Tacitly Communicating
with Smart Environments
via our Practical Behavior and its Traces
Cristiano Castelfranchi
CNR - Institute of Cognitive Sciences and Technologies
[email protected]
EBICC-Campinas’09 - Castelfranch
The thesis
A crucial part of the "intelligence" that we need in
our environment (like from our domestic robots) is a
specific form of social intelligence:
the ability to "read" our behavior and its
traces in terms of our actions, intentions, and
assumptions.
•
From this ‘intelligent’ Observation of the behavior a
special kind of Communication emerges: Behavioral
Implicit Communication (BIC).
EBICC-Campinas’09 - Castelfranch
because our claim is that
BIC is the most basic form of communication
from the analytical point of view,
and also
the most primitive
(both in evolutionary and in developmental sense).
BIC plays an irreplaceable and underestimated role in human interaction
and coordination, social order, cultural transmission,
and we do expect
an important role of BIC in social Agents, robot-robot coordination, and in
H-Agent and H-Robot interaction, and human smart-environment
interaction.
EBICC-Campinas’09 - Castelfranch
To define BIC making clear
• the fundamental
communication
distinction
between
signification
and
• why its is false that all behaviors in social contexts are
communication;
• why BIC has nothing to do with gestures and expressive
movements (the so called Non-Verbal-Communication);
• the ‘transition’ steps from non-communicative behavior to
intentional BIC;
• a few examples of how crucial BIC is in human coordination
and interaction will be provided.
It will clearly emerge
• Human capacity for behavior/mind-reading has been a
cognitive prerequisite for intentional communication.
• How this is a prerequisite for any efficient ‘coordination’
EBICC-Campinas’09 - Castelfranch
Is any behaviour communication?
EBICC-Campinas’09 - Castelfranch
Against Watclawicz: Damned to communicate?
 When behavior is communication and when is not.
 Is simple understanding already communication?
A famous thesis of Palo Alto psychotherapy school was that
"It is impossible do not communicate",
"any behaviour is communication" in social domain.
In this view, a non-communicative behaviour is a nonsense.
This claim is too strong. It gives us a notion of communication that
is useless because is non-discriminative.
In order to have communication having a "recipient" which
attributes some meaning to a certain sign or cue is a nonsufficient condition.
EBICC-Campinas’09 - Castelfranch
We cannot consider as communication
any information/sign arriving from A to B,
unless it is aimed at informing B.
A teleological (intentional or functional) "sending" action by the
source is needed.
The source has to perform a given behaviour "in order" the other
agent interprets it in a certain way, receives the “message” and its
meaning.
EBICC-Campinas’09 - Castelfranch
Is an escaping prey “communicating” to its predator/enemy its
position and move?
Watzlawich’s overgeneralization cannot avoid considering communication to
the enemy the fact that a predator can observe the movement of the prey.
Although this information is certainly very relevant and informative for
the enemy or predator, it is not communication.
Receiving the information is functional (adaptive) for the predator and for
that species which have developed such ability, but it is not functional at all, it
isn’t adaptive for the prey:
“Sending” that sign is not a functional (evolutionary) goal of the prey.
Is a pilferer informing or communicating to the guard about his presence and
moves?
Are a killer’s signs (very meaningful for the police) messages to it?
EBICC-Campinas’09 - Castelfranch
“Signification” – following (Eco, 1976)
 prints on the ground are signs for the hunter of the passage of
a deer
 smoke is the sign of a fire,
 some spots can mean "it is raining" (they are for Y signs of
the fact that it is raining);.
We have here simple processes of signification.
Notice that meanings are not conventional but simply based
upon natural perceptual experience and inference.
Notice also that the signal, the vehicle has not been
manufactured on purpose for conveying this meaning, it
doesn’t need to be “encoded” and “decoded” via some
conventional artificial rule.
