From: AAAI Technical Report WS-93-03. Compilation copyright © 1993, AAAI (www.aaai.org). All rights reserved.
A ! and Theories
of Groups and Organizations:
AAAI’93
Conceptual and Empirical Research
Workshop
Commitments:
from individual intentions to groups and organizations
Cristiano Castelfranchi
Project for the Simulationof Social Behaviour
Int. di Psicologia- CNR
viale Marx15 - 00137- Roma
- Italy
Email: PSCS@ IRMKANT.bimet
1. lntroductiveremarks
(Prelindnary Report)
Abstract
Theaimof these notes is to introducesomenotions of
Commitment
whichare crucial for the understandingof
groupsand organizations’functioning,andof the relations
betweenindividualagentsandcollective activity. I try to
identify someof the basic ingredientsof suchnotions, and
to makesomesteps towards their formalization. In
particular, I claim that a notion of Commitment
is needed
as a mediationbetweenthe individual and the collective
one. Beforeconstructinga notion of "Collectiveor Group
Commitment"
a notion of "Social Commitment"
is to be
defined. "Social commitment"is not an individual
Commitment shared by many agents; it is the
Commitment
of one agent to another agent. I stress the
normative
contents(entitlements/obligations)
of this social
relation, and its connections
withindividualintentionsand
collective activity. On that basis, a notion of
Organi~,ational Commitment
is proposed, that could
accountfor the structure of stable Organizations,and the
related notions of "role" and of "powerof command".
This
view,ba.q~don oneside on the intrinsic normativeaspects
of Social Commitment,and on the other side on the
dependenceand powerrelations that found a team, a
group, an Organization, is quite in contrast with the
"conversationar’ view of Organizationin whichagents
seemto be completelyfree to negotiate and establish any
.~rl of Commitment.
38
Thereis an implicit agreementabout Organizationsin
recent computationalstudies. Either in DAIthcorics of
organization [BON89]
[GAS91],or in formal thcories of
collective activity, teamor groupwork,joint intention,
and "social agents" [LEV90][RAO92], or in CSCW
approaches to cooperation [WIN87].Organization is
accounted for by means of the crucial notion of
"commitment"."Commitment"
is seen as the glue of the
group,of collective activity: it links the agent withthe
joint goal and the common
solution, it links members’
actions withthe collectiveplan, it links the members
with
eachother.
Unfortunately,the current analysis of Organizations
in
terms of Commitment
is quite unsatisfactory, for a number
ofreasons:
a) the current definitions of Commitment
are insufficient
to really accountfor stable groupconstitutionandactivity;
b) there is a dangerousconfusionbetweenthe notion of
"social" and the notion of "collective", so there is no
theory of "social commitment"
as a necessarypremisefor
a theety of collective or groupcommitment;
c) the relationships amongthe personalcommitment
to an
action (implied in the notions of "intention" and
"intentional act"), the group’s commitmentto the
collective act, and the commitment
of a memberto its
groupandthe collectiveactivity, are not clearlystated;
d) agents seem to be completely free (also
organizations) to negotiate and establish any sort of
Commitment
with any partner, without any constraint of
dependence
and ix)wcrrelations, of normsand procedures,
o[" pre-esulblishedplans and cooperationsICAS92b].
I wouldlike to discussthis problem
whichis crucial to
the relations between"AI and Theories of Groupand
Organization".
In particular, wewouldtry to analysea notionof "Social
Commitment"
as distinct on one side from the notion of
"Individual-Internal commitment",
and fromthe notion of
"CollectiveCommitment",
on the other side.
This notion of Social Commitment
is a relational one; it
cannotbe definedwithregardto one single agent (be it an
individual or a group);however,this notionis not purely
bchaviourist, like in [SIN92]:it is not reducible to the
individual commitments,but it implies individual
Commitment
and is analysedin terms of the mentalstates
of the panners.
On such a basis (and on the basis of the notion of
GenericCommitmen0
1 wouldpropose:
- a characterisation of the kind of Commitment
that
supportsthe structure of an Organization,including:
- the notionof "role";
- the claim that there is no Organization without
Obligations(norms);
- a criticism of the "conversational view" of
organization.
