OpenCulture

Following a brief Twitter discussion this afternoon, @miaridge asked me to put together a use case for an idea which has been rolling around my hind-brain for a good few years now. The idea first, then the use case.

The idea comes from four places:

1. The adage ‘knowledge grows through use’, which I acquired some years ago from a quotes website. The principle being that knowledge is dynamic and emergent, and that it thrives through the process of exchange
2. The fact that there is a physical manifestation of this principle in the way that neural pathways in the brain form, strengthen, detach and re-combine in response to changes in external stimulus
3. The way in which procedural AI in computer games can generate apparently complex, individual and motivated behaviours by combining a few simple starting conditions and essentially linear algorithmic rules
4. A bar in the City of London which shows prices for drinks on a ticker-tape - the prices fluctuating constantly in response to the demand for particular drinks

You can liken a user moving through a website to a chemical/electrical signal coursing through a brain. As a thought travels around the cells of the brain, it isn’t just being transported, it is actively changing the environment through which it travels.

Every change, fluctuation in power, redirection or bifurcation doesn’t just have an effect on the signal, it also affects the infrastructure, the axons and dendrites through which it travels. The more a pathway is used, the greater the nourishment it receives - there are even specific hormones such as Somatotrophin which bind to cells that receive more use than others.

In most cases, the process and structure of experiencing Collections through a web browser is essentially linear. At its most basic, a user types a pattern/alphanumeric string into a search box (or initiates a pre-defined pattern by selecting a browse option) - the underlying querying engine interrogates a data repository for matching patterns and returns the records/information which is linked to them.  This kind of classic pattern-matching search is the basis of the vast majority of online record-based experiences.

It carries with it some inherent flaws. The first is that it is a hit-and-miss affair, depending on someone having associated a pattern/search string/classification with a record. This makes searching really a game of ‘guess the pattern most likely to correspond to the thing you are looking for’. The second is that it is inherently inefficient - every search initiates a quest for matching patterns throughout the entire dataset.

But human beings don’t seem to work this way - either linguistically or in consciousness - because human information processing is context-aware. Evidence suggests that a significant part of our understanding of a given conversation, for example, is based on the starting conditions and context for that conversation. Put more simply, knowing how and why a conversation got started gives you a head-start on narrowing down the list of things that conversation might be about, and hence lets you respond more quickly.

The publishing of Collections online is still (even with options for User-Generated Response) a one-way affair, and one that resets every time a new user arrives at the point of entry. There is a beautiful aerial photograph of sub-Saharan watering hole (in the book The Earth from the Air: 365 Days) which shows how generations of migrating animals have worn into the landscape a huge network of interwoven paths. Imagine that after every visit to the water’s edge, a sandstorm arose and obliterated all evidence of previous use. Each new user would have to find their own path, with the attendant risk of missing their target entirely. This is the model for most current-generation Collection-browsing experieces.

So the idea, at heart, is to begin to work towards Collections-based systems that are intelligent about their own use and are capable of re-flowing their content and interface in a dynamic and emergent way to reflect how they are being used. Imagine records which accrete not just narrative meaning, but also contextual meaning based on a dynamic analysis of a combination of many user behaviours.

This takes us beyond the kind of Amazon recommendation-engine experience - from ‘users who bought this book also bought these other books’ to a behind-the-scenes re-flowing of the options and interfaces available to the user based on ‘users who exhibited broadly similar use patterns to you tended towards this type of interface’.

Of course, none of this is all that new. Personalisation, recommendation, fuzzy-logic interfaces, homonym and allophone search all exist - but they all suffer from two main drawbacks. On the first hand, they are active technologies requiring direct engagement from the user. On the other, they are (mostly) one-shot deals, supporting the user through a session/experience, but incapable of learning or improving by comparing the experiences of many users.

On the other hand, the skill of Web Managers in all industries has always been partly about interpreting from the mass of log files the broader themes and threads of user need and then of re-interpreting their Information Architecture both to satisfy these needs more quickly and effectively and to keep hold of the user’s attention by tantalising them with further relevant information. A big part of this idea, then, is simply to develop systems which respond automatically and dynamically to their own usage data - a kind of mechanised and incremental democracy of both interpretation and prioritisation.

And so to the use case…

Alice is a researcher in a biotech company. She uses museum collection databases to source type specimen for a research project she is working on. On Thursday morning, she logs onto the Natural Sciences Museum’s website, types in a taxonomic name into a search box, and clicks ‘go’.

At this point, the system knows nothing about Alice. But it does contain a history of the user pathways which have resulted from the same search string. It knows that last week Bob used the same string as part of his research, and that he ended up looking at two related taxa and even that he sent himself an email containing information about a specific plant.

Subtly, and without Alice really being aware, the interface reflows to prioritise options of the type which have proved useful to people searching for similar things in the past. It doesn’t remove alternate pathways, but it de-emphasises them. The idea, after all, is to offer Alice supportive guidance through the Collections system without constraining her choices.

As Alice moves through the Collections system’s information environment, she finds that it is opening up before her, giving her a sense of flow. Lists of results are re-sorted to prioritise records that have received more user-generated comments, or have been favourited more recently. By default, the system displays the longer text which experience suggests have been more useful to people with research interests like Alice’s.

And while the system is providing Alice with her enhanced browsing experience, it is also learning from her actions. At one point, she follows a blind alley, and has to backtrack. The system notes this from her overall user session, and adjusts the statistical relevance of the record accordingly.

And that’s about it for the use case - not much to go on, I know! The thing is that most of the technologies exist to deliver this kind of augmented browsing experience - user tracking, statistical relevance sorting - but they have yet to reach the point of maturity where they are combined into truly intelligent systems, or the price-point at which they are affordable to museums. Until they are, however, the online museum collection is likely to remain an impoverished experience, and one which is doomed to remain ignorant of one of its most important pieces of knowledge - the data about how it is used.

This entry was posted on Sunday, September 13th, 2009 at 8:21 pm and is filed under General, Nick Poole, Uncategorized. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.