EBICC-Campinas’09 - Castelfranch
From Observation & Signification
To Communication
EBICC-Campinas’09 - Castelfranch
1. Observation, and more specifically ‘signification’
(the capability to interpret and ascribe meaning to
observed facts)
is the basis of a very crucial form of Communication without
words or special protocols.
2. Efficient coordination – in humans but also in artificial
Agents – exploits or should exploit not just ‘observation’ but
more precisely this form of silent communication:
when Agent X relies on the fact that Agent Y is
observing her in order to let Y understand that p, i.e. for
communicating to Y that p.
The theory of this form of Communication and of its
relevance
EBICC-Campinas’09 - Castelfranch
Behavioral Implicit Communication Theory
Usual, practical, even non-social behaviors contextually used as messages for
communicating.
 Behavior can be communication
without any modification or any additional signal or mark.
I will call this form of communication without specialized symbols:
Behavioral - Implicit Communication (BIC).
“Behavioral” because it is just simple non-encoded behavior.
“Implicit” because – not being specialized and codified – its
communicative character is unmarket, undisclosed, not-manifest, and
thus not undeniable;
Normally communication actions are special and specialized behaviors (like
speech acts, gestures, signals, …).
EBICC-Campinas’09 - Castelfranch
I will illustrate
• The necessary steps to arrive from mere observation and
exposure to observation to Behavioral Communication, and why
this is crucial for mutual understanding and for commitment and
norms keeping, imitation and learning, etc.
• The message-sending paradigm dominating CSCW, MAS, HCI,
and H-Robot-I, is criticized; necessity and advantages of BIC for
coordination and cooperation are presented.
• The relationship with the ill-defined but important notion of
stigmergy (very used in ALife Agents) is analyzed: stigmergy results
to be a sub-case of BIC.
EBICC-Campinas’09 - Castelfranch
What Behavioral Implicit Communication is
…and what is not!
NOT Non Verbal Communication or Expressive Behavior, i.e. gestures,
conventionalized signals, facial expressions, and so forth.
All this sort of specialized messages (either cultural or inborn) are different
from simple practical actions like walking or sitting.
Usually, communication is based on specialized signals linked to specific meanings.
In order to use these signals, either we have to learn them or we refer to some sort of
innate knowledge.
In Behavioral Implicit Communication
there are NOT specialized signals,
but the practical behavior itself is the message.
EBICC-Campinas’09 - Castelfranch
The BIC message presupposes a more primitive and basic substrate
due to ‘observation’:
the unilateral capability of the agent to observe the other’s behavior and to “read” it,
to understand what she is doing, what she intends and plans to do (her
goals),
or at least (in rule-based systems or neural net systems)
to predict and expect her next position or action using this information
for example for “anticipatory coordination”
In other words,
communication is based on and exploits ‘signification’
(the semiotic ability of cognitive agents; for example the ability to take ‘smoke’ as a
sign of ‘fire’, or to ascribe thirst to a drinking agent)
that goes beyond simple perception but it not necessarily used in and for
communication.
EBICC-Campinas’09 - Castelfranch
Steps from Observation to Communication
0.
No Communication at all; simply exposure and observation with
interpretation activity by the observer ('signification').
1.
The weakest form of BIC:
the agent is not acting in order to let the other understand what she is
doing; she is simply aware of this possible result of her behaviour and lets it
happen. No real intention, the communicative result is just a known (not
desired) effect of the action.
2.
Side BIC:
The agent anticipates that the action will have also a communicative meaning
(additional result) and she likes this (it is a goal of her), but his action is not aimed to
reach this communicative goal; not motivated by it. The expected side-effects that I
let happen is a goal but neither necessary nor sufficient for my action.
EBICC-Campinas’09 - Castelfranch
3. True (intentional) BIC:
Communicating is part of my aims and motivates the action.
___________________________________________________________
The important point is that:
we have here a fully intentional communication act,
but without
the aim (intention) that the other understands
that X intends to communicate (by this act).
• ‘Intention of communicating’ and
‘communicating (this) intention’
are not one and the same thing.