Such an approach is not only necessary for a good
definition of our concepts and for developinga formal
theory of groups, organizations, collective actions, but
also for someconsequencesin applicative terms. For
example, without such a kind of analysis it seems
impo~ibleto accountfor different kinds of organization
(e.g. strictly cooperative Vs orchestrated), and
consequentlyfor different kinds of commitment
(by role
or byreciprocation,free, basedon interest or benevolence,
etc.), andthen for different waysto solve the conflicts,
both intra- and inter- agents. For ex. howto intervene
whenone of the agents decides to abandonthe group, or
the common
plan ? Wecannot influence appropriatelythis
agent if we don’t knowexactly its kind of commitment
andthe related different reasonsfor defection[CAS93a].
2. Kinds of Commitment: Internal,
Collective
Social, and
Weneed a notion of Commitmentas a mediation
betweenthe individual and the collective one. A "social
39
commitment"
is not an individual Commitment
shared by
manyagents.
In general, it is absolutely necessaryto distinguish
between the notion of "social" and the notion of
"collective". Thisis not the case in DAIandMAS:
for ex.
to designatethe notion of a goal or intention shared by
manyagents and that they can not achieveindependentof
eachother, it is usedthe notion of "social" intention or
goal (ex.[WER88]);
to designate an agent formedby many
individual agents, i.e. a groupor a team,it is usedthe
notionof "social agent"(ex. [RAO92]);
to denotethe idea
of reciwocal commitments
within a team of agents it is
usedthe notion of "social Commitment"
and "social plan".
"Social"is not a synonym
of"collective".
There is a very important level of "social action",
"social agent" and "social mind"whereaction and mind
remain"individual" but they are oriented towardanother
social entity. At this level, beforeconstructinga notionof
"collective or group Commitment"
we need a notion of
"social Commitment":
the Commitment
of one agent to an
other agent.
So let mestep by step to distinguish among:internal.
social, collective Commitments.
INTERNALCOMMITMENT
(l.Commitment)
Internal Commitment
--as [BOU92]
calls it-- corresponds
to the Commitment
defined by Cohenand Levesque(on
the basis of Bramum’s
analysis) [COLI90].It refers to
relation betweenan agent andan action.
The agent has decided to do something, the agent is
determinedto execute a certain action (at the scheduled
time), the goal(intention)that waspreferredis a persistent
one. Thewayto capturesuch a persistenceis to establish
that the intention will be abandoned
only if and whenthe
agentbelievesthat the goalhas beenreached,or that it is
impossible,or that it is no longermotivated.
Theterm "Internal" is to be preferred to "Individual"
becauseone mayattribute l-Commitments
also to a group.
Theterm"psychological", as opposedto "social", is quite
misleadingbecause also the Social and the Collective
Commitments
are relations amongminds.
SOCIAL COMMITMENT
(S.Commitment)
As said aboveS-Commitment
is a relational concept, h
expresses a relation betweenat least two agents. More
precisely, S-Commitment
is a 4 argumentrelation:
(S-COMM
xyaz
-
(S-COMM
x y a) ==>(GOALy (DONEx
whois committed
(x);
whataction x is committed
to do (a);
whox is committed
to (y);
before whom
x is committed
(z)
1.et us neglect(in this paper)the third agent(z), i.e.
witness, whohas a very crucial role in normativecontexts
(normsefficacy) [CON93]and in contractual contexts
implyingalso free riders andcheaters. Let us concentrate
on the relation betweenx andy.
COIJ.ECTIVE COMMITMENT
(C-Conunitmen0
Weshould differentiate
S-Commitment from the
CoUectiveCommitment
or GroupCommitment.
The latter
is just an Internal Commitment
of a Collective agent or
Group.in other terms, a set of agents is Internally
Committed
to a certain intention and (usually) there
mutualknowledge
about that.
it remains to be clarified: a) if such conditions are
sufficient to accountfor the commitment
to a collective
activity; b) which are the relationships betweenSCommitment
and C-Commitment
(later).
