Why does ‘Freesound’ succeed when so many learning object repositories fail?


Bryan Alexander posted a link to The Freesound Project and it was interesting to me for a whole slew of reasons.

It was interesting first off because I have been using the site myself for the last few months; I am getting more into making music with digital audio tools (yes, yes, I will post something, someday, give me time to build up my courage) and so turned to Freesound to find new samples for a drum machine. And it works; not perfectly maybe, but you can definitely find new samples fairly easily, and it has a number of other social affordances (‘users who downloaded this also…’ and folksonomies) that lead you to related stuff you might like.

I was interesting also on a personal level as it was built as part of the 2005 International Computer Music Conference. ICMC is dear to my heart because way back in 1995, I was responsible for building the first website to support a ICMC conference, when it was hosted in Banff (the only remnant of which I can now find is this reference, the ‘WayBack machine‘ not even going back that far, so safe to say Internet ancient history!)

And finally it’s also of interest as a ‘repository’ of shareable remixable content, and one that would have to be judged relatively successful at that, with around 10,000 ‘objects’ and almost a million downloads. So what makes it tick, why does it succeed when so many of our various ‘learning object repository’ projects are failing so miserably? Let’s consider (more)….

Is it the Technology?
Is it the technology that makes the difference? First let’s look on the contribution side of the equation. The uploading process certain doesn’t seem that ‘user friendly’ – it requires users to ftp files, then reconnect on the web to tag them. Yet on second glance, this does actually allow for ‘batch uploading’ to occur, something many LORs have stumbled with, and the user’s login account is common to both ftp and website, which brings the user back to their files fairly easily.

How about the metatagging itself? Again, some improvements maybe, but hardly enough to explain all of the success. Sure, they employ folksonomies with all of their related benefits (user controlled, easy, lots of good serendipity) and problems (stuff can get buried/lost, synonym issues, like between drum, drums and drumloop). But they have their issues too trying to educate their users on metadata conventions just like the rest of us. They don’t use words like ‘metadata’ with their end users, which seems pretty wise, nor do they seem overly concerned about adopting an infinitely interoperable metadata scheme. Instead they seem concerned with capturing just enough information to allow users to find records that might be what they need, and then an easy way to preview to decide yeah or neah, which seems pretty smart indeed.

And how about on the search and retrieval side, what parts of the technological ‘system’ can be seen as contributing to the success here? Well, there search interface should be credited with being pretty slick; I like the way the one search box can be quickly expanded to search various fields and how the advanced search fields are cached and revealed with one click. The fact that you can hear the sample with one click without having to download it has got to help users make quicker evaluations on the value of any sample. The ‘Sample Packs‘ are a nice feature that allow users to download a collection of samples at once, and one that I think we’ve done a poor job on in the LOR world. In my own use of Freesound, I actually found myself downloaidng these sample packs most often, as it was easier to download a collection, and then discard the pieces locally as I found they didn’t work in the context I wanted, or just left them on my hard drive for future use. Bandwidth and hard drive space are quickly becoming non-factors. The remix tree is also something to pay attention to, if only because it shows ‘remix’ culture at work and the social aspect of it too. I guess that’s part of the thinking behind the ‘Geotagged samples,‘ though I expect this is also related to both the conference and just a cool Googlemaps hack.

So I don’t want to diminish the effect that good technology has on the success of this project; I think there are many of the features above that do make it quite a usable and useful system, stuff that I can definitely learn from. And they’ve done it all on top of open source technology, and not one monolithic piece either, but a number of pieces woven together quite well.

But none of this seem to me to totally explain why there are 1 million downloads of of 10,000 samples here, and yet most of our LORs can’t break 1000 ‘objects’ and similarly dismal numbers of downloads and reuses. So what other reasons could be at play here?

What else is contributing to the success?

The License – a certain amount of credit has to be given to the Creative Commons license itself (An interesting side note; the license here is the Sampling+ license, different from that used for most learning content, and not without its detractors.) When you sign up for an account they have you acknowledge the rights and responsabilities of the license (another good innovation, instead of agreeing to it over and over) and then its off to free downloading. Free is good, right? And certainly, the LOR world is acknowledging this too, with more and more educational resources available under a similar license. So another factor, but not the entire one.

The use case – part of the success here is due to the use cases around using samples. There has been an explosion of tools to help users reuse samples in making music, and the use of samples in music has been gathering steam for the past 30 years to the point where it is pretty well ubiquitous. Not so the use case around reusing “learning objects.” Where a sample repository like Freesound is a solution to an existing problem that users have and want solved (where can I find free samples for my music that I don’t have to clear copyright for), LOs/LORs still represent a bit of a solution that’s trying to convince its users of the problem (hey, doesn’t it sound like a good idea to reuse and remix existing learning content instead of developing it all from scratch). It also likely helps that some of these sounds are now easy to produce with the right gear (though some are not, and there are some real dedicated folks doing a lot of work to create some of these samples).

The media itself – there’s a few aspects here that I think give Freesound an advantage over LORs. Relatively standardized media formats (sure there are a bunch of them, but there are some leaders and also free tools to convert between them). On the LOR side, what’s the standardized format for ‘learning content’? HTML? Flash? Powerpoint? PDF? Word? XML? IMS Content Packages? All of the above? But on top of that, the type of thing that is being shared and reused has a profound effect on the success here – I mentioned before how Freesound allows users to preview the sample right in the web page with one click. Not only does this make it easy to preview, the very fact that within a few seconds of hearing a sample you can often judge whether it will work or not gives them a huge advantage over ‘learning content’ of any granularity, which will almost certainly take more time to assess. So a corollary to the famous ‘reusability paradox‘ – not only are things of smaller granularity more reusable, it is very likely much easier to assess their reusability. (An additional area worth investigating is the effort/reward ratio of finding higher level objects to reuse and how this is effected by not accounting for the costs of our time, but that’s another thought for another time).

I was going to write something about how the users themselves are different, but I’ll admit to chickening out (and ran out of time too). But check out the discussion forums here and compare them with your typical higher ed CMS forum, as an example. I don’t want to ‘blame the users’ (actually, I kind of do sometimes) but maybe just leave it that Freesound seems to be addressing what feels like a more sophisticated user base.

So, why write all of this? Well, I do think there are lots of good things to learn from projects like Freesound (and it’s not the only one; people have pointed me to sites like Deviant Art and Canada’s own Zed TV site as other great examples, to name a very few). And I also think it is useful (at least for my own tired brain) to disentangle what works generally from what works specifically in this realm of shareable, remixable content. Hope it was for you too. – SWL