176 Environments of intelligence
the degree of habituation and accommodation that is required, the less profound
the changes in the user’s informational environment will be, and vice versa.
If we look from the other side of the interactive relation, from the system to
the user, the requirements will be partly complementary. Quite obviously, coming
to correctly anticipate and interpret human behaviours will be crucial to the system’s success. To the extent that an AR system falls short of correctly interpreting
human perceptions and actions, and to the extent that it does not correctly anticipate the effects of its own behaviours on human perception and action, it will be
confined to some kind of cognitive trial-end-error game. The resulting behaviour
is likely to create irritating effects on the side of the users which will have to be
dealt with but from which the system, if so capable, may also learn.
In the SL case, although one may teleport and fly in this environment, and
although avatars may assume the shape of animals, robots or even potted plants on
wheels, one expects to interact with other natural human beings. However, much
of the information that would be used in face-to-face interactions is not present
in this environment, or it is transmitted through channels that would otherwise be
complemented by information arriving through other channels (a gesture or wink
to see, an exaggerated intonation to hear). Moreover, SL residents encounter the
entire virtual environment in which they act with a sense of uncertainty towards
its ontological status that can hardly be fixed in-world. The domains of informational relations in that world, both in terms of the unequivocality and stability of
the correlations involved and in terms of the degree of isomorphism to natural
informational domains, remain in constant jeopardy.
In all cases under discussion here, if the conditions of ambiguity of the informational relations were selectively relevant in a biological sense, such ambiguity
would not persist for long. Still, actions or entire practices may fail in painful and
embarrassing ways. An evolving robot will be at a selective disadvantage when
being unable to resolve the ambiguity in edge detection from certain angles, so
his artificial genes are unlikely to be replicated in the next generation of the simulated evolutionary process. In the SL case, residents with different perceptions of
some game-relevant conditions, after a series of communicative and interactive
mishaps that they encounter as “newbies”, will be found either to quickly drop out
of SL altogether, or to coalesce with like-minded residents, save the occasional
clash with members of variant subcultures in this essentially social environment.
In turn, conversational agents and social robots are boundary cases in which
the counterpart of interaction is known not to be a human being but supposed to
be human-like in appearance and behaviour in important respects. It will be found
being treated unambiguously neither as a person nor as a machine – an ambiguity
that may turn out to be of somewhat more general relevance than in SL encounters. Even the most natural and familiar appearance would not warrant the agent’s
or robot’s integration into common modes of social interaction, as the interaction
remains asymmetrical, and as the counterpart’s artificiality remains discernible.
Some of the behavioural cues by which human beings identify each other as such
are present but, first, one cannot be sure that they convey information on any
sort of analogue of a real human being’s beliefs and desires. Second, whatever
the degree of habituation and accommodation that is required, the less profound
the changes in the user’s informational environment will be, and vice versa.
If we look from the other side of the interactive relation, from the system to
the user, the requirements will be partly complementary. Quite obviously, coming
to correctly anticipate and interpret human behaviours will be crucial to the system’s success. To the extent that an AR system falls short of correctly interpreting
human perceptions and actions, and to the extent that it does not correctly anticipate the effects of its own behaviours on human perception and action, it will be
confined to some kind of cognitive trial-end-error game. The resulting behaviour
is likely to create irritating effects on the side of the users which will have to be
dealt with but from which the system, if so capable, may also learn.
In the SL case, although one may teleport and fly in this environment, and
although avatars may assume the shape of animals, robots or even potted plants on
wheels, one expects to interact with other natural human beings. However, much
of the information that would be used in face-to-face interactions is not present
in this environment, or it is transmitted through channels that would otherwise be
complemented by information arriving through other channels (a gesture or wink
to see, an exaggerated intonation to hear). Moreover, SL residents encounter the
entire virtual environment in which they act with a sense of uncertainty towards
its ontological status that can hardly be fixed in-world. The domains of informational relations in that world, both in terms of the unequivocality and stability of
the correlations involved and in terms of the degree of isomorphism to natural
informational domains, remain in constant jeopardy.
In all cases under discussion here, if the conditions of ambiguity of the informational relations were selectively relevant in a biological sense, such ambiguity
would not persist for long. Still, actions or entire practices may fail in painful and
embarrassing ways. An evolving robot will be at a selective disadvantage when
being unable to resolve the ambiguity in edge detection from certain angles, so
his artificial genes are unlikely to be replicated in the next generation of the simulated evolutionary process. In the SL case, residents with different perceptions of
some game-relevant conditions, after a series of communicative and interactive
mishaps that they encounter as “newbies”, will be found either to quickly drop out
of SL altogether, or to coalesce with like-minded residents, save the occasional
clash with members of variant subcultures in this essentially social environment.
In turn, conversational agents and social robots are boundary cases in which
the counterpart of interaction is known not to be a human being but supposed to
be human-like in appearance and behaviour in important respects. It will be found
being treated unambiguously neither as a person nor as a machine – an ambiguity
that may turn out to be of somewhat more general relevance than in SL encounters. Even the most natural and familiar appearance would not warrant the agent’s
or robot’s integration into common modes of social interaction, as the interaction
remains asymmetrical, and as the counterpart’s artificiality remains discernible.
Some of the behavioural cues by which human beings identify each other as such
are present but, first, one cannot be sure that they convey information on any
sort of analogue of a real human being’s beliefs and desires. Second, whatever