So, x andy sharea goal (DONE
x a). Noticethat:
- this goalis an Intentionofx’s(giventhat it is his own
action);
- x and y havean identical goal, butxadoptedthis goal
from y (in our terms [CAS90][CON91],x has the
supergoal: (OBTAIN
y (DONE
x a))
b) lf x is S-Commined
to y, theny can(is entitled to):
- control ifx does whathe "promised";
- exact/require
that he doesit;
- complain/protest
withx if he doesn’tdo a;
- (in certain cases) makegoodhis losses (pledges,
compensations,
retaliations, ..)
I will comehacklater on this necessaryand veryrelevant
ingredient of the S-Commitment:
the rights of y on x
created by the S-Commitment
of x to y.
Let meintroduce, just for the sake of exposition, this
quasi-formal and incompletedefmition of S-COMM:
(MKxy ((INTENDx (DONE
x a))
(GOALy (DONEx a)))
&
(ENTITLED
y x as for a))
3. Towarda Definition of Social Commitment
S-Commitment
is muchmore than an l-Commitment
of x to a knownby y. Let us supposethat y is awareof
someintention of x to do an action a, nevertheless,this
does not allowto say that "x is committed
to y to do a".
Noris it sufficient to supposethat x is awarethat y knows
his intention, etc. (MutualKnowledge).Let megive
example:! haverealised that Johnintends to play a bad
joke on Paul, and Johnbecomesawareof the fact that I
understood
his intention; nevertheless,Johnpersists in his
plan, relying on mysilence. I maykeepsilent and become
party to his trick. Evenin this ease it wouldbe false that
"Johnis committed
to meto play a joke to Paul". if John
drops out his l-Commitment(that I know), I can
surprised,butI amnotentitled to protest.
in this situation twoqualifyingaspectsof the social
relation of "being committed
to somebody"
are lacking:
a) A social Commitment
is a formof "GoalAdoption".In
other terms:x is committed
to y todo a, if y is interested
in a. Theresult of a is one of the goals of y; for this
reason,y hasthe goalthat x doesa.
40
where ENTITLEDmeans that y has the rights o f
controllinga, of exactinga, of protesting(andpunishing).
ENTITLED
meansalso that x is S-Commitment
to y to not
oppose to these rights of y (in such a way,
"acknowledges"these rights of y). So, interestingly
enough,the definition of S-Commitment
seemsto require
a recarsivecall of anotherS-Commitment
of x itself.
4. Other Social Aspects of Social Commitment
I will considerjust three mainaspects:
a) Not all the adoptionsof a goal of y by x imply a SCommitment
ofx to y. Whatelse is required .9 First, the
Mutual KnowledgeI already mentioned. Second, y’s
agreement
In fact, if x has just the l-Commitment
to favouroneof y’s
goals (to do somethingso that (OBTAIN
y p) ), this is not
sufficient (even if there is common
awareness).Yshould
acccpt to (is l-Committedto) achieve her goal p by means
of x’s action. So, the S-Commitmentof x implies a SCommitmentof y to x to accept his action (y doesn’t
refuse, doesn’t protest, doesn’t say "who toldyouV,..).
Without such (often implicit) agreement (which is
reciprocal S-Commitmen0
no true S-Commitmentof x to
y ha.~ beenestablished.
b) The very act of committing oneself is a "rights.
producing" act: before the S-Commitment, before the
"promise"[SEA69],y has no rights over x, y is not entitled
(by x and before z ) to exact this action. After the
Commitmentit exists such a new and crucial social
relation: y has somerights on x, on the very basis of the
entitlement by x himself. The "title" of y is x’s act of
Commitment. So, the notion of S.Commitmentis well
definedonly if it implies these other relations:
- y is entitled to whatI already mentioned;
- x is in debt to y;
- x acknowledges
to be in debt to y and y’s rights.
