (computers were brought into schools already in the 1950s), and although today computer-based and Web-based learning seems quite natural - it has been criticized that generally and in a large scale, technology (almost) didn’t change a thing in the way we learn. The criticism often comes from researchers within the technology-enhanced learning community.

So, we have some 30 years of experience with computers in the learning process, and on the other hand we have constant skepticism from allover around. Therefore, it is of no wonder that many researchers offensively try to prove that technology does work! Unfortunately, this battle looks more like a defensive to me, because what I feel from the large set of such studies is quite the opposite, and to me their approach seems like being taken from an insecure point of view. They try to say: “Technology works for learning!”, but I hear the between-the-lines “Does it really work?!”.

This feeling of mine becomes more stable when I observe that many of the technology researchers I saw were not fluent with the technology (e.g., they couldn’t solve extremely simple problems during their PowerPoint presentations). It was Dan Halutz, the former Chief of Staff of the Israel Defense Forces, who said (after being criticized that an officer who served only in the air force and has no ground combat experience could not chief the army) that “you don’t need experience as a sheep to be appointed a shepherd”. And I must wonder whether this “unsheepingness” of the researchers I saw yetserday (which are not, by any means, a representative sample of the Israeli research community) has to do with the defensive attitude they promoted.

One of the main limitations of the human brain, and I guess it has much to do also in this case, is that new situations are being examined and treated by means of old paradigms (which is quite understandable, of course). This is why innovations diffuse rather then just appear allover¹. Using Roger’s Diffusion of Innovations terms, I must ask whether the reserachers I mentioned earlier, those who took the somehow-offensive-somehow-defensive approach, are the “early adopters” or are they actually the “laggards”?

And for not ending with quite a pesimistic mood, I’ll let you enjoy this (quite famous already but still) hilarious short clip² about adopting new technologies. It is called: “Introducing the Book”, or “Medieval Helpdesk”:

1. Regarding the evolving use of Web pedagogies on university course Websites, see:

   Shemla, A. & Nachmias, R. (2006). How Do Lecturers Integrate The Web in Their Courses? Web-Supported Courses at Tel-Aviv University. In E. Pearson & P. Bohman (Eds.), Proceedings of World Conference on Educational Multimedia, Hypermedia and Telecommunications 2006 (pp. 347-354). Chesapeake, VA: AACE.

   As for innovative pedagogical practices in schools, you might be interested in:

   Mioduser, D., Nachmias, R., Tubin D., & Forkosh, A. (2006). Innovative Pedagogical Practices using Information and Communication Technologies. Tel Aviv: Ramot - Tel Aviv University (in Hebrew). [↩]

2. Broadcasted back in 2001 in the show “Øystein og jeg” on Norwegian Broadcasting (NRK). [↩]

Two news articles published during the last few weeks have attracted my attention. Although not mentioning edumining research or research of logged online activity at all, they (immediately) made me wonder about several aspects of our research. Let me just summarize the two stories:

1. Speed tracking technology using (existing) cellular broadcasts. “Or Yarok” (in Hebrew: green light), an Israeli non-profit road safety organization, promotes a revolutionary technology: using existing cellular broadcasts for measuring drivers’ driving-speed. This application is quite simple to understand. Cellular broadcasts can be gathered and used for knowing the position of the car in certain times. Given coordinates and timestamps of a certain car traveling between them, and knowing the physical figure of the road - velocity calculation is enabled. This way, the organization’s researchers can measure average driving speeds for each road (for which data is available) and may, for example, correlate it with number of accidents occured on that road.
2. Restaurant of the Future knows exactly what, when and how (but not why) you eat and drink. This fascinating project, held by Wageningen University (Netherlands), is probably one of its kind (if not counting The Truman Show). The project description, citing from the project’s homepage, is as simple as that: “The Restaurant of the Future is not just a place to experiment with new food products, preparation methods and self-service systems, but also a facility allowing close observation of consumer eating and drinking behavior”. If not understood, “the restaurant” tracks each and every movement (both of food and of people) within it. This complicated research facility, as summarized in the article published in the New York Times, aims on answering a simple yet complicated question: what makes people eat and drink the way they do?

One main difference distinguishes between these two studies reported, and it is due to this difference that the first one got many aggressive talkbacks in the online version of the newspaper, while the second one was treated as an anecdote: while every visitor in the Restaurant of the Future is a research subject by agreement, none of the drivers the cellular broadcasts of whom were analyzed knew about any use of their cellular (talks-independent) activity.

