Information Technology Reference
In-Depth Information
A large amount of AI research is justified or motivated by pragmatic goals
and there may in fact be pragmatic goals that would justify why we need a new
technology of story understanding. In contrast, the poetics of AI have almost
always been articulated around the need to get to know ourselves better. This
poetics of the design and construction of intelligent, non-human entities has
long been a theme of science fiction and science fantasy (not to mention its im-
portance in philosophy since at least the time of Socrates when it was expressed
as the ethical imperative “Know yourself.”) Sherry Turkle nicely illustrates the
how AI programs can function as a “second self ” (Turkle 1984). It is within this
tradition of poetics - what the philosopher Michel Foucault has described as
“technologies of the self ” (Foucault 1997) - that I would argue we need a new
technology of story understanding. As new narrative forms are developed and
new media proliferate, we need to invent new means for understanding how
and where we are located in the emerging social networks.
.
Methodology
Methodology = Computational Sociolinguistics = Computational Linguistics
+
Quantitative Sociology
.
Within the field of sociology, a number of computational approaches to
understanding the social significance of literatures have been developed. Most
prominently these methods have been applied to the literatures of science. For
example, the methods of co-citation analysis (Garfield 1979) are routinely ap-
plied to determine the relative importance of a scientific article: its significance
is thought to be a function of the number of other articles that cite it. AI elab-
orations of the techniques of co-citation analysis include (Lehnert et. al. 1990).
The methods of social network theory (Wasserman & Galaskiewicz 1994) and
actor-network theory (Callon et. al. 1986; Latour & Teil 1995) provide tech-
nologies akin to co-citation analysis, but have their own particular strengths
and weaknesses. Co-word analysis, the computational technique associated
with actor-network analysis, is basically the calculation of mutual probabili-
ties between nouns in scientific abstracts and so this technique probably has
more affinities with techniques in computational linguistics than with those
developed by other sociologists.
These sorts of sociological “story understanding” technologies are very dif-
ferent from the story understanding technologies of an older, symbolic AI, but
they have some affinities with techniques of newer AI work in agent-based ar-
chitectures for information filtering and recommendation. Thus, for example,