In other terms, x cannot protest (or better he is committed
to not protesting}if y protests (exacts, etc.).
c) What I said just now implies also that if x is SCommitment
to y, he has a duty, an obligation, he ought to
do what he is Committedto.
(S-COMM xy a) ==> (OUGHTx (DONEx
So, whenx is commilted,a is morethen an Intention of x,
it is a special kind of goal, morecogent [CON93].
5. Relationships
Commitment
between Social
and Internal
S-Commitmentit is established through an overt or
tacit communicativeact. It implies the intention of x that
"y believes that x is S-Committed
to y to do a" and "llmt y
believes x is l-Commiuedto do a". This is because in SCommittinghimself x want that y intend to accept and to
control his behaviour(so that y’s knowledgeis a necessary
condition for x’s S-Committing).
This communicative
act has very relevant
consequences on the relations between S-Commitment
and l-Commitment. S.Commitmentmay be "sincere" or
"insincere": x lets y believe that he intends to do what y
4l
needs. If the communicative act and the S-Commitment
arc "sincere", x’s S-Commitment
to y to do a also implies
an l-Commitment
to do a. But. ff x is not sincere, if hc is
deceptive or just a light person, in this case he is not lCommiuedto do a.
it is really remarkable that even in that case, x is SCommittedto y to do a, he actually got an Obligation to
do a. So, x’s l-Commitment
on a is neither a necessary nor
a sufficient condition for his S-Commitment
on a. Just y’s
belief that x is l-Cuminto a, is a necessarycondition of x’s
S-Commitment.
Anyway,let me simplify. Let me consider that our
agents are always sincere (like in other models of
Commitment). That is why I putt Mutual Knowledge in
my temptative definition of S-Commitment.Given this
postulate, we mayremarkthat the S- Commitmentof x to y
to a, implies an l-Commitmentof x to a.
In our approach, it is possible to slate such a precise
relation between the two notions, because also the SCommitment
is analysed in terms of the menial states of
the partners. It is not a primitive notion (like in [SIN921),
but is not "reducible" to the I-Commitment,because it is
an intrinsically relational/social notion (amongagents),
and it contains muchmore than the I-Commitmentof the
involved agents.
I should also mention the Self-Commitment,which is
the S-Commitment
of an agent to itself (x=y), with the
same deontic consequences of the S-Commitment to
another agent. This kind of Commitmentis specially
interesting in humanpsychology, I mentionedit to avoid
confusion with the simpler I-Commitment.
6. Relationships amongInternal, Social, and Collective
Commitment
Is a true CollectiveCommitmentof a group of agents
just the sum of the l-Commitments of the members, or
does it require S-Commitmentsamongthose members?
! think that there is no univocal answer, it dependson the
kind and nature of the group (there are many kinds of
groups or collective agents [CON91
] [CAS93a]).
Commitments
in strictly "cooperative" groups
Let us considerfirst a true "cooperative"
group which
in our sense is based on a CommonGoal and Mutual
Dependence.
Moreprecisely, true cooperationis defined in terms of
MutualKnowledgeof MutualDependencerelative to an
Identicalgoalof x andy.
MutualDependence
is defined as follows: x Dependson y
for doing the action al relative to his goal p, and y
Ikpendson x for doingthe action a2 relative to her goal
p.
Anagent x Dependson an agent y for doingan action a
relative to a state p ICAS92a]
when:
- p is a goalof x’s
- x is unableto doa (andto realise p)
- y is ableto doso
- actiona is in a planusefulto achievep
If everybodyknowsto be dependenton the actions of
other agents, everybodywantsthat these others do their
share and wantsto do its own,for the common
goal. Then,
in a fully cooperative group, a S-Commitmentof
everybody
to everybody
arises: each one has to do its own
job. Giventhat the membersform the group, we maysay
that each memberis S-Committed
to the groupto do his
share[S IN92].