In Grice-inspired view of ‘communication’ these two very different things are usually
mixed up
OK for Language, but not for all forms of Communication
EBICC-Campinas’09 - Castelfranch
Behavioral Implicit Communication Theory
_________________________________________________________________________
•
This is the 'intentional' path in BIC evolution, when the
communication is intentional;
more suitable for BDI-like Agents.
•
more primitive forms of BIC, based on reinforcement learning or
on selection
fit well with reactive or ALife Agents.
EBICC-Campinas’09 - Castelfranch
We have two basic kinds of communication:
Intentional (or better "goal-governed") communication and
Functional (or "merely goal-oriented") communication.
FUNCTIONAL BIC
with several sub-type: by evolution-selection; or by design; or by reinforcement learning based on the
effects
INTENTIONAL BIC
Intentional BIC usually presupposes an intentional stance and more precisely a
«theory of mind» not only in the sender but also in the interpreter,
since the message bring by the action can be about the mind of the source: his
intention, or emotion, or motives, or assumptions, etc.
EBICC-Campinas’09 - Castelfranch
Both: Intentional vs. Functional BIC
________________________________
CLAIM:
In Human-Human, HCI, H-Agent, H-Robot, Ag-Ag,
interaction, H-Smart Environment Interaction
Robot-Robot
the possibility of communicating through an action is very interesting:
•
the solution for several coordination problems, and,
•
the interaction becomes natural and intuitive.
via Evolution, Learning or Reasoning and Planning
EBICC-Campinas’09 - Castelfranch
X’s  aimed at
producing  or
X’s side
Y’s side
Y’s perception
of X’s behavior

perception
Y’s reaction  to
or its
‘recognition’ and
‘interpretation’ 
X’s perception of 
or

4

Y
5
to 
3
2
1
EBICC-Campinas’09 - Castelfranch
4. meta-BIC:
In meta-BIC, there is meta-communication,
typical of higher forms of communication like language.
BIC meta-message is as follows:
"this is communication, this is a message not just
behavior; it is aimed at informing you".
Frequently BIC has such a high level (Grice’s way) nature.
For example the act of giving or handing is not only a practical one,
but is a meta-communicative act where X intends that Y
understands that she is putting something closer to Y in order Y
(understanding that she intends so) takes it.
EBICC-Campinas’09 - Castelfranch
5. Beyond BIC:
The last step is when
the behavior is performed only for communication and looses
its practical purpose (or even some practical effect).
In the last case, the act either is
- just faked (simulation, bluff) (which is very important in
conflict coordination es), or:
- it is just a ritual, i.e. the action has fully become Non Verbal
Communication or a conventional symbolic gesture with a
practical origin.
EBICC-Campinas’09 - Castelfranch
STIGMERGY
stigmergy is communication via long term traces, physical
practical outcomes (objects), useful environment modifications,
not mere signals.
Stigmergy is defined as “indirect communication through the environment”.
(Holland and Beckers)
A wrong definition!!
Stigmergy is just a sub-case of BIC,
since in fact any BIC is based on the perception of an action that necessarily means
the perception of some “trace” of that action in the environment (for example air
vibrations).
EBICC-Campinas’09 - Castelfranch
STIGMERGY
Stigmergy is not only for insects, birds, or non-cognitive agents.
There are very close examples also in human behavior.
In animals stigmergy is non intentional, but intentional forms of it are
possible.
Consider:
• bestsellers
• turning the lights or TV on in one’s house when one is going out, as a (deceptive) ‘signal’
for possible intruders.
• a sergeant that – while crossing a mined ground –says to his soldiers: “walk on my
prints!”. From that very moment any print is a mere consequence of a step, plus a
stigmergic message to the followers (descriptive “here I put my foot” and imperative “put
your foot here!”).
Human merely ‘functional’ forms: ex. Paths
EBICC-Campinas’09 - Castelfranch
Stigmergy in humans:
some nice example with deontic components
Leaving the coat on the seat
“already taken, not free”. This is a sign, deliberately used for meaning
(signalling) this. It's communication.
