| Task: Define Agent Decisions based on Trust Values |
| |
 |
| Define which decisions agents make depending on the trust value of potential interaction partners. |
| Disciplines: Development |
|
Purpose
| For agents to make use of the trust values supplied by the trust values, their decisions should be based on the semantics
of the values. |
Relationships
| Parent Practices |
|
| Inputs | Mandatory:
| Optional:
|
| Outputs |
|
Main Description
Agents make a number of important decisions that can be based on trust values:
-
Which agent to interact with
-
Which precautions to take when interacting with an agent (such as checking the result of a computation, specifying
more exact conditions in the contract, etc.)
-
Whether to regard the result of an interaction in other decisions (e.g., when requesting data from sensors)
-
How the system will probably behave in the future
The system designer has to define how the agents will behave when confronted with different trust values in these
situations. Often, it will be possible to make binary decisions based on a threshold, but determining this threshold
can be non-trivial. In other cases, the exact semantics of the trust value need to be known to allow a staged
reaction in different situations and react appropriately.
The result of this task will be an updated agent architecture description that captures the choices and their relation
to trust values.
|
Steps
|
Identify agent decisions influenced by trust values
| Not all decisions an agent makes are influenced by trust values. Use the trust model and the requirements documents, as
well as the sources of uncertainty identified earlier to identify those decisions that need to be made under uncertainty
and that thus benefit from trust values. If decisions are identified that are not yet covered by the trust model, it might
have to be redefined or an additional trust model for a previously unidentified source of uncertainty has to be
introduced. |
Define the different choices an agent has in each decision
In many cases, agents only have a limited set of options to choose from in a decision. If that is the case, these
possible choices should be catalogued and their consequences and the circumstances under which they are useful should
be identified.
Should the decision have an arbitrary number of possible outcomes (e.g., when the agent has to decide on a value for a
continuous variable), identify the function that maps input data to the variable.
|
Map ranges of trust values or thresholds to the different choices
If a set of discrete choices has been found, relate the trust values to the choices by identifying how the trust value
influences the decision and which ranges of trust values warrant which choice. If additional factors are considered in
the decision, identify the correlation between trust values and these factors. It is often useful to define the
decision procedure formally to avoid ambiguity and to enable formal analysis.
In case a function is used for the decision, augment the function definition by including the trust value as an input.
|
|
|
