Duck City : avatar and software agent programming
The success of Duck City's "Identiduck" prompted us to look more closely at the reasons for the phenomenon.
Eventually I decided that its success was based in a cognitive principle known as the "levels of processing" framework, (Craik & Lockhart, 1972).
According to the levels of processing framework, stimulus information is processed at multiple levels simultaneously depending upon its characteristics. Furthermore, the "deeper" the processing, the more that will be remembered.
For example, information that involves strong visual images or many associations with existing knowledge will be processed at a deeper level.
Using Identiduck to create a duck first involves making visually-based decisions but importantly, is terminated by a set of verbal decisions: choosing a first and last names, a job and an address. The stimulus array was sufficiently varied for the final decision to be processed at a significant depth, rendering it memorable. The result was that users recognised "their" duck when they revisited the site.
We became interested in whether we could use the ducks to invest Duck City with a slightly more ambitious purpose: if we turned the ducks into pukka software agents, we would imbue them with the ability to perform tasks on behalf of the owner.
We were keenly interested in what might be involved in developing a successful means by which people could easily direct the ducks' behaviour without having to learn a programming language.
The Formation Flying section offered an analogy we could use to get the ball rolling and introduce the ducks' extended abilities. In the stands, the ducks in the crowd are visually adjacent and are plausibly analogised as standing next to each other, providing the opportunity to swap jokes.
The proposal was: develop a facility by which the owner can tell a joke to their duck (via form input). The next time their duck is in the Flying Formation crowd, it swaps jokes with the next duck. When the owner revisits the site, their duck now has jokes to tell them.
Search engines are major features of the Internet these days but the search facilities they offer are both limited and commercially biased. None offer data-gathering facilities, each works on the assumption that its database is canonical and crucially, all assume that a one-off search will suffice.
One response is to give the ducks a text analysis capability along the lines of the webrats, then owners could ask their duck to keep an eye out for particular words appearing in texts. The ducks would use their innate ability as software agents to roam the Net, reading and analysing texts and marking the relevant ones for the owner's attention. If targets were shared between ducks, efficiency would be increased: compared to the computational cost of processing the text, checking the high-frequency uncommon words for multiple targets is cheap.
The growth of blogging has seen a related growth of RSS, a simple means of providing on-demand notification of change. For many years page "mind-it" services have been providing a similar email-based facility: RSS feeds can be integrated into desktop applications, providing almost instant notification of change. Providing each duck with its own RSS output enables the owner use a standard RSS-enabled application to receive communication from their duck.
Another notion was to give the ducks access to wordnet (a free-to-use semantic dictionary) and Roget's thesaurus so that the ducks can make a rich response when asked about a particular word.
These notions will eventually be implemented. After all, we're only ten years into the life of Bel EPA and those early years were a little too frenetic to allow many of these questions to be properly explored. Solutions remain to be discovered. The results could be quite interesting.
References
Craik, F. & Lockhart, R. (1972). Levels of processing: A framework for memory research. Journal of Verbal Learning & Verbal Behavior, 11, 671-684.