Cybernetics — What? The Domain of Cybernetics — Cybernetics may be defined as the science of describing goal-directed systems. The term comes from the Greek word "kubernetes", meaning "steersmanship." (From the same Greek word but through Latin, English gets the word "governor.") In general (center of diagram), cybernetics is useful in understanding a goal-directed system that exists in an environment which, because it is dynamic, makes an unchanging set of actions by the system inadequate to achieve the system's goal. Initially (left side of diagram) cybernetics was concerned with systems that were observed (such as thermostats). It soon became clear that one could also observe systems that were themselves observing (right side of diagram). This brought issues of language, meaning and subjectivity into sharp focus. I. Lines of Enquiry These statements suggest how cybernetics could offer perspectives on interaction design: When a system has a goal, cybernetics offers definitions of "information," "feedback" and "interaction." When interaction includes a person, cybernetics offers useful models of "personalization," "learning" and "understanding." When software is in the way of understanding (between person and person, or person and machine), cybernetics may open the flat interface into a semantic space; develop a history of clicks into a relationship; and transform a computer into a partner with a shared goal.   Modeling Goal-Directed Systems — Systems that have a goal may be modeled by this notational scheme [adopted from Gordon Pask, in Soft Architecture Machines, edited by Nicholas Negroponte, MIT Press, 1975]. An "intelligent system" (left side) is defined as a system that maintains at least one goal and calls upon a set of actions (gold lines) that may achieve that goal. An intelligent system must also review the results of actions (blue lines) and compare (black-and-white circle) those results to the desired goal. If the action fails to bring about the goal, alternative actions must be tried and may involve a request to other systems (right side). Such other systems can be partners if they understand not just the requested action, but also the goal that the action is intended to achieve. This requires another type of interaction (green line) that takes place in language and that may be explicit ("I'm trying to keep the room cool.") or implicit, that is, not stated but implied by the circumstances.   It is an enigma to me that goals are not an explicit part of the discussion around user interface design, no less an explicit part of the interface functions themselves. This is not a rant in support of impractical abstractions, but a lament that so many opportunities are lost. It would take some work to connect the "Lines of Enquiry" statements and the model above to interaction design, but for now I offer an anecdote: How many times have you dispatched an email without the attachment that you wanted to send along? Is this your mistake, or the error of the UI designer for failing to allow you to assert your goal and to let you assert it at the moment it was clear in your mind — just as you open a new message. A new menu command, "Send New Message with Attachment", would not let you send a message without an attachment. Even in a simple case, a goal-oriented design process leads to a modest, but I believe useful, insight.   II. Architecture of Interaction Human-to-human, but with intermediation — When the conversation is human-to-human and in real time, something novel may arise from the interaction. When technology is interposed, problems arise, especially since both participants are not available at the same time to react dynamically to each other. To compensate, the technology can bring some responsiveness back by allowing the selection processes (arrows) to be variable and based on a history of clicks, rather than fixed functions such as links. Collaborative filtering, for example, takes input from the participant on the right and makes selections of content for display based on statistical "similarities" that are the essence of this filtering technique. The disadvantage is that the history used as the basis of the dynamic responsiveness is aggregated across many people, and the individual participant on the right may not be represented by the resulting statistical aggregate. Alternatively, the selection process could be based on the local context of the individual participant. In this case, more is required than current systems deliver. In the process of assembling and organizing the content, the participant on the left can be aided by authoring tools that capture the content into a structure that allows for adaptive display. In the process of delivering content based on clicks from the participant on the right, the interface must accept signals and track behaviors that express the user's context, which becomes the basis of the "personalization."   III. On-Line References Materials on cybernetics are only slowly being made available on-line, and these limited references cannot reflect the depth and breadth of the field. And, as in any field with many contributors, not everything under the name of cybernetics is consistent, and some sites may give a confusing impression of cybernetics. Heinz von Foerster There are many "giants" of cybernetics, but none so crystal-clear in thinking and writing as Heinz von Foerster. "Ethics and Second-Order Cybernetics", an amazing talk about the relationship of goal-directed systems to ethical behavior A lovely and illuminating interview with von Foerster Gordon Pask Eccentric beyond description and 30 years before his time, Gordon Pask has influenced generations of innovators, including Ted Nelson and Nicholas Negroponte (to name the most well-known figures). Unfortunately very little of Pask's own writing is on-line. "A comment on the cybernetic psychology of pleasure" is from the 1960s and only introduces concepts that are developed further elsewhere (see the first Off-Line reference, below). Beware an apparent mislabeling of the 2nd numbered list, which should be 'a,b,c and d' instead of '1,2,3 and 4.' Personal Links These links are from my site, which mainly chronicles work from the 1980s and 90s. They introduce ideas related to the Design Summit. A definition of cybernetics A summary of software implementations and approaches to user interface design The remainder of these links are included on the summary page just mentioned: A talk given at BayCHI specifically about goal-directed user interface design Some polemical remarks about software design An aging but still interesting prototype 'search tool' Ideas for moving 'Beyond Menus' Description of an adaptive hypermedia system based on cybernetics, and how it came about   IV. Off-Line References This paper describes in detail an interaction environment (that I mentioned at the Design Summit), where the user playing a musical instrument is engaged in a conversation with a mechanism that focuses the user's attention, gets bored, and otherwise stimulates the user in novel and unexpected ways. It is a fabulous proposal for a type of video game, though Pask built the original version of this system in the 1950s and carted it around England in vans for installation in music halls. Gordon Pask: "A Comment, a Case History and a Plan". In Cybernetic Serendipity, edited by J. Reichardt. Rapp and Carroll, 1970. Reprinted in Cybernetic Art and Ideas, edited by J. Reichardt. London: Studio Vista, 1971, 76-99. Special thanks to Hugh Dubberly for asking the first question, and Clarke Robinson for designing the graphics and handout.
© Copyright Paul Pangaro 1999. All Rights Reserved.