Leaving an object on a given place
“this is its right place”. This is a sign, deliberately used for meaning
(signalling) this. It's communication.
Parking marks
They are not merely messages; in fact, they cannot be replaced by a simple poster
illustrating the prescribed car disposition in that street.
They also have the practical function of visual reference point in the manoeuvre
to be used during the act of parking.
>> Prescribing and teaching by doing
(instructions by imitation ; see later)
EBICC-Campinas’09 - Castelfranch
Stigmergy in humans: playing soccer
BIC actions + Stigmergic Communication Through the Ball
OBS ERVATION
COMMUNICATIO N
EBICC-Campinas’09 - Castelfranch
BIC
Coordination and Social Order
- BIC and coordination
- BIC and implicit agreement
- BIC and conventions
- BIC and normative behavior
EBICC-Campinas’09 - Castelfranch
HOW DO GROUPS COORDINATE
THEIR COLLECTIVE ACTIVITY?
Mainly by exploiting mutual observation, behavior interpretation,
and outcomes and traces: that is
by sending behavioral and stigmergic messages
Ex.
Soccer
Interference in a corridor
Sitting at table,…
EBICC-Campinas’09 - Castelfranch
MOVEMENTS + NOICE
EBICC-Campinas’09 - Castelfranch
Coordination is that additional part or aspect of the
activity of an agent specifically devoted to deal and cope
with dynamic environment interferences.
Living in a common world entails that there exists
‘interference’ between the behaviors of the agents (the
action of an Agent Y could affect the goal of another Agent
X);
 ‘interference’ can be either ‘negative’ (Y’s action creates obstacles
to X’s action or damage her goals)
 or ‘positive’ (Y’s actions realize X’s goals or create opportunities
for her actions).
Thus X has to perceive (or infer) those ‘interferences’ in
order to avoid or exploit them.
EBICC-Campinas’09 - Castelfranch
The basic forms of coordination are:
Unilateral: X just coordinates her own behavior with Y’s or environmental
dynamics, ignoring Y’s coordination or non-coordination activity.
Bilateral: X coordinates his behavior with Y’s observed behavior; and Y does the
same. Bilateral but independent: X coordinates his behavior with Y’s observed
behavior; and Y does the same in an independent way.
Reciprocal: X coordinates his behavior with Y’s behavior by taking into account the
fact that Y is coordinating her behavior with X’s behavior.
Mutual: it is based on symmetric and interdependent intentions and mutual
awareness (shared beliefs). Both X and Y wants the other to coordinates with his/her
own behavior and understand that s/he intends to coordinate with the her/his own
behavior.
EBICC-Campinas’09 - Castelfranch
(X) = possible; X = necessary
Observation
based
BIC
BIC
(X)
COORDINATION
S
O
C
I
A
L
UNILATERAL
BILATERAL
RECIPROCAL
MUTUAL
X
X
X
(X)
(X)
(X)
X
EBICC-Campinas’09 - Castelfranch
Coordination is possible without any communication both in human and
artificial societies (Franklin).
However, usually coordination exploits communication.
Since BIC is
i) a very economic (parasitic),
ii) a very spontaneous,
iii) a very practice and rather effective form of communication
just exploiting side effects of acts, traces, and the natural disposition of agents to
observe and interpret the behavior of the interfering others,
a rather important prediction follows.
‘Agents’ acting and perceiving in a common world
will use a lot of BIC and will spontaneously develop it
Actually a very large part of communication for coordination in situated and
embodied agents exploits reciprocal perception of behavior or of its traces and
products; i.e. it is just BIC.
EBICC-Campinas’09 - Castelfranch
HOW DO GROUPS ESTABLISH
THEIR "AGREEMENTS” ?