So, the Collective-Commitment
(defined as the ICommitment
of a collective agent) will imply(at least
the case of a fully cooperativegroup):
a) the S-Commitment
of each member
to the group: z is SCommitted
not simplyto anothermember
y, but to the all
set/groupXhe belongsto;
b) the S-Commitmentof each memberto each member
(includeditself); then .also manyReciprocalCommitments
and the SeifCommitments
of each member;
c) the l-Commitmem
of each member
to do his action.
Whois entitled to protest, giventhat S-Commitment
is
characterized by entitlements ? The group: each member
or the authorityin charge(if suchauthorityexists).
Commitment
and the notion of Collective Agent
To define a notion of Collective Agentjust on the
basis of the l-Commitment
(or of the individualintention),
of its sharing amongthe members,and on the members’
mutualknowledge,
is, in myview,a fallacious attempt.
Members
of a "social agent" -- in [RAO92]’s
terms -or of a group or "team" -- in [LEV90]’s -- have a
JointPersistentGoai,
that is, a realizable achievement
goal
associatedwitha mutualbelief that other members
of the
teamhavean equal goal and belief. Now,it is possible to
showthat this notion is not sufficient to accountfor a
teamworklCAS93b].
/42
Considerthis real life example:the case of a coupleof
people stranger to each other standing at the samebus
stop. It is an optionalstop for the driver (i.e. he will stop
only if people makesomesignal to stop). Theyknowthat
both are waiting for the same bus: so, they have an
identical intention ("to take the bus n" 3") and they
mutually knowthat. is this a sufficient basis for a
collective activity ? Dothey forma collective agent, a
team ? Unfortunately, each of them(becauseeach knows
the goal of eachother andthat it is identical withits own
goal) has got the expectation that the other does the
necessaryactions; each of themrelies uponeachother for
achieving the shared goal. The result is that the bus
arrives, nobodydoesthe signal, the bus doesn’tstop: both
of themlose the opportunity to reach the shared goal.
Whatis needed? It is neededan (implicit or explicit)
agreement about a commonactivity, based on the
awarenessof the reciprocal dependence
relations.
Let me give another crucial example. Prof.
Montaigner,of the Institute Pasteur in France,and Prof.
Galloin the USboth havethe final goal p "vaccineantiAIDS
be foundout" relative to the belief q that "if vaccine
is discovered,AIDSis wipedout". Theyshare all three
mental attitudes described [LEV90]as necessary and
sufficient conditionsfor a Joint Persistent Goaland then
for a team:
1) they mutuallybelievethatp is currentlyfalse;
2) they mutuallyknowthey all want p to eventually be
true;
3) it is true (and mutually knowledge)that until they
cometo believeeither that p is true, that p will neverbe
true, or that q is false, they will continue to mutually
believe that they each have p as a "weakachievement
goal"relative to q andwithrespectto the team.
Wherea "weakachievementgoal" with respect to a team
has beendefinedas "a goal that the status of p be mutually
believedby all the teammembers".
But no-onewouldstretch oneself up to saying that Prof.
Gallo and Prof.Montaignerform a team. Indeed, given
their "parallel goals"CI discoverthe vaccine"),they might
cometo strongly competewitheach other.
Whatelse is neededfog themto forma teamwogk?
Withoutthe belief about the mutualdependencealso
the Commitment
to participate, to do one’s ownshare is
unmotivated,it is irrational. Thebelief in an dependence
relation or in a "necessityto collaborate"is a fundational
condition for a really cooperativeworkand team[JEN92]
[CON91].
Now,mypoint is the following: if the members
acknowledge
their mutualdependence
and the needto
collaborate,
if consequently
they(implicitly or explicitly)
agreeabouta (already specified or to bc specified)
common
activity, havetheygot only an l-Commitmcnt
or
better havethey got S-Commitments
to the others ? Is the
S-Commitment
a necessarycondition for the constitution
of a CollectiveAgent?
Commitments
in other kinds of collective activity and
groupstructure
In myview, the current characterizationof the group
activity andcollective agentsis neither sufficient (as we
saw)nor necessary.In manykinds of natural group,team,
organizationspeopleparticipate in a collective activity
without sharing the samegoals or the final end of the
groupor organization.Norare they evenexpectedto have
such joint mentalstates. Let mementionjust an other
"cooperation" modell call "orchestrated cooperation"
[CAS92c].
