Trust, Control and Autonomy: A Dialectic Relationship - Trust Theory: A Socio-Cognitive and Computational Model - page 200

In sum, as for the dynamics of such a relation, we explained how:
X 's Control over Y denounces and derives from a lack of X 's trust in Y;
X 's Control over Y can increase X 's trust in Y;
X 's Control over Y increases X 's trust in deciding to delegate to Y (her global trust);
Control over Y by X can both increase and decrease Y 's trust in X ; in case that control
decreases Y 's trust in X , this should also affect X 's trust in Y (thus this effect is the opposite
of the second one);
X 's control over Y improves Y 's performance, or makes it worse;
X 's control over Y improves Y 's willingness, or makes him more demotivated.
7.1.11 Resuming the Relationships between Trust and Control
As we saw, relationships between trust and control are rather complicated. In this paragraph
(see also Figure 7.9) we resume the different role that control can play with respect to trust.
In fact, as shown in Figure 7.9 the control can increase or decrease and in both the cases we
can evaluate the potential influence on the two aspects of trust (strict and broad trust).
7.2 Adjusting Autonomy and Delegation on the Basis of Trust in Y
In this part we are going to analyze the complex scenario in which a cognitive agent (an agent
with its own beliefs and goals) has the necessity to decide if and how to delegate/adopt a task
to/for another agent in a given context. How much autonomy is necessary for a given task. How
could this autonomy be changed (by both the trustor and the trustee) during the realization
of the task. How trust and control play a relevant role in this decision and how important are
their relationships and reciprocal influences.
Autonomy is very useful in cooperation (why someone should have an intelligent collaborator
without exploiting its intelligence?) and even necessary in several cases (situatedness, different
competence, local information and reactivity, decentralization, etc.), but it is also risky because
of misunderstandings, disagreements and conflicts, mistakes, private utility, etc. A very good
solution to this conflict is to maintain a high degree of interactivity during the collaboration,
by providing both the man/user/client and the machine/delegee/contractor the possibility of
taking the initiative in interaction and help ( mixed initiative (Ferguson and Allen, 1998),
(Hearst, 1999)) and of adjusting (Hexmoor, 2000) the kind/level of delegation and help, and
the degree of autonomy run time.
We will analyze a specific view of autonomy which is strictly based on the notions of
delegation and adoption (Castelfranchi and Falcone, 1998). In fact, in several situations the
multi-agent plan, the cooperation between the delegating agent ( delegator ) and the delegated
one ( delegee ), requires a strict collaboration and a control flow between the partners, in order to
either maintain the delegator's trust or avoid breakdowns, failures, damages, and unsatisfactory
solutions.
Software and autonomous agents will not only be useful for relieving human agents from
boring and repetitive tasks; they will be mainly useful for situations where delegation and au-
tonomy are necessary (' strong dependence' , Section 2.9) because the user/client/delegator
does not have the local, decentralized and updated knowledge, or the expertise, or the