Groups (in strict sense) and group activities are
based on members’ agreement (CollInt)
However, such an agreement and commitment is rarely achieved by
an 'explicit' negotiation and communication (Scanlon).
Usually the real mechanism is a 'tacit consensus'
but
• HOW TO ACHIEVE A CONSENSUS 'TACITLY'?!
• HOW CAN WE KNOW THAT THERE IS A CONSENSUS?
The answer is
via BIC:
when doing or not doing is communicating
EBICC-Campinas’09 - Castelfranch
“Qui Tacet Consentire Videtur”
Tacit agreement are based on a lot of common knowledge or at least
of shared beliefs
but also on two basic implicit communication acts:
 x’s “proposal” (or request of consent):
by starting doing b (and checking that y is checking) x intents
to implicitly inform y that he intends to do b and that he is
seeking for y’s agreement;
 y’s “agreement” (or permission/consent):
by not-opposing y intends to implicitly inform x that she agrees
with his doing b.
EBICC-Campinas’09 - Castelfranch
Fulfilling Social Commitments as BIC
“ I did it”
• Psychiatric patient shows to the nurse that he is drinking his
drug
• My secretary leaves the letter on my desk: “Done!”
The satisfaction of social commitments and obligations
EBICC-Campinas’09 - Castelfranch
Obeying Norms as BIC
One of the functions of norm obedience is the confirmation of the norm
itself, of the normative authority of the group, and of conformity in
general (E. Posner; Conte)
•
thus one of the functions of norm obeying behaviours is that of
informing the others about norm obedience.
x is worrying about social monitoring and sanctions or seeking for
social approval, and he wants the others see and realize that he is
obeying the norms.
Of course, X can also simulate his respect of the norms
EBICC-Campinas’09 - Castelfranch
• Imitation-BIC
as convention establishment and memetic agreement
Imitation (i.e. repeating the observed beahavior of Y – the model)
has several possible BIC valences
Communicative goals:
a) In learning-teaching via imitation. X communicate to Y “I’m trying to do like you;
check it: it is correct?” (also practically teaching, ‘demonstration’)
b) In convention establishment and propagation. “I use the same behavior as you, I
accept (and spread) it as convention; I conform to it”.
c) In imitation as emulation and identification: “I want to be and to behave like you”
d) In imitation as membership: “I want to be and to behave like you; since I’m one of
you”.
Will Artificial Intelligence creatures learn by imitation??
YES, for improving performace, for conformity and social order, for establishing
conventions, etc.
EBICC-Campinas’09 - Castelfranch
BIC basement of Social Order & Interaction
BIC has a privileged role in social order,
in establishing commitments,
in negotiating rules,
in monitoring correct behaviors,
in enforcing laws,
in spontaneous emergence of behavioral conventions
If there is a ‘Social Contract’ at the basement of society
it has been established by BIC
and is just tacitly signed and renewed
EBICC-Campinas’09 - Castelfranch
Back to Human-Artifacts Interaction
EBICC-Campinas’09 - Castelfranch
This kind of observation-based, non-special-message-based
communication should be much more exploited in CSCW and
computer/net mediated interaction, in Multi-robot coordination,
in Human-robot coordination, in MA systems.
• CSCW:
-allowing BIC messages,
or
- Agents reading Y’s BIC messages (actions of Y) and sending explicit messages to the user X
• Virtual environments, and software Agents:
- providing Observability! And “behavior reading”
• Multi-robots
- situated, physical communication
•Smart Environments
- situated, physical communication, not boring verbal
interaction only
EBICC-Campinas’09 - Castelfranch
Ambient Intelligence
will necessarily enable this form of communication.
• to establish with our intelligent, proactive,
cooperative environment a good
‘understanding’ of what we are doing and a
good coordination;
• to tacitly negotiate agreements and
conventions;
for example, about the habitual location of people or
objects, preferences, or habits and practices.
EBICC-Campinas’09 - Castelfranch
Ambient Intelligence
This form of ‘social intelligence’ and of tacit ‘communication’ is a
necessary condition for an active environment to coordinate its
activities and changes with our behavior in an effective way.