Supposethere are three agents : a boss A, and
two executors B and C. Supposethat only A knowsthe
final goal of the coordinatedactivity he requires fromB
and C, only A knowsthe completeplan and the reciprocal
dependence
relations betweenB and C. It is evenpossible
that B and C ignore eachother. Thecooperativeplan is in
one and only one mind:the boss’. Betweenrespectively A
and B, and A and C, there ate S-Commitments
based on
social exchangerelations (not on true cooperation):they
are not interestedin the result of the planbut just in their
personal benefits (rewards). There is no S-Commitment
betweenB and C, whoin fact collaborate in a coordinate
way,and are members
of a team.
Evenin these cases we maysay that the group is ICommmitted
to do a (or to achieve a certain goal),
because this is the explicit plan and intention that
organizesthe action of the participantsanddeterminetheir
S-Commitments
and their l-Commintmcnts.
But, in these
case, the I-Commitmentof the group (C-Commitment)
doesn’t correspond to identical I-Commitments
of the
members.
llowever, there is an inportant constraint about the
relation betweenthe members’Commitments
and the CCommitment:
the formershould be "instrumental" to the
latter. In other terms, the S-Commitments
of the members
(and their consequent /-Commitments) should
Commitmentsto an action wich is part of the plan
(complexaction) the groupis C-Committed
to achieve.
43
7. Generic Commitment, and Structural
Organizational Commitment
or
A GenericCommitment
is a Commitment
to a class of
actions: x is Committed
to do any instanceof such a class
A,anyaction of that kind.
(GenericCOMM
x A) ffi for all a (wherea is an instance
of A) (COMM
x
TrueOrganizations
are not extemporary,
built up at the
moment.They are not made of Commitmentsto do a
specific action at a specific moment:so they use
GenericCommitments.
Moreprecisely, they are madeof
Commitments
to Commitoneself to do the right thing at
the right moment.
These Commitments
to Commitonese(f determinethe
"structure" of the Organization.Theyare different from
the running Commitments
involved in the structuring of
the collective activities of the Organization.So, the
structure of the Organization is different from the
structuresof its activities. Theformerpartially determines
thelatter.
TheseOrganizational Commitments
arc also Generic
SCommiunents:
(OrgCOMMxXA)
foralla(((REQUESTXx(DONExa))
~=>
~=> (S-COMM
x X (DONEx a))
wherea is an instanceof A, and x is a member
of X(set,
collective agen0.
Whenthe member
x of the groupX is organizationally
Committed
to his group, he is Committed
to accept
the
requests of the group(or of a specific member,e.g. the
boss) withina certain class of actions (his office). Then,
x’s OrgCommitment
to X implies that if there is a request
of Xto x aboutan action of the class A, x is automatically
S-Committed
to Xto this action, he automaticallygets an
obligationto do a.
The nature of the GenericCommitment
maygive rise
to controversies.In fact, there maybe different points of
view betweenx and X (or y) as to whether a specific
requestedactiona is or isn’t a member
of the class A.
The"Role"of an agent is relative to the groupor the
organization.
Itisthesetofthestructural
Commitments
of
themember
tohisgroup.
So,theorganizational
roleis
neithera functionalbehavior,nor a task or a set of ~,;ks. It
is a "normative" notion: it is a set of behavioral
obligations basedon the structural agent’s Commitments
to the grouprelative to certain classes of possibleactions
to berequestedor expected(tasks)
A GenericS-Commitment
of x to y to A (like the
OrgCommitment), gives to y a special "power of
influencing" x ICAS90I,a "Command
Power"over x. In
fact, x is committed
to complywith any requestofy as for
A: x is "benevolent"for special reasonstowardsy as for
A. Thispreventsy fromthe needfor negotiatingeachtime
x’s compliance with y’s requests. This kind of
"benevolence"based on such a Commitment-obligation
is
whatwecall "obedience".
