Rabbit

Uit Aardnoot
Versie door Martien (overleg | bijdragen) op 21 okt 2008 om 12:33 (Nulde versie.)
(wijz) ← Oudere versie | Huidige versie (wijz) | Nieuwere versie → (wijz)
Naar navigatie springen Naar zoeken springen


One of my long-term favourite examples of a system that attempts to provide a soft, appropriate technology is illustrated by a computer software system developed by Michael Williams and Frederich Tou, then at the Xerox Palo Alto Research Center. This system, which was named “Rabbit”, attempts to provide information that you need to make a decision. Unlike other such services, it does not assume that you know anything at all about the way it stores its information. Instead, you simply request whatever you think is relevant, and the system will respond with samples.

Suppose you were hungry, in a strange town, and wanted a restaurant. You came across Rabbit. In most systems, you would have to provide a precise query, and so you would first have to learn how the system was organized. Not so with Rabbit. You could start off simply by saying, "I would like to go to a restaurant." This is really an inadequate request, because it does not specify the location, type of food, and other variables based on which the system expects to select. Worse, there might be a thousand restaurants in the database. How would the system know what to do? Most systems would either reject the request or provide all thousand entries. Neither is satisfactory.

Rabbit, however, reverses the normal interaction. Rabbit is not trying to choose for you, but rather help you choose for yourself. So if you simply ask for a restaurant, Rabbit will give you one, any one. It doesn't even matter what restaurant it gives you, because it is teaching you about the kinds of information it needs. Suppose it responds to your general query with a very formal Chinese restaurant located on the other side of the city that requires advance reservations:

Name: Chan Wo
Location: 1800 W. Far Side Drive
Price: €€€€ (very expensive)
Formality: Suit required
Type of food: Chinese
Service: Superb
Reservations: Highly recommended
Credit cards: No

You may still not know what you want, but one thing you do know is that this is not it. Good! Rabbit works this way deliberatgely, for that it can learn your desires by taking account of the things you don't like. Either you can simply reject this choice and ask for another, or better yet, you can use the restaurant description as a starting point by indicating just what features of this selection you like, dislike, or are indifferent to. The way this works is that you can select any of the descriptive dimensions and modify it. Rabbit helps by giving a set of alternatives: For _Location), it can list the areas of the city and let you say for each area whether it is acceptable, not acceptable, or doesn't matter. For _Type of food_, once you saw the list of possibilities, you could rule out Chinese while accepting Thai, rule out French while accepting Northern Italian, and say "Don't care" regarding the rest.

Location: I want the East Side.
Price: Must be inexpensive (€).
Formality: I prefer informal.
Type of food: -
Service: Don't care.
Reservations: Must not be required.
Credit cards: Must be allowed.

Rabbit is educating you, without you realizing it, but on your own terms, in your own language. The suggested alternatives are structured to teach you about the way Rabbit organizes them, allowing you to adjust and use its organizational structure to help refine your search. "Oh," you might think, "credit card. I forgot about that." You then tell Rabbit, "I don't have much cash, so only show me the places that accept credit cards." Notice too that the system doesn't ever have to get a unique answer. At some point, you could simply ask for several examples, read them, and decide among them, never telling Rabbit your decision.

Although Rabbit never asks its user to present a request using the language of logic, Rabbit is internally constructing its own logical expressions to describe your likes and dislikes. That is fine, for logic offers an important precision and value to the computer system. Logic is not the appropriate tool for most people, but it is quite appropriate for machines.

Rabbit provides a positive example of a person-centered, appropriate technology. Give control to us as individuals, in our language, and on our terms. The technology does the translation from its internal machine-centered logic to a form appropriate for the person. Rabbit was only used in the research laboratory as a demonstration: It has never been used in a commercial system. Too bad.


The Future (according to The Rock):

  • Provided the system that exactly allow Rabbit's behaviour to be used in many commercial systems.
  • Make these scenarios ”affordable”: "The affordances of an object refers to its possible functions: A chair “affords” support, whether for standing, sitting, or the placement of objects. A pencil affords lifting, grasping, turning, poking, supporting, tapping, and of course, writing. In design, the critical issue is perceived affordances: what people perceive the object can do. We tend to use objects, suggested by the most salient perceived affordances, not in ways that are difficult to discover (hence the fact that many owners of electronic devices often fail to some of their most powerful features--indeed, often do not even know of their existence.", Donald Norman, Things that make us smart, 1994, page 105.
  • Foster a peer-2-peer or crowdsourcing network to help you make even better and quicker decisions by including preferences and experiences from peers.
  • Connected enterprises (restaurants in this example) only show up if they still have room, showing availability, by having their reservation system connected to Rabbit's services.



Bron: rabbit uit Things that make us smart, pagina 240.