A cognitive analysis of tagging is a defining blog post focusing specifically on the initial input part of tagging. It compares tagging to categorising, and the mental process that we deal with, as well as differentiating between physical, and digital realms.
Tagging is described as a 1 stage process, where you think of many concepts and tag them, seems natural, intuitive, easy, and fun (well fun for librarians!)
Seems like an easy process; although when considering later retrieval you have to think carefully about the tag names for future findability…but I guess the first concepts that come to mind are probably going to work best
Categorising is described as a 2 step process, where you think of many concepts but you only choose one concept for the item.
Now even if an application like Outlook enabled multiple folders for an item, it doesn’t mean you are tagging…to me a folder is physically a bucket or a container to put stuff in, whereas tagging is sticking a label on something, like a post-it note (more specific).
When you categorise in the physical realm, it happens naturally, you don’t think of the overall scheme, and how you will remember it again, or which file in your memory to recall the experience.
But Rashmi suggests with digital objects there is the findability issue, so at the time of tagging/categorising we freeze up in trying to decide a label to fit for future recall.
But he mentions that with this freezing up condition called “post activation analysis paralysis”, the use of tagging helps eliminate the decision of choosing a category, which makes saving an item for future reference less daunting (cognitive cost).
We also understand that whether using categories or tags, our set of labels need periodic reviewing as our understanding of concepts may change, or the specificity, or scope of our collection may change.
ie. What I used to call “audiofeed” I now call “podcast”, so I have to review my collection
…and now that my interest in bicycles is growing, I’m going to have to review my “bike” tags, and start labelling, “bmx”, “racer”, etc…
This notion also spills over into folksonomies, even if people are tagging for personal reasons, a folksonomy easily aggregates these accounts for sharing, it’s an experimemt, it’s not a system where all tags are overhauled and merged and all content is reviewed to make sure they are all in the right tags, and then presented to the public.
It is a tool of chance, serendipity…it’s not meant to be perfect…for a starter many people are using different tag names for the same thing (car, automobile), or the specifity may be different. Someone might use the tag “bike” for an item about a bmx, whereas someone else will use the tag “bmx”…searching the tag “bike” to find a bmx item might take a long time or you might not even know there is a bmx item in a tag called “bike” (unlike a thesaurus, folksonomies aren’t exhaustive), but a folksonomy is what it is, each personal account is accurate for the owner (both creating the tag name, and choosing what item belongs in which tag within your tag set), but may not be for others, the fact we can join up all these accounts is fascinating.
Tag set relationship
Also on the personal level how do your tags relate, if my collection is about bikes, I might use specific brand name tags for the brands I like and need to retrieve quite often, for other brands I might just use the tag “bikes”, now in a formal taxonomy, this doesn’t make sense as naturally all the items in the brand name tags, should also be included in the more general “bike” tag.
But this is because formal taxonomies are usually made for a collective mind, whereas a personal folksonomy only needs to make sense to one person (unless you tag for others or for the benefit of an emerging vocabulary), so the tags relationships don’t have to make sense in the traditional way.
You could make a personal taxonomy out of your tags, but if the purpose is for retrieval and not necessarily browsing, then just apply the tag name that you will think of later on when wanting to retrieve that item. In this way you don’t really care what your tag set looks like, you just want to type in a tag and find that item and the job is done.
A post by Otis (Simpy) describes how he uses hundreds of tags, as findability is much easier, whereas some people still use tags as containers or categories, tags really work well for specificity…and big deal if you have a massive tag list, just search a tag, or search full-text.
More from the post on how he uses tags:
“…I tend to use them as keywords, as I place a lot of value on being able to find my tagged data with a search, and I place less value on having cleanly organized items. The side-effect of this is that I have a lot of tags with low occurrence counts. Some may think of this as a bad thing - my tag cloud is huuuuge, but the advantage of this approach is that I can use search to quickly find exactly what I need, in my archive of 1000+ bookmarks. I don’t care to group things, I care to find precisely what I need, when I need it. The Web is a mess, but you still use Google and not DMOZ, to find information on the Web. Search is the King (Kong).”
So he is tagging according to the purpose of his needs, this works for him.
This is the difference for some people, it is in completeness
…if I tagged a past item about bmx bikes with the tag “bike”, and now I tag a new item about bmx bikes with the tag “bmx”, does this make sense (of course you could review your collection)…well this might happen if you have hundreds of tags, as now the tagging process is harder as there are so many tags to choose from, but then you have full-text search on a service like Simpy
…but the problem is completeness, your collection is getting messy for browsing purposes, but then again you may not use it to browse.
So it comes down to the purpose of, why are you saving this item?…or the purpose of your collection.
Are you building a directory type tag set for browsing content or are you labelling things for quick findability regardless of its relationship to other labels.
If you are saving for personal findability then tag and forget is the answer.
If you are saving to make topic buckets then quick tags only go so far, as tags generally scatter a topic (you need to use them more as categories, like in email), that’s why tag bundles in del.icio.us are effective, as you can put your tags in a greater concept.
eg. if I want to give someone a synopsis on RSS search engines I can show them my RSSsearch tag, but what if I don’t use an RSSsearch tag, but I tag items under the various RSS search engine names, then I have to try and collate all the tags into a topic, whereas tagging them RSSsearch would have already done this. But then if I want to retrieve an article about just one particular RSS engine it will take longer to find.
The weigh up is how you are going to most use your collection, if you are locating one item, then it doesn’t matter if your collection is messy, but if you want to browse a tag like a topic, being messy might not be effective as the topic is scattered across lots of tags (if you can find them all).
An answer might be to have 2 tag sets one for tags (quick findability), and one for tag containers (browsing), these containers would act as a higher level tag.
So in your tag set you could create 2 tag bundles, one for the high level containers, and one for the specific tags, this way you can use one bundle to browse a topic, and the other bundle for quick finding of an item, this isn’t to say that a tag in the quick find bundle couldn’t become a specific topic of its own.
So every time you bookmark an item, it goes in at least one tag in both tag bundles.
For this to work you’d need a bookmarklet that seperates your tags into both tag bundles so you can choose a tag/s from each bundle, otherwise it’s too hard to differentiate.
If you are saving items for a public research portal then tags may not be the answer, you may want a controlled vocabulary…why…because when people research they like to be exhaustive, they want to know they can now leave this portal knowing they have seen every possible item related to their research topic (this is considering the thesaurus is intuitive, this is another folksonomy issue).
So in comparison, the unique features of a folksonomy are more for personal recall, quick finding, serendipity, sharing, we are getting value out of the chaos…folksonomies aren’t always intended to be built like a topic portal or journal database, they aren’t even built really, they just become from the unison of user collections, and they keep becoming…they are like a manifestation or a materialization.
Categories vs. tags
[ADDED - also see Quick Study: Tag Use Patterns]