Disembodied Intelligence
The sciences of cognition have tended to examine a disembodied intelligence,
a pure intelligence isolated from the world. It is time to question this approach,
to provide a critique of pure reason, if you will. Humans operate within the
physical world. We use the physical world and one another as sources of in-
formation, as reminders, and in general as extensions of our own knowledge
and reasoning systems. People operate as a type of distributed intelligence,
where much of our intelligent behavior results from the interaction of mental
processes with the objects and constraints of the world and where much be-
havior takes place through a cooperative process with others.
In the research areas studied by experimental psychologists, linguists, and
workers in the field of artificial intelligence, thought and understanding are
assumed to take place with little or no hesitation, little or no error, and little or
no doubt. Scientists make these assumptions in order to simplify their task.
‘‘After all,’’ they will state, ‘‘the phenomena we are studying are so complex
that it is essential to look at them first without all those other complicating
factors. Then, after we have understood the isolated case, we can move on to
the more realistic and complex situations.’’ The problem with this point of view
is that the so-called simplification may be making the task more difficult.
With a disembodied intellect, isolated from the world, intelligent behavior
requires a tremendous amount of knowledge, lots of deep planning and deci-
sion making, and efficient memory storage and retrieval. When the intellect is
tightly coupled to the world, decision making and action can take place within
the context established by the physical environment, where the structures can
often act as a distributed intelligence, taking some of the memory and compu-
tational burden off the human. To give one example: Linguists are continually
worried about the amount of ambiguity that exists within language. Ahuge
amount of scientific research has gone into developing schemes for under-
standing and trying to minimize this ambiguity. But the ambiguity almost al-
ways results from the analysis of single, isolated sentences: in real situations,
where several interacting people deal with real events, the sentences usually
have only one meaningful interpretation. Actually, even when communications
are ambiguous, they are usually not perceived as such by either speaker or lis-
tener, even though both may have different interpretations of the meaning. It is
this lack of perception of ambiguity that is important, and it derives from the
communicative, social nature of language, something that is entirely missed
when the language is studied as isolated, ‘‘simplified’’ printed sentences or
utterances, completely abstracted from the real, social setting.
Information in the world can be thought of as a kind of storehouse of data.
This has many advantages. The world remembers things for us, just by being
there. When we need a particular piece of information, we simply look around,
and there it is. Do I need to repair my car? I don’t have to remember the exact
shape of the part, because when the time comes for me to do the task, the shape
is there in front of me. This eases the burden on initial data collection, eases
the requirements on learning and memory, and avoids the need for complex
indexing or retrieval schemes. Moreover, it guarantees that the values so
obtained will be the most timely available at the moment of need.
448 Donald A. Norman