This "Command
Power"gives y (or X) a very special
faculty: that of takingdecisionsfor x, of decidingthat x
does something.This possibility of deciding aboutother
agents’actionsis quite strange: y has an "intention"about
X’Saction.
Morethan this, y can take a S-Commitment
for x, she
can Commit
x to wto do something.Tobe moreprecise: y
S-Commits
herself to w aboutan action of x, or better to
require and to influence x to do this action; y is
"responsible"for x’s action. ThisS-Commitment
of y to w
implies an l-Commitment
of y to require/influencex to do
a; then, y will require x to do a. Given that x is
OrgCommitted
to y, from y’s request a S-Commitment
derives of x to y to do a. So, in fact, y has indirectly
Committed
x to do a.
This kind of Indirect Commitments,
as the Command
Powerand the faculty to decide the actions of other
people, are obviously a crucial and very well known
feature of Organizations. This feature dependson the
OrgCommitments
of the members,and on their being
boundby obligations
Let me,at this point, just mentiona criticismaboutthe
"conversational view" of Organization. In this view,
agents seem to be completely free to negotiate and
establish any sort of Commitment
with any partner,
withoutany constraint ofdependence
and powerrelations,
of normsand procedures, of pre-established plans and
cooperations.
44
Not only behavior in Organizations is bound by
"external" procedures, normsand rules, but even in a
purely contractual perspective --based on agents’
Commitments
and their mentalattitudes-- we sawthat the
agent is not free to establish his specific and extemporary
Commitments. lte is bound by his role (previous
OrgCommitments which are Generic) and by the
consequentexpectationsand duties. In this sense, there is
no OrganizationwithoutObligations.For this reasontoo,
the Commitments
an agent establishes in his cooperative
workin an Organization,are not all equivalent,nor canbe
handled in the same way. For example, it should he
distinguished whethera Commitment
is merely"personal"
(both by friendship or by social exchange)or it is "by
role", or it is "free" or it is "due" (by role or just by
reciprocation).
Conclusions
AI current interest in groupworkand Organizationis
basically motivatedby the fact that the introduction of
computersandtheir role in supporting andmediating
cooperation, absolutely requires a theoretical
understanding of humancooperative activity and of
Organizations.I think that AI is right in identifyingthe
core of this problemin the notion of Commitment
and in
the mentalrepresentationsof the agents. I tried to show
that currentnotionsare insufficient for accountingfor the
link between individual mind and collectivity. In
particular, I introducedthe intrinsically relational notion
of Social Commitment(which is neither mutual nor
collective) as an intermediary betweenthe Internal
Commitment
and the Collective one. 1 analysed many
relevant aspect of this social relation, with particular
attention to its normatives ingredient (obligations,
expectations).I think that a formalizationof these notions,
as well as of the notions of Organizational Commitment
and of Role, could help both in the theory of
Organizations and Groups, and in the computational
supportingof cooperativework.Myattempt has beenjust
to clarify someuseful concepts. Current views of
Organization
risktobetoo "subjective"
andtoobased
on
conununication.
Theyrisktoneglect,
ononeside,
the
objective
basis
ofsocial
interaction
(dependence
and
power relations), on the other side, its normative
components.The notions I proposedtry to relate also to
theselevels of analysis.
AcknowledgemenLs
I wouldlike to thankmyfriends of the PSCS
Project at IPCNR,Amedeo
Cesta, MariaMiceli and Rosaria Conte, for
our old discussionson this topic, and for their precious
comments
on these notes. Recklessnessand mistakes are
just mine.
References
[BON89]
Bond,A.ll.,
Commitments,
SomeDAIinsigths
^
fromSymbolic
Interactionist
Sociology,
9
International AAAIWorkshopon Distributed
Al~,
1989,
pp.239-261
[BOU92]
Bouron,
Th.,Structures
decommunication
et
d’organisation
pourlacooperation
dansun
univers
multi-agents.
, Thesede Doctorat,
Universite’
Paris
6,nov.
1992
C.,Social
Power.