* * *

Before inferring anything on this blog’s domain, let me first examine some more similarities and differences between these two researches:
Data collected is “anonymous”. It is clear that data regarding the same person should be recognized as such, hence data atoms cannot be identified by random numbers. However, there is no need in recognizing the data with a “true” person, and it is not relevant who was the person driving the car (or even which car was it) or who is the one eating the sushi. Any identifying parameters may be replaced by random identifiers (keeping in mind this replacement should be injective, at least for each “session”).

Costs of research pretty much differ. The Wageningen sushi restaurant’s project is estimated in 2.3 Million Euro (according to this report). It is clear that the cost of the driving-speed research is much lower. The main cost-gap is, obviously, in the collecting mechanisms.

Scaling. Both the restaurant and the roads researched have physical capacity limitations, being serial in nature (each coordinate in every lane of the road can “carry” at most one car at a certain time; number of people dining at the restaurant in any given moment is, of course, limited by terms of chairs), and therefore the research population size is, theoretically, limited by a number known in advance.

Research questions may be asked in retrospect. In both cases, the data is collected in the most atomic level possible by the research designers, allowing them later to ask many quetsions they didn’t think of beforehand.

Nature of data collected. The cars research uses very basic information, each raw of which documents (so I guess) mainly place and time indicators. The restaurant research uses much more complicated data (imagined e.g., 3-D location, weight, facial expression, tables/chairs configuration).

Continuousness of data collecting. In both cases, data is collected continuously, but in contrast to the 24/7-available roads, the restaurant has specific opening hours, out of which data is not collected.

Multidicliplinarity. The restaurant research is truly multidisciplinary (e.g., psychology, anatomy, computer science, culinary). The velocity research is quite straightforward.

* * *

I must admit now that I really intended to infer from these two totally-offline-totally-not-educational researches on edumining, but after revisiting the points I’ve just named, it seems really unnecessary. Those topics of comparison (and maybe some more) refer to each and every one of our researches as well. Although I wanted this post to discuss the great benefits we have using Web mining techniques in education comparing to similar research in the “real world”, I’m now standing - after finishing it - quite confused. It seems that research is research is research, no matter what is the subject matter, which is the population investigated and which methods are being used.

Each topic mentioned above referring the velocity/eating research, might also be (and I’m sure is indeed most of the times) in mind when planning a research in edumining. There is no set of answers relevant to all the edumining studies: some of them are anonymous by the nature of the data and some of them use data that should be anonymousize (is there such a word?!); some logs are cheap to collect, but some actions might be expensice to track; some Web-based learning enviroments are parallel to all users (e.g., fully on-line course), some are quasi-serials (e.g., Wiki platforms); often research questions may also be asked even after data was collected; there is huge differences in the complexity of the logged data, and it is sometimes being fit to the research purposes; some online educational media are open to the public 24/7, but some are limited by time, space (e.g., by IP) or identity (e.g., for current students only); and although edumining is mutidisciplinary by its nature, some research within it uses only a small portion of the big package.

So, is edumining special in any manner? I guess there is a short answer and a long answer to that question. The short one will be: “Yes!”, the longer: “Well, it’s hard to explain it in only a few words at the botton of a long post; better to dedicate a special post for it”.

They say a picture is worth a thousand words, but they never got to calculate how much a video is worth. However, the video I’ll show you in a minute demonstrates that the idea of visualizing log files is limited only by our minds.

Here is a little exercise. Take yourself a few seconds of quiet, close your eyes and let your imagination lead you. Mumble the words “visualization” and “logfile”, and try to picture the very best solution that will enable you to better understand those thousands of raw data. Now, open your eyes, take a glass of cold water and keep reading. You will be stunned.

How about “realtime logfile visualization”? This is the solution offered by the cool tool named glTail.rb (I must admit I didn’t understand the meaning of that name nor did I find an explanation anywhere). And it is free.

Here is a demonstration of the output, watch it once before the explanations:

Now for the explanations. What you just saw is a visualization - recorded in relatime - of HTTP traffic on certain URLs. Each flying blob represents a hit on a Website, moving from the referrer (on the left hand side) to the requested URL (on the right hand side), color indicates the referrer. The blob size indicates the response time (small blob equals short response time). Further blobs on the right hand side indicates URL which were accessed directly. There are more data on the screen, and it is all explained in the project Website.

Now, take yourself a minute, open your eyes and watch this video again. So, how many words is a video worth?

I am peeking carefully to the table that the programmer sent me… After days of planning, determining goals, collecting, clearing and transforming the data with the programmer – It is on my desktop… Rows of data, which expose every action of the kids in the site… a golden treasure.

