Trust, Control and Autonomy: A Dialectic Relationship - Trust Theory: A Socio-Cognitive and Computational Model - page 224

b) or they revise their reliance beliefs about: i) their dependence on Ag 2 , or ii) their preference
to delegate to Ag 2 rather than to do the job themselves, or to delegate to Ag 3 (a third agent)
or to renounce the goal;
c) or they changes their risk policy and is more or less likely to accept the estimated risk
(this means that the trustor changes either their set of utilities ( U(Ag 1 ,t 0 )) or their set of
thresholds. In other words, either Ag 1 's trust on Ag 2 is the same but their preferences have
changed (including their attitude towards risk), or Ag 1 has changed their evaluations and
predictions about relying on Ag 2 .
The modifications showed in the cases (a, b, c) might produce delegation adjustments but also
they could suggest to the trustor the introduction of control actions (either as monitoring or
as intervention). So the relationships among trust, control, autonomy, and delegation are very
complex and not so simple to predict: also the trustor and trustee personalities can play a
relevant role in these relationships.
7.3 Conclusions
As already shown, the relationships among Trust , Control and Autonomy are very complex and
interesting. Autonomy is very useful in collaboration and even necessary in several cases but
it is also risky - because of misunderstandings, disagreements and conflicts, mistakes, private
utility, etc. The utility and the risk of having an autonomous collaborator can be the object of a
trade-off by maintaining a high level of interactivity during the collaboration, by providing both
the trustor/delegator/client and the trustee/delegee/contractor with the possibility of having
initiative in interaction and help ( mixed initiative ) and of adjusting the kind/level of delegation
and help, and the degree of autonomy run-time. This also means providing channels and
protocols - on the delegator's side - for monitoring (reporting, observing, inspecting), and
for intervention (instructions, guidance, helps, repair, brake); and - on the delegee's side -
providing some room for discretion and practical innovation; for both client and contractor,
channels and protocols are needed for communication and re-negotiation during the role-
playing and the task execution.
Our model also provides a principled framework for adjusting autonomy on the basis of
the degree of trust and of the control's level of the trustor. In particular we have shown
that in order to adjust autonomy one should in fact adjust the delegation/help relationship.
Thus a precise characterization of different dimensions of delegation and of goal-adoption is
necessary. Moreover, we argued that adjustment is bi-directional (one can expand or reduce
the delegee's autonomy) and is bilateral ; not only the trustor or the delegator but also an
adaptive/intelligent delegee, the trustee (the 'agent') can change or try to change its level
of autonomy by modifying the received delegation or the previous level/kind of help. This
initiative is an additional and important aspect of its autonomy. We showed how trust, being also
the mental ground and counterpart of delegation, plays a major role in adjustment: limiting
autonomy is usually due to a trust crisis, while expanding autonomy is usually due to an
increased trust. Collaborative conflicts are mainly due to some disagreement about the agent's
trustworthiness.
We assume that this theoretical framework can also be useful for developing principled
systems.