A point
[CASP0I Castelfranchi,
missedin Multi-Agent,
DAI and HCI.In
Y.Demazeau
& J.P.Muller
(Eds.),
A.I.,
Amsterdam,
Elsevier,
1990(pp.49-62)
[CAS92a] Castelfranchi,C.,
Miceli,M., Cesta,A,
Dependence
Relations
amongAutonomous
Agents.
In Y.Demazeau
& J.P.Muller
(Eds.),
/2gggnll~liggg[.A,L~,
Amsterdam,
Elsevier,
1992
(pp. 49.62)
ICOII90| Cohen, P.R., Levesqoe, H.J., Intention is
Choice with Commitmentitment,
llllgUiggfl~, 42, 1990,pp.213-261
ICON9
! ]
Conte,R., Miceli,M.,Castelfranchi,C.,Limits
and Levels of Cooperation: Disentangling
various types of prosocial interaction. In
Y.Demazeau
&J.P.Muller (Eds.),
AJ.2, Amsterdam,
Elsevier, 1991(pp. 147-157)
[CON93]
Conte,
R.,Castclfranchl,
C.,Norms
asmental
objects.
Fromnormative
beliefs
tonormative
goals. "Reasnnin~about MentalStales: Formal
Theories & Applications’.
AAAISpring
~O~J~,~, March 23-25 1993, Stanford,
CA
[GAS91] Gasser, L., Social conceptionsof knowledge
and action: DAI foundations and open systems
semantics.Artificial Inlelli2ence, 47, 1991,pp.
107-138
[JEN92]
Jennings, N.R., Mamdani,
E.H., Using Joint
Responsibility to CoordinateCollaborative
Problem Solving in DynamicEnvironments,
Proceedingsof AAAI-92.
ILEV90] Lcvesque,J.H., Cohen,P.R.,Nunes,J.H., On
Acting Together, Prooeedin~sof AAhJ-9(},1990
[RAO92]Rao,A.S., Georgeff, M.P., Sonenberg, E.A.,
Social Plans: A preliminary Report. In E.
Werner &Y.Demazeau(Eds.), Decentralized
A.I.3, Amsterdam,
Elsevier, 1992
[CAS92b]Castelfranchi,C., Conte,R., Cesta,A, The
Organization
asa Structure
ofNegotiated
and
NonNegotiated
Binds.
In G.C.van
derVeer,
M.J.Tauber,
S.Bagnara,
M.Antalovits
(Eds.),
|luman Computer Interaction: Tasks and
Or e.’mizalionProceedingsof the Sixth European
Ctalferenccon Cognitive,Ergonomics~CCF_.6),
Balatonfured,llungary, September6-11, 1992
[SEA69] Searle, J.R., Speech Acts, Cambridge
UniversityPress, Cambridge,1969.
[CAS92c] Castelfranchi,C.,
Conte,R., Emergent
Functionality AmongIntelligent Systems:
Cooperation Within and Without Minds, A1&
Society. 6, 1992,pp.78-93
[WER88]Werner, E., Social Intentions,
Munich,WG,pp. 719-723,1988.
ICAS93a]Castelfranchi, C., Cesta, A., On Rational
Reasons for DroppingCommitmentsin Joint
Activities, HCAI1993. Workshop
on "Conflicts
mane~emen!
in cooperative problemsolving"
(submitted)
[CAS93b]
Castelfranchi, C., Cesta, A., Conte,R.,Miceli,
M., Foundations for Interaction:
the
DependenceTheory, AI*IA3/‘ Conference.
Torino, 1993(submitted)
45
1SIN921Singh, M.P., Social and Psychological
Commitmentitmentsin Multiagent Systems. In
"Knowledge and Action at Social &
Organizational Levels’. AAAIFall Symposium
Series. 1991, pp.104-106
ECAI-88,
DYINg7]Winograd,T.A.,LanguagelAction
perspective
ontheDesign of Cooperative
Work,llumanComnuler
Interaction.
vol.3,
I,pp.3-30,
1987