Not only
• to avoid interferences and obstacles to our situated activity,
but possibly
• to be cooperative and even pro-active, and for favoring our
activity by removing obstacles, creating the needed conditions,
anticipating our steps, and
deeply co-operating and collaborating with us;
not: “orders --> execution”
EBICC-Campinas’09 - Castelfranch
Ambient Intelligence
Ideally, this advanced capacity requires the adoption of an “intentional stance” towards
The future smart
environments will be able to observe our behavior, and to
understand and anticipate it.
humans and some capability of mind reading as part of its intelligence.
HOWEVER,
It has been less emphasized that once this form of ambient intelligence
will be achieved, humans will be able to exploit it in new ways: i.e.
by performing our actions while knowing and expecting that the
environment will notice and understand what we are doing.
Our behavior - and its physical traces - will thus become a “message”, a
“signal” sent to the environment itself in order to obtain collaboration,
although remaining a concrete practical action, not symbolic gestures or
mimics.
EBICC-Campinas’09 - Castelfranch
We sketch a computational framework to formalize the basic
cognitive ability that the environment should display. (Lorini, Pezzulo)
the formalism of graphical models, and in particular Bayesian networks
which couple the power of a sound statistical formalism for reasoning
under uncertainty, with the intuitiveness of graphical representations.
“Intentional stance”: attributing to the human hidden (latent)
cognitive variables, namely beliefs (B), intentions (I), and actions (A),
instead than simply monitoring the observable effects of their actions
(O).
EBICC-Campinas’09 - Castelfranch
a generative process, in that latent variables are estimated based on observables.
Indeed, within the Bayesian framework, an observable Ot can be considered a measurement
of At, which can be estimated by computing the posterior probability P(A|O).
Standard statistical (Bayesian) inference methods permit to do so efficiently. The same
techniques also permit to predict P(At+1|At) and to estimate I and B in addition to A.
EBICC-Campinas’09 - Castelfranch
SORRY
for mainly presenting the theory
but I think that it is really important and deserves a specific attention:

in Cognitive Science (the most basic form of (intentional) communication, with
specific cognitive requirements)
 in Semiotics
 in the Social sciences (the strongest basis for norm emergence and tacit agreement
in social order)
 in ICT: a natural form of interaction that cannot be missed.
Of course, several important antecedents:
Semiotics; “Signaling” theory (Schelling, Spence); in Ethology (Hinde, etc.); Goffman’s
theory of social representation; something in NVB theory, etc.
EBICC-Campinas’09 - Castelfranch
to remain silent might BIC-mean:
“I have finished”
EBICC-Campinas’09 - Castelfranch
to remain silent might BIC-mean:
“I have finished”
but I prefer an explicit message:
EBICC-Campinas’09 - Castelfranch
to remain silent might BIC-mean:
“I have finished”
but I prefer an explicit message:
the
END
EBICC-Campinas’09 - Castelfranch
The Stigmergic Over-generalization
The notion of Stigmergy comes from biological studies on social insects, and more
precisely the term has been introduced to characterize how termites (unintentionally)
coordinate themselves in the reconstruction of their nest, without sending direct messages
to each other.
Stigmergy essentially is the production of a certain behaviour in agents as a consequence
of the effects produced in the local environment by previous behaviour.
This characterization of Stigmergy is not able to discriminate between simple
signification and true communication, and between prosocial and antisocial behavior.
• prey-predator coordination and
• a pilfer (unintentionally) leaving footprints very precious for the police.
In order to have “communication”, it is not enough that an agent coordinates its
behavior with the behavior or thanks to the traces of the behavior of another agent.
____________________________________________
Stigmergy is defined as “indirect communication through the environment”.
(Holland and Beckers)
A wrong definition!!