Taking a deep breath and few minutes to wonder before I’ll examine every detail in it… What I am going to see? What I am going to feel about it? A lot of information is stored in this data, such as: the time the learner spent on each online activity, the pace of his learning, the mistakes he made and the feedbacks that he got. Information that can tell a lot on his learning strategies, his motivation and preferences…

I am sorting the table by users and time of action, and reading row after row. The first thought that is passing throw my mind is that the kids are learning! Ok, maybe this is too soon to tell, but they are “here”, they are doing the activities, they are engaged with the content, and they want to have a good computerized feedback in the end. That was my first impression- without any data mining technique, yet (!).

1. A simulation that demonstrates the concept of the Hebrew month:

2. Extracted (row) data that presents the usage of the simulation by one of the students:

(Click on the picture to get it bigger)

So now, after the data has been extracted and stored, we can begin to mine…
The first method that will be used is the Usage Clusters, in which we will try to find groups of users that share common surfing patterns in the educational activities.
Some findings will be presented in the future posts.

One of the first things that convinced me to choose the path of research I’ve been following for the last two years, was a simple question raised in a meeting with my current-advisor-and-then-a-lecture-in-a-course-I-took-within-a-pre-PhD program. After presenting me with the idea of EduMining, Rafi asked this question out loud. “Will we be able to say that a certain learner is terrified, or on the other hand, very motivated, during an online learning session, without seeing him or talking with him?”

The moment that questions was suggested - I now know - was the moment my professional life took a significant turn. A few weeks later, I quit my job as an Algorithm Engineer in a big Israeli company and officially filled-up the PhD program forms. The turn was even bigger when realizing that I registered to the School of Education, while my MA and BA were under the Exact Sciences. And this question has been incessantly gnawing at my mind.

That very same conversation, during which the potential (and then still theoretical) implications of Web Mining in Education also were mentioned, didn’t came to the “How?”, but rather focused on the “What?”. But even then, our envy at the Cardiologists was spoken.

Take a look at the cardiologist. A patient is coming to be examined, and the doctor should check a very specific and unseen part of this patient: his or her heart. And how does the cardiologist check the person’s heart? By connecting him or her to a strange machine which reflects the heart activity and plots some strange curves (ECG/EKG - Electrocardiograms). By examining these curves - which are a visual representation of the heart activity - the doctor says all sorts of things about the heart activity itself. Can we imitate this very same process regarding our field of interest?

The answer to this question is based on a few assumptions:

1. the online learner is connected by default to a strange machine which reflects his or her activity, the output of which is the log files
2. a visual representation of the online learner’s activity might help us in saying something meaningful about him or her
3. the instructor (or the resaercher) should be able to understand the visual representation of the online learner activity in order to tell something about him or her

According to this analysis, we are still missing both the visual representation and its interpertation tools. This is exactly the gap we aim to bridge by using the Learnograms. A learnogram is a visual representation of learning-related variables which are based on data hidden in the log files. Learning variables may be simple and directly extractable from the log files (e.g., time pattern, pace of action), or high-level ones (e.g., motivation, frustration, learning strategy). It is the latter which are of our main interest and, of course, less trivial to produce. However, we have began in the process of defining such variables, and it sure feels like it is not impossible to do it.

Here is four learnograms of Johnny, a student in a fully-online learning unit (click to enlarge):

We may see (from top to bottom) Johnny’s mode of learning, pace of activity, time pattern and (subjective) knowledge (the system investigated keeps this information, and it was a shame not to use it) for his learning period of more than two months. This tool gives us the opportunity to “meet” Johnny without actually seeing him or taking with him, and to try and say something about his behavior, and from that - about his learning.

It will be, of course, an understatement to say that the learnograms, reflecting activity, may render an understanding of the learning, but this is one of our main challenges and the focus of my PhD resaerch. I truly believe this is not impossible, and in future posts I will tell you more about Johnny and his learning, as an example to the knowledge we may gain by analyzing the learnograms.

Well, thinking again about the question from my meeting with Rafi some three years ago. Will it be possible to tell something about Johnny’s motivation or frustration during his online learning only by analyzing his logs, just like Johnny’s cardiologist can learn about his heart activity from his ECG? I think there is still a long way until this vision becomes a reality. But I also think that we are on the right path for making it true.

For making this vision clear, take a look at this overused short clip, but think of it from a new angle: imagine that this guy is taking an online learning course right now (within his organization); ask yourself whether we will be able to recognize his distress, and if so - will we be able to prevent the disaster?

Imaginery? I don’t think so…