| dc.description.abstract | Abstract
Information theory (Shannon, 1948) and relative similarity (Rosch and Mervis, 1975) are
used to investigate the problem of how people communicate about ambiguous, unstructured
stimuli. Specifically, the following one-way communication setting is considered: given a set
of ambiguous items available to both a sender and a receiver, the sender uses messages to
describe one of the items from the set, such that the receiver is able to identify this target
item correctly. It is argued that information reduces uncertainty in ambiguous communication
settings through the development of structure, which is conceptualized in terms of a system
of categories. During the communication process, distinguishing an item from a set of other
items involves clustering items into subsets; that is, grouping items together that are similar
to one another and leaving out items that are dissimilar. It is proposed that relative similarity
is the cognitive mechanism involved in the development of those categories. It is further
proposed that perceptions of relative similarity are made with respect to one attribute. People
take advantage of the perceived structure of the ambiguous stimulus and form categories in a
goal-directed manner, focusing on whatever attribute best enables them to distinguish the
target item from the others (Barsalou, 1983). Each time one attribute is used in
communication to refer to a cluster or subset of items, uncertainty is reduced. Furthermore, it
is postulated that when people have the choice of one attribute from multiple possible
attributes to communicate about unstructured stimuli, two categorization logics operate to
reduce uncertainty. Hypothesis 1 predicts that people maximize distinctiveness by choosing
the attribute that allows for a larger gap (i.e., greater dissimilarity) between the subset
containing the target item and the subset of remaining items. Hypothesis 2 predicts that
people maximize information gain by choosing the attribute that allows for the smaller subset
containing the target item. The theoretical framework is built upon illustrative examples.
Three experiments were conducted to investigate the cognitive mechanisms that
people use to identify and describe a specific ambiguous item from among a set of
ambiguous items during communication. Participants took the role of either sender or
receiver in a one-way communication situation. In the role of sender they ranked a set of
descriptions/attributes based on the degree to which they thought the descriptions would
enable an imagined receiver to identify a specific target item from a set of ambiguous items.
As receiver, they were given specific descriptions/attributes from an imagined sender, and
ranked a set of ambiguous items based on which of the items they thought the sender was
referring to. Experiments 1 and 2 used two different kinds of unstructured stimuli;
experiment 3 utilized a structured way of manipulating unstructured/ambiguous stimulus.
The results of experiments 1 and 2 were consistent with hypotheses 1 and 2. There were
strong consistency between the two different stimuli and, therefore, strong evidence for
generalizability of the observed effects. The results of experiment 3 gave inconsistent support
for the hypotheses. The theoretical framework, the design of the experiments, and the results
are discussed. | en |