EBICC-Campinas’09 - Castelfranch
STIGMERGY 2
The real difference is that in Stigmergic communication we do not have specialized
communicative actions, specialized messages (that unambiguously would be “direct”
messages because would be just messages);
we just have practical behaviors (like nest building actions) and objects, that are also
endowed with communicative functions.
In this sense communication is not “direct” (special communicative acts or objects) and is
“via the environment” (i.e. via actions aimed at a physical and practical transformation of
the environment).
stigmergy is communication via long term traces, physical
practical outcomes, useful environment modifications, not mere
signals.
To be true, perceiving behavior is always perceiving traces and environmental modifications due
to it; the distinction is just a matter of perception time and of duration of the trace.
Stigmergy is just a sub-case of BIC,
since in fact any BIC is based on the perception of an action that necessarily means
the perception of some “trace” of that action in the environment (for example air
vibrations).
EBICC-Campinas’09 - Castelfranch
STIGMERGY 3
Stigmergy is not only for insects, birds, or non-cognitive agents.
Also for software Agents
There are very close examples also in human behavior.
• In animal stigmergy is non intentional, but intentional forms of it are
possible.
Consider
• a sergeant that – while crossing a mined ground –says to his soldiers: “walk on my prints!”.
From that very moment any print is a mere consequence of a step, plus a stigmergic (descriptive
“here I put my foot” and prescriptive “put your foot here!”) message to the followers.
• the double function of guard-rails: on the one side they physically prevent car from invading
the other lane and physically constrain their way, on the other side they in fact also communicate
that “it is forbidden to go there” and also normatively prevent that behavior. This is what law
theorist call materalization of the norm: the norm its “hardwired” since either the external or the
internal condition for the agent’s doing differently are excluded; there is no possible decision to
violating. In fact the communicative function is a parasitic effects of the practical act and of its
long-term physical products.
EBICC-Campinas’09 - Castelfranch
BIC for coordination (in particular Stigmergy) has some nice properties and
advantages
BIC is naturally and intrinsically ‘situated’ merged in concrete location, time,
objects, thus it transmits this kind of information in a perceptual, immediate way
without any special, abstract, arbitrary codification.
The receivers would be obliged by more explicit and symbolic form of communication
to reinterpret those signs and to re-apply them to the concrete context, via deictic
reference. While using BIC the addressee extracts the information directly from the
environment (while/when ‘observing’ it) and it is ‘ready to use’ in that
environment.
This is also a possible memory charge advantage.
This information has also the very nice property of not being discrete, digitalized,
that sometimes might be critical.
Suppose that you have to move a heavy table with another guy: you will use the table itself as a
coordination device, and the physical sensations (that you know that the table will give to the other
and that the other will take into account for adjusting his behavior) as messages to him. He will
‘feel’ the direction and the acceleration that you impose to the table; he will feel the evolving
equilibrium . Imagine that on the contrary you have to give him verbal instructions and
reinsegnments. The messages might be not enough precise, or not enough fast, or to be cognitively
interpreted at the symbolic level before being translated in motor-commands; etc.
EBICC-Campinas’09 - Castelfranch
BIC for coordination (in particular Stigmergy) has some nice properties and
advantages
-no costs
-spontaneous, parasitic
BIC is naturally and intrinsically ‘situated’ merged in concrete location, time,
objects, thus it transmits this kind of information in a perceptual, immediate way
without any special, abstract, arbitrary codification.
The receivers would be obliged by more explicit and symbolic form of communication
to reinterpret those signs and to re-apply them to the concrete context, via deictic
reference.
This information has also the very nice property of not being discrete, digitalized,
that sometimes might be critical.
Suppose that you have to move a heavy table with another guy: you will use the table itself as a
coordination device, and the physical sensations (that you know that the table will give to the other
and that the other will take into account for adjusting his behavior) as messages to him. He will
‘feel’ the direction and the acceleration that you impose to the table; he will feel the evolving
equilibrium . Imagine that on the contrary you have to give him verbal instructions and
reinsegnments. The messages might be not enough precise, or not enough fast, or to be cognitively
interpreted at the symbolic level before being translated in motor-commands; etc.
