Hi Jon!

re McDonalds the world over...

It's good to see that you are still not letting considerations of health or personal wellbeing stand in the way of knowledge. More seriously, though, I wonder whether the model of McDonalds you are using is a formulaic one - ie there's a McD formula A+B+C which you must follow, but beyond that you can do what you like - or whether its a totalitarian one, where everything is supposed to be the same. If the latter, then it's interesting that all the culturally significant lessons you're learning are probably signs of things going wrong with the master plan. The minimalistic formulaic approach ought to work better... but we rarely build this sort of lack of control into our systems because, pedagogically, we've never needed to - it'salways been there already due to the inherent chaos of educational practice in the human-to-human world (and I'm speaking from personal experience, as you know). So you've got me thinking about the differences between flexible learning environments and ones that are truly minimalistic....

...but what do I know? I had a big Mac in (from memory) 1977 in Milton Keynes, didn't like it much, and have no relevant experience since....and at the back of my mind is the idea that McDonald's cooking had something to do with the Glencoe massacre...

Best Wishes,

Chris

- C W Boyne

Hello Jon. I read this post and absolutely think that viewers or visitors should be able to self-curate. This does present some challenges - and I like the idea of creating holes in the filters (or perhaps resetting parts of the filters) from time to time is a good way of introducing serendipitous possibilities into the mix.

One other situation where curation of a site seems to be applicable is when instructors want a learner to "see" something specific. I know of a situation where an instructor of a very very large course (1000+ at the same time) of students is needing to curate what the Tutors for the course, and finally the students in the course "see". The reason for this appears to be twofold - one to manage the infomation to ensure it is in line with the course outcomes and that in fact the examples are correct (where an example is being provided) and another is to manage the timing and amount of materials being provided to the students. There simply is TOO much information in a whole newspaper to read it all - unless one has the benefit of unlimited time...

Do you think, or know of, any ways in which filter bubbles could be used to provide students with personalized learning materials from a large curated list?  Would that be a particular kind of situation where students could choose what they want to see or not (based perhaps on their prior learning?) 

Finally - are you aware of any social networking site kind of implementation that is working currently that has the backend engines like Amazon which could be used to filter and curate large collections of relevant data for delivery in a personalized fashion to students?

By an odd coincidence I wrote a system or two that tried to do that in the early 2000s, CoFIND being the main one, that were used in precisely this kind of context. I wrote a PhD thesis around it. One of the notable things about CoFIND was that it inverted the usual approach to a user model. Most adaptive and collaborative filtering systems keep the user model hidden, and that's a big problem because we may not even know that we are in a filter bubble and, even if we do, we may not know much about the algorithms and data points that are being used to provide us with recommendations. My approach was to treat the user model as a set of user-generated tags (not just discrete categories but also fuzzy categories) that could be added, aggregated in new ways, and chosen according to context. For instance, learners could seek resources that were 'good for beginners' or 'funny' or whatever tag had been applied by others in a similar role, so you were not treated as one unified persona but as a constantly shifting human being with ever changing and adapting needs. Unfortunately, though I still think it was conceptually cool, the system was incredibly hard to use and exponentially magnified the cold-start problem, mainly because each combination of tags created its own micro-ecosystem. Any ways that I found to reduce complexity also reduced the value of the recommendations. Judy Kay's notion of scrutable user models was better, but ran into similar difficulties. Giving control over your user model adds too much complexity.

There have been many attempts to produce collaborative filters for education, maybe starting with CoFIND (which stands for 'Collaborative Filter in N Dimensions'). At least, it was among the first used in formal educaiton, though there were plenty of earlier informal learning/knowledge management systems like PHOAKS ('People Helping One Another Know Stuff') and even Tapestry, that started it all in about 1992, that explicitly tried to recommend things to learn from. There has been quite a bit of work on it at OUNL - I helped to examine Hendrik Drachsler's PhD on that subject, for instance (the link is to some of his papers). There are some huge issues using such things in education though: previous preference is way less useful in determining future needs in learning being perhaps the biggest one. One or two of Hendrik's papers have a good review of the issues -see http://inderscience.metapress.com/content/n63082472v3210l0/ for instance. Timing remains an enormous obstacle: these systems are OK if you are riding the wave and want to know where to go next, but hopeless at showing paths (and no, aggregation of paths doesn't work!). I don't think anyone has really successfully overcome the problems yet, but I and others are still working on it.