EBICC-Campinas’09 - Castelfranch
Fulfilling Social Commitments as BIC
another kind of demonstrative act,
intended to show that one have done the expected action.
A Social-Commitment of X to Y of doing the act, in order to
be really (socially) fulfilled, requires not only that agent X
performs the promised action , but also that the agent Y knows
this.
Thus the performance of the act is also aimed at informing
that it has been performed!
(If there are no explicit and specific messages) any act of SCommitment fulfilment is also an implicit communication act
about that fulfilment.
EBICC-Campinas’09 - Castelfranch
BIC theory looks quite relevant in several application domains in
IT, especially with Agents or Smart Environments, and for several
important issues like:
-the problem of social order and social control in MAS, CSCW, virtual
organizations.
-One cannot believe that social order will be created and maintained mainly by
explicit and formal norms, supported by a centralized control, formal monitoring,
reporting and surveillance protocols, etc. Explicit negotiations and agreements.
-Social order will mainly be self-organizing, spontaneous and informal, with
spontaneous and decentralized forms of control and of sanction. In this approach BIC
will play a crucial role. Sanctions like the act of excluding, avoiding the bad guys
will be messages…The act of monitoring the others behavior will be a message for
social order; the act of fulfilling commitments, obeying to norms, etc. will be all BIC
acts.
-We should design those systems allowing not only message sending but the
possibility for reciprocal observation of behaviors or of their outcomes and traces.
- the theory of ‘observability’ and the appropriate design of the ‘environment’ and of
‘coordination artefacts’ become crucial
EBICC-Campinas’09 - Castelfranch
friendly and natural HM interaction
• Machine collaboration and initiative (for example ‘Over-Help’)
should be based upon the possibility to “observe” and understand what we are
attempting to do, and in anticipating or correcting it.
• The same holds in Human-Robot interaction
It is not simply a matter of specialized and artificial messages (words or gestures); this
seems rather unnatural.
Also expressive NVC signals (faces, emotions) are not enough.
Before this one should provide the robot – for example for coordination with humans in a
physical environment –
 the ability to interpret human movements,
understand them and react appropriately.
At that point the human action in presence of the robot will be performed also for its
understanding, i.e. as a BIC message to it.
Analogously the human should be in condition to monitor what the robot is doing and to
intervene on it by adjusting its autonomy. At this point the robot behaviors – in front of the
human – might become message for approval, help, coordination, and so on.
EBICC-Campinas’09 - Castelfranch
The Big Problem:Social Order
how obtaining from local design and programming, and from
local actions, interests, and views, some desirable and relatively
predictable/stable emergent results.
- Emergent computation and indirect programming (Forest, 1990);
- reconciling individual and global goals;
- the trade-off between initiative and control; etc.)
Precisely because agents are relatively autonomous, act in an open word, on
the basis of their subjective and limited point of views and for their own
interests or goals, Social Order is a problem. There is no possibility for a predetermined, “hardwired” or designed social order.
formal
social control
SOCIAL
ORDER
informal
social control
orchestrated
selforganising
the invisible hand
EBICC-Campinas’09 - Castelfranch
COMMUNICATING without Meta-Communicating (without Grice)
Consider for example the habit of turning the lights on in one’s house
when one is going out. When a room is clearly visible from the
outside, leaving a light on is a signal that is left for a possible intruder
to mean that somebody is still at home. The light in itself has not a
conventional meaning but the possible inferences that can be drawn by
observing it are exploited to send a ‘deceiving’ message. Moreover, in
this case, although the real goal of the practical action is informative,
one does not want to be understood as communicating. This example
is also useful to stress that sometimes it is not even desirable that the
addressee understands that one is communicating. The light is
intended to be understood as a simple trace and not as a trace
intentionally emitted for some informative goal (the signal).
EBICC-Campinas’09 - Castelfranch